10.1007@978 3 030 33229 7 PDF
10.1007@978 3 030 33229 7 PDF
10.1007@978 3 030 33229 7 PDF
Roberto Zagal-Flores
Cristian Barría-Huidobro (Eds.)
123
Editors
Miguel Felix Mata-Rivera Roberto Zagal-Flores
Instituto Politécnico Nacional Instituto Politécnico Nacional
Mexico City, Mexico Mexico City, Mexico
Cristian Barría-Huidobro
Universidad Mayor
Santiago, Chile
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Organizing Committee
General Chair
Miguel Félix Mata-Rivera UPIITA-IPN, Mexico
Academic Chair
Roberto Zagal-Flores ESCOM-IPN, Mexico
Local Manager
Jairo Zagal-Flores UNADM, Mexico
Posters Chair
Noe Sierra UPIITA-IPN, Mexico
Sponsors
ANTACOM A.C.
UPIITA-IPN
GIS LATAM
Collaborators
Alldatum Systems
Samani
Contents
Text and Image Transmission and Reception Using Light from LEDs
and a Light Sensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
Sergio Sandoval-Reyes
Mobile Application for the Support in the Learning of the Alphabet, Verbs
and Pronouns of the Mexican Sign Language Based on Augmented Reality . . . 183
A. Gordillo-Ramírez, O. Alonso-Cuevas, D. Ortega-Pacheco,
and U. Vélez-Saldaña
Internet of Things
Traffic Light Control Through Intelligent Agents, Sensors and Arduino . . . . . 258
René Alberto Ojeda-Cepeda, Larissa Jeanette Peniche-Ruiz,
Iván Antonio Castro-Ynurreta, and Edwin Moises Pool-Couoh
Informatics Security
Security Hierarchy for the NDN Network: Description and Experimentation . . . 335
P. David Arjona-Villicaña and Alejandra G. Silva-Trujillo
xiv Contents
Abstract. Propelled by the video game industries, virtual reality (VR) systems
are on the verge of entering the electronic mass media market. VR not just
allows generating imaginary worlds, but also visualizing real geographies and
possible past and future scenarios of real-world environments. However, VR
capabilities of current geographic information systems (GIS) are limited as yet,
thus challenging the transport of geodata into VR applications. Given the
shortcomings of GIS, the present article proposes a practical workflow for the
visualization of geographic information with immersive environments. Principal
work steps are 3D-modeling, composition, performance optimization and the
final dissemination on VR platforms. We will illustrate all elements of our
workflow using the example of an immersive VR application of a coral reef in
the Mexican Caribbean, thus providing both theoretical and practical back-
ground on the visualization of real-world data in VR space.
1 Introduction
1.1 Immersive VR
Recent literature on geographic information science (GIScience) documents an
increasing interest in the application of immersive virtual reality (VR) technologies for
geovisualization matters [1–3]. This interest is based upon the assumption that
immersive VR stimulates understanding through the formation of spatial presence,
defined as the sense of “being there” in a virtually mediated environment [4]. In line
with current research [4, 5], immersion can be considered a technological quality of
media that describes the extent to which an illusion of reality is delivered to the user.
Fig. 1. Workflow for geovisualization with immersive VR - principal work steps: (a) modelling,
(b) composition, (c) performance optimization and (d) dissemination
Real Geographies in Virtual Space 5
We cannot fail to mention that parts of the workflow shown in Fig. 1 have been
already discussed by other authors (e.g. [7]). However, given the spatial extension and
complexity of the underlying coral reef, this text will complement former work by
several fundamental elements (e.g. ecosystem composition, collective movement,
performance optimization).
2 3D-Object Modelling
2.1 GIS Base Data
Using basic features of points, lines and polygons, GIS data represents real-world
phenomena in a categorized manner and from a bird’s eye perspective. To generate
meaning, GIS users have to interpret these features, e.g. in accordance with the map
legend. However, in real space we rather see 3D objects than 2D features. Hence, in
order to visualize GIS-based data within an immersive VR environment, points, lines
and polygons need to be translated into the most important and (proto)typical objects of
the real-world environment being represented.
Moreover, these objects are not distributed in a neutral and flat map space, but
rather integrated into physical world’s topography – information typically managed
within digital terrain models (DTM) by GIS. In the present case of a Caribbean coral
reef, high resolution bathymetry information, derived from WorldView-2 data, was
available [8] (Fig. 2, left).
More arbitrary is the disaggregation of points, lines and areas into 3D models of
real-world elements. In our model case, classes of benthic habits (Fig. 2, right) were
translated into lists of species documented by the Mexican National Biodiversity
Information System [9] for each habitat within the study area shown in Fig. 2. Based
on both ecosystemic and aesthetic criteria, the most important species of these listings
were then selected for further 3D modelling (cf. Sect. 2.2).
6 F. Hruby et al.
Fig. 3. Object-based modelling of the Queen Conch (Lobatus gigas) with SfM - sample of input
scenes (left) and render of the resulting 3D model (right)
Fig. 4. Box-modelling (left) vs. sculpting (right) – exemplified by models of loggerhead sea
turtle (Caretta caretta) and black iguana (Ctenosaura similis)
2.3 Animation
In Sect. 1.1 we argued that – in order to generate an illusion of geographic reality in
VR space – users must perceive the virtual representation of a real place at a level of
detail as they would do being physically there. This means not just realistically
modelling all objects and agents (i.e. the computed objects that represent the real-world
creature) of a geographic environment but also to simulate their motion sequences, that
is to animate the 3D models created in the previous step (Sect. 2.2).
Skeletal Animation. Skeletal animation is a commonly used approach to convert
static into dynamic 3D objects [15]. As indicated by the technique’s name, a skeleton-
like structure is defined for each object, representing its main parts (e.g. head, legs, and
arms of a sea turtle) by so-called bones. Transforming position, scale and orientation of
each bone in a systematic, frequent and hierarchical manner, motion patterns of the
real-world referent can be represented (Fig. 5).
Fig. 5. Animation of the loggerhead sea turtle (Caretta caretta): Skeleton of the 3D model (left)
and scenes from a resulting animation sequence (right)
8 F. Hruby et al.
Vertex Animation. Directly controlling the movement of vertices rather than bones,
the vertex animation is another widely applied 3D computer animation technique
relevant to immersive VR environments. Vertex positions are stored for typical stages
of a motion sequence (e.g. the left- and rightmost position of a fish’s caudal fin
movement) and then interpolated between these stages during the animation process.
Morph animation and pose animation (especially used to animate facial expressions)
are two subtypes of vertex animation [16].
Having defined relevant GIS data sets (incl. a high-resolution DTM) and disaggregated
them into 3D models (Sect. 2), we can start blending these basic elements into a highly
immersive and realistic environment. While this blending process cannot be done with
the limited VR capabilities of current GIS, computer game engines (e.g. Unreal
Engine® or Unity 3D®) provide an alternative approach to close the gap between GIS
and VR. Game engines not just allow for a geographically referenced distribution of 3D
models, but also a realistic setting of ambience parameters. We shall discuss these main
steps at the composition stage of our workflow subsequently in further detail.
3.1 Topography
Since DTM contain 3D information (x/y position + z height) by definition, their
visualization in VR systems is a straightforward process whose results depend on
horizontal and vertical DTM resolution. Game engines can interpret topographical
information in several vector and raster formats. In the present project, a grayscaled
heightmap of the reef bathymetry (Fig. 2) was converted into a 3D mesh in Unreal
Engine (cf. Fig. 6).
distribution maps can, then, be used as a reference for allocating all 3D-models
correctly.
We can exemplify the process of static element distribution relying once again on
the model case of a VR coral reef. In a first step, information on the benthic habitats
shown in Fig. 2 was (temporarily) mapped onto the bathymetry mesh. Since each
habitat/color on the reference map had been already translated into 3D models of the
most typical species (Sect. 2), we now distributed these models upon the bathymetry
mesh by assigning each model to the accordant habitats/colors of the reference map.
Figure 6 illustrates a result of this process: Areas shown in red in Fig. 2 refer to
coral colonies. Since Elkhorn corals (acropora palmata) are typical of our study area,
they had been modelled in 3D and were then assigned to all areas labelled in red on the
reference map.
3.4 Ambience
The Oxford English Dictionary defines ambience as “[t]he character and atmosphere of
a place.” [18]. Character and atmosphere of a geographic place depend, for instance, on
lighting conditions, ambient sound and atmospheric parameters. Including these ele-
ments into an immersive environment will strongly affect whether a given VR scene is
perceived as realistic or not (cf. Fig. 6: left vs. right). Computer game engines provide a
rich set of tools to simulate ambience by land and sea, and even underwater.
In our model case, we simulated color and visibility range of Caribbean waters by
post-process-volume and exponential height fog functions in Unreal Engine. Under-
water caustics were visualized with an animated material of the principal VR light
source, while the effect of waves, e.g. on seagrass, was simulated by a virtual wind
function. Suspended solids and air bubbles were represented by particle systems.
Moreover, non-visual ambience settings were implemented via stereo scuba sounds
[19].
4 Performance Optimization
Complex immersive geovisualizations, e.g. the VR coral reef discussed throughout this
paper, are often constructed with millions of static objects and thousands of dynamic
agents. All these elements have to be rendered in real time and at a high frame rate onto
the HMD to sustain an illusion of reality. To guarantee these frame rates not just on
highly specialized VR systems, but also on consumer-level gaming PCs or even mobile
VR devices, several improvement measures can be implemented along the whole
workflow.
At the stage of object modelling, different techniques have been tested within the
present project (cf. Sect. 2). However, SfM and sculpting result in significantly higher
polygon counts than box-modelling, so that we preferred the latter over the former
where possible. In the case of objects rich in detail (Fig. 4), the 3D mesh was sculpted
in a first step, and retopologized afterwards in order to reduce the sculpted model’s
complexity while conserving its details. Additionally, all objects were generated at
different levels of detail (LOD) both regarding their polygon mesh and textures.
At the stage of dynamic agents, skeletal animation techniques provide higher
control of motion patterns but require considerable amounts of processing power, since
three parameters (rotation, translation and scale) need to be calculated per bone, per
object and per frame. Texture based vertex animations, on the other hand, are more
complex to define, but more flexible to integrate into advanced processes of ecosystem
composition. For instance, vertex animated models can be used as hierarchically
instanced static mesh components, where just a single object needs to be processed by
the GPU while several instances of this object can be rendered. We use this method to
simulate the shoaling behavior of fish.
A final step of optimization is the relative and/or absolute segmentation of the VR
environment. In the case of relative segmentation, only objects visible from the user’s
current position in VR space will be rendered at different levels of detail. Absolute
segmentation pre-divides the environment into regions the user enters or leaves.
Real Geographies in Virtual Space 11
Figure 8 illustrates this approach, which we adopted for the MXVR project:
Suppose that Fig. 8 represents VR space in a map-like manner. Suppose further, a
given user is virtually located in the central hexagon 1 (H1). Only within this hexagon,
both static and dynamic elements are rendered, while neighboring hexagons (H1a-f)
and hexagon spots (H2 and H3) remain empty, since they are beyond the user’s range
of vision. Imagine the user now moving from H1 towards H1c: As soon as the user
leaves H1, all objects of H1 will be eliminated from the PC’s working memory, and all
objects of H1c will be loaded and rendered as soon as this hexagon is entered. The
interspace between neighboring hexagons shown in Fig. 8 corresponds with the visi-
bility range within the VR environment to avoid abrupt (dis)appearance of 3D objects.
For the areas between hexagon spots (e.g. between 1a-f and 2a-f) only static elements
are rendered. Users can traverse these areas or switch between different hexagon spots
via teleports.
5 Dissemination
Moreover, a small information dashboard advises the users on their position, diving
depth and direction in VR space. Additional information on the species modelled is
provided via 3D icons giving access to fact sheets in audio and text format.
5.2 Release
Consumer analysis on VR systems indicates a still small but constantly growing
market, which is clearly dominated by the Oculus Rift® and HTC Vive ® platforms
[23]. Both HMD systems are broadly supported by game engine software. However,
when it comes about rendering stereo images on the fly, the actual bottleneck of
immersive VR is not so much the availability of a suitable HMD, but rather the
performance limits of current gaming GPUs. While these limits might be overcome – in
accordance with Moore’s law [24] – during the next years, streaming platforms provide
an alternative approach to deliver high resolution VR environments to a broad audi-
ence, shifting the aforementioned bottleneck from user’s hardware to digital
bandwidth.
The Oculus Rift HMD is the principal platform, the present project has been
developed for. NVidia’s GeForce GTX 1060® (or equal GPUs with at least 6 GB
memory) defines the minimum requirements to run the application fluently. For users
with lower GPU characteristics or no type of HMD at hand, also a low-immersive
desktop VR version of the MXVR project has been published. Both full- and low-
immersive applications are available online [25] under a CC BY-NC-ND 4.0 license.
Fig. 9. Non-stereoscopic (left: desktop) and stereoscopic (right: Oculus Rift®) renders from the
MXVR project (release candidate)
6 Conclusion
In this paper, we presented a practical workflow on how to convert GIS data into a
realistic immersive VR application. This workflow is just an early and only one pos-
sible way of bridging geographic and virtual reality. However, paths are made by
walking, and during this text we tried to take a few steps further to make this path.
These steps have proven feasible and effective (Fig. 9), but future improvements on our
workflow are foreseeable at all stages.
At the stage of geographic base data, next generation GIS may already provide
significantly higher VR capabilities, thus speeding up the information transport
Real Geographies in Virtual Space 13
between real and virtual space. ESRI’s CityEngine®, for instance, already offers plug-
ins to support GIS integration with game engines [26], indicating that GIScience may –
in line with our workflow – rather opt for an indirect connection between GIS and VR
via game engines than trying to make GIS directly compatible with HMDs.
At the stage of object modelling, currently common but time-consuming reference-
based techniques may be displaced by semi- and full-automated approaches like SfM.
Recently presented mobile devices with (stereo) pairs of cameras could convert SfM
based modelling into a common form of 3D data acquisition and communication.
At the stage of ecosystem composition, we mapped interactions between flora and
fauna of the MXVR project only at a very coarse level, e.g. in terms of path animations.
Implementing artificially intelligent (AI) behavior will allow us to realistically simulate
even higher order relationships, e.g. food chains or predator-prey-relationships, as well
as future or past environmental (e.g. climate change) scenarios [17].
At the stage of dissemination, increasingly powerful and affordable hardware could
further broaden the application range of VR technologies. For example, wireless HMDs
and also mixed-reality headsets (e.g. Microsoft Hololens®) may facilitate new forms of
interaction and interfaces between real and virtual space [27, 28], that is: between
human and computer.
References
1. Çöltekin, A., Oprean, D., Wallgrün, J.O., Klippel, A.: Where are we now? Re-visiting the
digital earth through human-centered virtual and augmented reality geovisualization
environments, 119–122 (2019). https://doi.org/10.1080/17538947.2018.1560986
2. Hruby, F., Ressl, R., De la Borbolla del Valle, G.: Geovisualization with immersive virtual
environments in theory and practice. Int. J. Digit. Earth 12(2), 123–136 (2019). https://doi.
org/10.1080/17538947.2018.1501106
3. Klippel, A., et al.: Transforming earth science education through immersive experiences:
delivering on a long held promise. J. Educ. Comput. Res. (2019). https://doi.org/10.1177/
0735633119854025
4. Skarbez, R., Brooks Jr., F.P., Whitton, M.C.: A survey of presence and related concepts.
ACM Comput. Surv. (CSUR) 50(6), 96 (2017). https://doi.org/10.1145/3134301
5. Cummings, J.J., Bailenson, J.N.: How immersive is enough? A meta-analysis of the effect of
immersive technology on user presence. Media Psychol. 19(2), 272–309 (2016). https://doi.
org/10.1080/15213269.2015.1015740
6. Hruby, F.: From third-person to first-person cartographies with immersive virtual
environments. Proc. Int. Cartogr. Assoc. 2, 44 (2019). https://doi.org/10.5194/ica-proc-2-
44-2019
7. Edler, D., Husar, A., Keil, J., Vetter, M., Dickmann, F.: Virtual Reality (VR) and open
source software: a workflow for constructing an interactive cartographic VR environment to
explore urban landscapes. Kartographische Nachrichten 68(1), 3–11 (2018)
8. Cerdeira-Estrada, et al.: Benthic habitat and bathymetry mapping of shallow waters in Puerto
Morelos reefs using remote sensing with a physics based data processing. In: IEEE
International Geoscience and Remote Sensing Symposium, pp. 4383–4386. IEEE, Munich
(2012). https://doi.org/10.1109/igarss.2012.6350402
14 F. Hruby et al.
9. Sarukhán, J., Jiménez, R.: Generating intelligence for decision making and sustainable use of
natural capital in Mexico. Curr. Opin. Environ. Sustain. 19, 153–159 (2016). https://doi.org/
10.1016/j.cosust.2016.02.002
10. Carrivick, J.L., Smith, M.W., Quincey, D.J.: Structure from Motion in the Geosciences.
Wiley, Chichester (2016)
11. Digital Life. http://digitallife3d.org. Accessed 11 June 2019
12. Bot, J.A., Irschick, D.J.: Using 3D photogrammetry to create open-access models of live
animals: 2D and 3D software solutions. In: Grayburn, J., Lischer-Katz, Z., Golubiewski-
Davis, K., Ikeshoji-Orlati, V. (eds.) 3D/VR in the Academic Library: Emerging Practices
and Trends, pp. 54–72. Council on Library and Information Resources, Arlington (2019)
13. Tran, M.H., Vu, H.M.Q.: A research on 3D model construction from 2D DICOM. In:
International Conference on Advanced Computing and Applications (ACOMP), pp. 158–
163. IEEE, Can Tho City (2016). https://doi.org/10.1109/acomp.2016.031
14. Knopf, G.K., Igwe, P.C.: Deformable mesh for virtual shape sculpting. Robot. Comput.-
Integr. Manuf. 21(4–5), 302–311 (2005). https://doi.org/10.1016/j.rcim.2004.11.002
15. Jensen, J.A., Burton, R.P.: Fourveo: integration of 4D animation into conventional 3D
animation workflows. Comput. Animat. Virtual Worlds 29(3–4), e1816 (2018). https://doi.
org/10.1002/cav.1816
16. Basori, A.H., Al Jahdali, H.M.A.: Emotional facial expression based on action units and
facial muscle. Int. J. Electr. Comput. Eng. 6(5), 2478–2487 (2016). https://doi.org/10.11591/
ijece.v6i5.12135
17. Magallanes Guijón, G., Hruby, F., Ressl, R., Aguilar Sierra, V., De la Borbolla del Valle, G.,
Rodríguez Paz, O.: Modelling of collective movement in immersive environments. In:
International Archives of the Photogrammetry, Remote Sensing & Spatial Information
Sciences, vol. XLII–4, pp. 397–402 (2018). https://doi.org/10.5194/isprs-archives-xlii-4-
397-2018
18. OED: The New Oxford Dictionary of English. Oxford University Press, Oxford (2001)
19. Hruby, F.: The sound of being there: audiovisual cartography with immersive virtual
environments. KN-J. Cartogr. Geogr. Inf. 69(1), 19–28 (2019). https://doi.org/10.1007/
s42489-019-00003-5
20. Wilk, M.P., Torres-Sanchez, J., Tedesco, S., O’Flynn, B.: Wearable human computer
interface for control within immersive VAMR gaming environments using data glove and
hand gestures. In: 2018 IEEE Games, Entertainment, Media Conference (GEM), pp. 1–9.
IEEE (2018). https://doi.org/10.1109/gem.2018.8516521
21. Ahn, S.J.G., Bostick, J., Ogle, E., Nowak, K.L., McGillicuddy, K.T., Bailenson, J.N.:
Experiencing nature: embodying animals in immersive virtual environments increases
inclusion of nature in self and involvement with nature. J. Comput.-Mediat. Commun. 21(6),
399–419 (2016). https://doi.org/10.1111/jcc4.12173
22. Duan, Q., Gong, J., Li, W., Shen, S., Li, R.: Improved cubemap model for 3D navigation in
geo-virtual reality. Int. J. Digit. Earth 8(11), 877–900 (2015). https://doi.org/10.1080/
17538947.2014.947339
23. Steam Hardware & Software Survey. https://store.steampowered.com/hwsurvey/Steam-
Hardware-Software-Survey-Welcome-to-Steam. Accessed 11 June 2019
24. Moore, G.E.: Progress in digital integrated electronics. Electron Devices Meet. 21, 11–13
(1975). https://doi.org/10.1109/N-SSC.2006.4804410
25. CONABIO Geovisualization. https://www.biodiversidad.gob.mx/v_ingles/region/geoviz.
html. Accessed 11 June 2019
Real Geographies in Virtual Space 15
26. Boulos, M., Lu, Z., Guerrero, P., Jennett, C., Steed, A.: From urban planning and emergency
training to Pokémon Go: applications of virtual reality GIS (VRGIS) and augmented reality
GIS (ARGIS) in personal, public and environmental health. Int. J. Health Geogr. 16, 7
(2017). https://doi.org/10.1186/s12942-017-0081-0
27. Keil, J., Edler, D., Dickmann, F.: Preparing the HoloLens for user studies: an augmented
reality interface for the spatial adjustment of holographic objects in 3D indoor environments.
KN-J. Cartogr. Geogr. Inf. 69, 1–11 (2019). https://doi.org/10.1007/s42489-019-00025-z
28. Satriadi, K.A., Ens, B., Cordeil, M., Czauderna, T., Willett, W.J., Jenny, B.: Augmented
reality map navigation with freehand gestures. In: IEEE Virtual Reality (2019, forthcoming)
Cross-Layer Architecture Applied
to Geospatial Information Communication
in Embedded Systems
1 Introduction
Currently, air navigation is mostly supported by satellite systems; among these systems
are the Global Positioning System (GPS), the European Geostationary Navigation
Overlay Service (EGNOS), Galileo and others [1]. However, data gathered from these
systems are useful for final users such as pilots and drivers, above is caused by the lack
of proper communication systems that allow sharing information in order to know
nearby vehicles position.
According to [2], communication is one of the most important aspects to get
success in an operation where many vehicles participate. Therefore, different mecha-
nisms, systems, and approaches have been studied to provide an effective data link
between vehicles. In [3] for software reconfigurable is proposed in order to rearrange
logically communications devices on board when one or more of these systems fail. On
the other hand, in [4] an additional communication system based on WiMAX Tech-
nology was proposed, in order to attain a decentralized and unstructured mechanism
ADS-B is divided mainly in two modules ADS-B OUT and ADS-B IN [9], first module
gathers information from GNSS or other Geospatial Information provider and then
transmit that information even if there are not nearby receivers; the second module
receives data and process it to be interpreted by a final user, in Fig. 1 general data flow
between vehicles is depicted, vehicles are going to be named nodes henceforth.
The Wi-Fi technology selection was based on the study presented in [10] where a
data link of 279 km was established as a result of this research, thence, in a smaller area
Wi-Fi turns into a more suitable alternative.
Fig. 1. Data Flow in an ADS-B system. A2A Comm does not need a centralized point forward
gathered data from GNSS.
For the ADS-B IN module, Wi-Fi adapter receives packages and the standard
802.11 treats them to the upper layer, socket listeners trigger the process to recognize
the received package and if it is not repeated security standard decipher the message. At
this point, and in the sender counterpart, the Advanced Encryption Standard (AES) [13]
is used for this purpose due to it belongs to symmetric key authentication mechanisms,
i.e., the receiver needs to have the same transmitter’s key (e.g., passphrase, shared
token or others), the above pursues bringing data privacy due to simple Ad Hoc
configuration does not provide authentication mechanisms such as WPA [14]. After
that, the actual information is obtained and finally, the software provides data as a
service that web application consumes to show on a web browser where digital car-
tography is deployed from a Geographic Information System (GIS). In Fig. 4 this
process is depicted.
20 L. A. Camacho Campos et al.
After implementation, for avoiding possible collisions, accuracy test was simulated,
above due to there is a delay between position acquisition and information displaying
which derives in position difference, one aspect to consider for avoiding crashes.
Although there are many variables to take into account such as reflected signals [18], a
time-based accuracy factor is described starting in (1).
Dt ¼ tf ti ð1Þ
Where tf is the time where data is displayed and ti is the time where GNSS data is
received, both actions take place in different nodes. For (1), clock synchronization is
needed, so UTC could be used but its accuracy is 1 s, thence, Dt could take a value of
0, in small areas or high dynamic scenarios. In order to obtain a difference estimation
f
Dt, (1) is rewritten like a sum of times involved since acquisition to display, (2)
describe this sum.
e ¼ tp1 þ tp2 þ te
Dt ð2Þ
de
et e ¼ ð3Þ
c
Based on (3), now distance accuracy approximation can be calculated, this distance
could be represented as a geofencing which alert pilots or drivers of nearby nodes.
Therefore (2) is rewritten and in (4) this is shown.
e ¼ tp1 þ tp2 þ et e
Dt ð4Þ
With (5) is possible to get radius estimation Ud, where s is the speed of the node. Also,
taking into account data described in Table 1 some estimation is given [4, 19, 20].
e ðsÞ
Ud ð DtÞ ð5Þ
Using JavaScript tools the average of tp1 and tp1 were settled in 0.012858 and
0.210177 respectively after 10 simulations from data gathering to displaying. Distance
is showed in meters and time in seconds, so Ud is shown in meters.
Based on Table 1, radius Ud is different for each vehicle, while speed increments a
major geofencing radius is necessary in order to increase accuracy. According to
related work, this paper presents an augmented fencing solution which increases
avoidance for possible collisions; also, this work provides a low-cost solution, unlike
satellite terminals proposed in [5], on the other hand, proposed system is added on,
however, proposed solution in [3] could be used jointly thought different communi-
cation interfaces.
Despite using the same principle of Ad Hoc network, this work offers an easy and
cheap implementation alternative and proposed solution augments traditional ADS-B
single-hop communication, due to nodes relay the received package, becoming in a
multi-hop communication system, and it gives wider area coverage than the traditional
one. The accuracy is a stochastic value because many factors impact it, however, the
presented time-based accuracy bias is lower than their WiMax counterparts presented
in the results of [4].
4 Conclusions
References
1. Groves, P.D.: Principles of GNSS, Inertial, and Multisensor Integrated Navigation System,
2nd edn. A. H, Boston (2013)
2. Mexican Air Force: private interview (2018)
3. Jayachandran, M., Manikandan, J.: Software reconfigurable state-of-the-art communication
suite for fighter aircraft. In: 2011 International Conference on Communication Systems and
Network Technologies, pp. 729–733 (2011)
4. Vengadesh, A., Gunasekaran, P.: Design and development of aircraft to aircraft commu-
nication by ad-hoc networks. In: 2015 International Conference on Circuit, Power and
Computing Technologies (2015)
5. Lee, L.-N., et al.: Micro satellite terminal-based high data rate communication for rotary
wing aircraft. In: 2014 IEEE Military Communications Conference, pp. 1698–1703 (2014)
6. Deepu, G.K. Sethuraman, N.R.: A study of various aircraft communication technologies. In:
International Conference on Communication and Electronics Systems (2017)
7. Federal Aviation Administration ADS Page. https://www.faa.gov/nextgen/programs/adsb/.
Accessed 15 Mar 2019
8. Amodu, O.A., Othman, M., Noordin, N.K., Idawaty, A.: A primer on design aspects, recent
advances, and challenges in cellular device-to-device communication. Ad Hoc Netw. 94,
101938 (2019)
9. Busyairah, S.A.: ADS-B system failure modes and model. J. Navig. 27, 995–1017 (2014)
10. Flickenger, R.: Very long-distance Wi-Fi networks. In: ACM SIGCOMM Workshop on
Networked Systems for Developing Regions (2008)
11. IEEE Standard for Information Technology. IEEE Standard 802.11 (2012)
12. NMEA data. https://www.gpsinformation.org/dale/nmea.htm. Accessed 02 Apr 2019
13. Crypto. https://nodejs.org/api/crypto.html. Accessed 20 Mar 2019
14. Kaur, R., Singh, T.P., Khajuria, V.: Security issues in vehicular ad-hoc network (VANET).
In: 2nd International Conference on Trends in Electronics and Informatics, pp. 884–889
(2018)
15. RPI 3 Model B Product. https://www.raspberrypi.org/products/raspberry-pi-3-model-b.
Accessed 10 Feb 2019
16. Zhou, H., Ni, L.M., Mutka, M.W.: Prophet address allocation for large scale MANETs. Ad
Hoc Netw. J. 1, 423–434 (2003)
17. GeoServer. http://geoserver.org. Accessed 01 Feb 2019
18. Yu, P., Yao, Y., Lou, M., Zhang, L.: Analysis and research of GNSS near ground signal
based on the dilution of precision. In: IEEE ICCC, pp. 1790–1794 (2016)
19. Excelsior. https://www.excelsior.com.mx/comunidad/2016/11/14/1128068. Accessed 29
Apr 2019
20. FAA. https://www.faa.gov/air_traffic/separation_standards/. Accessed 01 May 2019
21. Rathod, N., Dongre, N.: MANET routing protocol performance for video streaming. In:
2017 International Conference on Nascent Technologies in the Engineering Field (2017)
Methodology to Create Geospatial
MODIS Dataset
1 Introduction
Studies of earth observation are carried out with open-access optical satellite
data, highlighting the sensors AQUA-MODIS, TERRA-MODIS, and Landsat-
OLI [1–5]. Spectral and radiometric resolutions of MODIS sensor have been
sub-used since a few bands have been limited to the use of RGB images and
some indexes. For example, the MODIS sensor allows 38 bands with 12 bits
of radiometric resolution [6], which in principle mean 155648 possible values
for every pixel. Therefore, it is advisable to work with the data that composes
every band. Therefore, this study propose to build a dataset with the spectral
and radiometric MODIS data, to training, testing, and validate data science
algorithms. As a very first stage, to build the data set, it is necessary to define:
c Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 25–33, 2019.
https://doi.org/10.1007/978-3-030-33229-7_3
26 G. Álvarez-Carranza and H. E. Lazcano-Hernández
Fig. 1. Study Area: Represent by 115 pixels of 1 km2 (red squares) along some sections
of the coastal zone of Quintana Roo. The challenge is to extract the data from the
different MODIS bands for the pixels of interest. (Figure credit: Holger Weissenberger).
(Color figure online)
In this stage, a list of latitude and longitude of every study pixels was built. The
list built in this stage allows knowing which data will be extracted from the files
in the processing stage. The visualization, selection, and ordering of the data
were done with SeaDAS. To validate the pixels position, Google Earth Pro [14]
software was used. Once the pixels have been selected and validated the file was
converted into a “.csv” format and finally downloaded. Figure 2 shows the pixel
selection and validation stage.
The file and pixel information generated in the previous stages feeds process-
ing stage. To automate the processing for the study, a software tool developed
at home was used (L2-JEP-MODIS-Processor)2 , which uses the SeaDAS soft-
ware as CORE. During this stage, the raw files go through several levels of
processing. The original raw data is re-projected, radiometric and atmospheric
2
INDAUTOR registry number: 03-2019-032810002000-01.
28 G. Álvarez-Carranza and H. E. Lazcano-Hernández
Fig. 2. Study Area: (a) 115 selected pixels using SeaDAS. (b) Approach to the southern
area close Mahahual using SeaDAS. (c) Approach to the southern area close Mahahual
using Google Earth.
Fig. 3. Sample of Aqua-MODIS imagery used to build the dataset. rhos product at
different wavelengths: (a) 412 nm. (b) 469 nm. (c) 555 nm. (d) 645 nm. (e) 859 nm. (f)
1240 nm. (g) 2130 nm and (h) Reference map.
Fig. 4. Pixels from 718 to 728 were shown, every pixel is related to a specific area.
For every file, the values related to every pixel could be different. The pixel values for
different files (dates) feed the dataset.
30 G. Álvarez-Carranza and H. E. Lazcano-Hernández
Fig. 5. Presence and absence of Sargasso along the coast line is depicted in a and b
respectively. The difference in the color between the sea with presence and with the
absence of Sargasso is indeed the difference in reflectance between both conditions and
that’s just what the sensor measures. (Photographs credit: Hugo Lazcano) (Color figure
online)
Using this methodology, the first dataset for Quintana Roo coastline with
rhos and rhot reflectance values was built. The dataset values are labeled with
presence/absence of Sargasso between 2014 and 2018 with latitude and longi-
tude data for each pixel (from 21.496124 Latitude, −87.546677 Longitude, to
18.477211 Latitude, −88.293625 Longitude), bordering the coast of Quintana
Roo. The area is located where massive arrivals of Sargassum were recorded in
2015 and 2018. This dataset allowed us to train, test and validate the ERISNet
algorithm, which is an algorithm based on deep learning techniques that allow
assigning a probability of presence/absence of Sargasso to each of the pixels used
for the study. ERISNet was successful and finally published [10].
Clases 2 Presence/Absence
Wavelengths 7 412, 469, 555, 645, 859, 1240 and 2130 nm
Attributes 14 7 rhos and 7 rhot
Instance with Sargasso 2306 Reflectance values with Sargasso
Instance no Sargasso 2209 Reflectance values without Sargasso
Total instance 4515 Total rows
Total data 63210 Total reflectance values
The Dataset3 is available for download and can be used for training and
testing other types of algorithms. The dataset is a multivariable one, real and
useful for classification. Additional features are shown in Table 1.
3
DOI: 10.7717/peerj.6842/supp-1.
Methodology to Create Geospatial MODIS Dataset 31
The built dataset was used successfully in the training, testing, and validation
of ERISNet, an algorithm based on convolutional and recurrent neural networks
to classify pixels with presence/absence of Sargasso [10]. Figure 6 shows a dataset
example of some values corresponding to rhos.
As summary Fig. 7 shows some details of every stage of the proposal method-
ology: (1) selection and download of files (stages with light yellow fill), (2) selec-
tion and labeling of pixels (stages with light rose fill) and (3) processing (white).
The software “Simple web crawler for MODIS data ver 1.0” and L2-JEP-MODIS-
Processor were developed at home using batch and python (Anaconda 3) lan-
guages. Lenovo Workstation (with Intel Xeon EP processor, 64 GB of RAM and
Ubuntu 18.04 64 bits) was used for software developed.
32 G. Álvarez-Carranza and H. E. Lazcano-Hernández
Fig. 7. Process Flow Diagram: Selection and download of files (light yellow), selection
and labeling of pixels (light rose) and processing (white) (Color figure online)
4 Conclusions
This study shows the methodology to build a geospatial dataset using as input
Aqua-MODIS data. The dataset it is valuable in itself. Therefore was registered
in INDAUTOR with the name of CaribeMX-Dataset4 . The dataset is composed
of reflectance values at seven different wavelengths (412, 469, 555, 645, 859, 1,240
and 2,130 nm.) from two different MODIS products (rhos and rhot). According
to fieldwork the dataset values was labeled as Sargasso or No Sargasso (2306 rows
related to presence of Sargasso and 2209 rows related to absence of Sargasso).
Finally the dataset has 63210 values.
4
INDAUTOR registry number:03-2019-010810340300-01.
Methodology to Create Geospatial MODIS Dataset 33
The built dataset was used successfully in the training, testing, and valida-
tion of ERISNet algorithm. The data set is available for their download and
can be applied to other studies. The proposed methodology can be used in the
construction of different dataset from other areas and for other objectives.
According to the data source and the final application, the method to build
geospatial dataset could be different. There is no better method than another,
the study emphasizes the use of open-source data and tools.
References
1. Nathan, F.: Simulating transport pathways of pelagic Sargasso from the Equatorial
Atlantic into the Caribbean Sea. Prog. Ocean. 165, 205–214 (2018). https://doi.
org/10.1016/j.pocean.2018.06.009
2. Schell, J.M., Goodwin, D.S., Siuda, A.N.: Recent Sargassum inundation events in
the Caribbean: shipboard observations reveal dominance of a previously rare form.
Oceanography 28(3), 8–11 (2015)
3. Gower, J.: Satellite images suggest a new Sargassum source region in 2011. Remote
Sens. Lett. 4(8), 764–73 (2013)
4. Wang, M.: Mapping and quantifying Sargassum distribution and coverage in the
Central West Atlantic using MODIS observations. Remote Sens. Environ. 15(183),
350–67 (2016)
5. Wang, M.: Predicting Sargassum blooms in the Caribbean Sea from MODIS obser-
vations. Geophys. Res. Lett. 44(7), 3265–73 (2017)
6. Garcı́a-Mora, T., Jean-Francois, M.: Aplicaciones del sensor MODIS para el mon-
itoreo del territorio, 1st edn. Secretarı́a de Medio Ambiente y Recursos Naturales
(Semarnat), México (2011)
7. SeaDAS. https://seadas.gsfc.nasa.gov/. Accessed June 2019
8. Lance MODIS. https://lance-modis.eosdis.nasa.gov/. Accessed June 2019
9. Ocean Color. https://oceandata.sci.gsfc.nasa.gov/MODIS-Aqua. Accessed June
2019
10. Arellano-Verdejo, J.: ERISNet: deep neural network for Sargasso detection along
the coastline of the Mexican Caribbean. PeerJ, 7, e6842, 1–19. (2019). https://doi.
org/10.7717/peerj.6842
11. MODIS-Aqua, M.-A.O.C.: NASA goddard space flight center, ocean ecology lab-
oratory, ocean biology processing group. Moderate-resolution imaging spectrora-
diometer (MODIS) Aqua l0 Data; NASA OB.DAAC, Greenbelt, MD, USA (2018).
Accessed 01 Nov 2018
12. Van Tussenbroek, B.: Severe impacts of brown tides caused by Sargasso spp. On
near-shore caribbean seagrass communities. Mar. Pollut. Bull. 122(1–2), 272–281
(2017)
13. Rodrı́guez-Martı́nez, R.E.: Afluencia masiva de sargazo pelágico a la costa del
Caribe mexicano (2014–2015). In: Florecimientos algales nocivos en México,
CICESE, Ensenada, Mexico, pp. 352–365 (2016)
14. Google earth. https://www.google.com/intl/es/earth/. Accessed June 2019
Data Acquisition System for Solar Panel
Analysis
Abstract. This project introduces a data acquisition system for solar panel
technologies, mainly for analysis and report purposes. The measured variables
are the current and voltage generated by the panel so that the power and voltage
curve can be plotted to analyze. The system is made up of two parts, the
hardware that will be designed using an Arduino circuit board with a voltage
sensor, a current sensor and a temperature sensor that feedback the modeled
circuit based on Matlab and Simscape platforms. As an addition to the hardware
system, an embedded system is designed using operational amplifiers to simu-
late a solar panel, this emulation is based on the mathematical model that gives
our values for variables generated by solar technologies. For simulation of the
modeling and plotting of the generated data, we propose the use of Wahba’s
method, for smoothing noisy data using spline function. This model will serve
as a research/experimentation tool for technologies that improve the efficiency
range of a solar cell.
1 Introduction
In the last 15 years, there has been an unprecedented change in the consumption of
energy resources. The high growth in the renewable energy market, in terms of
investment, new capacities and high growth rates in developing countries, has changed
the landscape for the energy sector. We have seen the growth of unconventional
resources and improvements in the evolution of technology for all forms of energy
resources [1]. In particular, solar energy has proven to meet up this energy scenario [2].
Solar cells, also called photovoltaic (PV) cells, convert solar radiation directly into
electricity. They capture photons by exciting electrons through the band gap of a
semiconductor, which creates electron pairs and holes that are then separated by
charge, usually by PN junctions introduced by doping. The bonding interface conducts
the electrons in one direction and the holes in the other, which creates a potential
difference in the external electrodes equal to the band gap.
The concept and configuration are similar to those of a semiconductor diode, except
that the electrons and holes are introduced into the junction by the excitation of the
photon and are removed at the electrodes [3]. Currently solar cells reach a wide variety
of applications and contribute to the massive or isolated generation of electrical energy.
Crystalline silicon is a well-studied material and its technology is mature, it is the most
common material used in Photo Voltaic (PV) technologies, the global production of
solar cells increases rapidly and the amount of silicon used exceeds the set of semi-
conductor devices such as diodes, transistors and circuits.
The huge gap between the potential of solar energy and its use is due to the cost and
conversion capacity. Fossil fuels satisfy our energy demands at lower cost than solar
alternatives, in part because fossil fuel deposits are concentrated sources of energy,
while the sun distributes photons quite evenly on Earth at a more modest energy
density. The use of biomass as a fuel is limited by the production capacity of the
available land and water. The limitations of cost and capacity in the use of solar energy
are addressed more effectively with a single research objective: to increase conversion
efficiency in a cost-effective manner.
Several mathematical models have been developed that describe the behavior of a
solar cell. All the models are based on the current-voltage ratio resulting from a diode,
whose manufacture represents the n-type and p-type doped semiconductors, used in the
manufacture of a solar cell [4]. These models can be used in the design and devel-
opment of data acquisition (DAQ) systems which measure physical phenomenon such
as voltage, current, temperature, pressure, or sound converting them into digital signals.
These signals in turn, are processed and conditioned by a computer with programmable
software to allow their easy analysis and visualization.
In this work a circuit is designed in Simscape to simulate the behavior of a real PV
panel. A sensor board was constructed to collect the current, voltage and power values
generated by the photovoltaic panel. The DAQ system was completed by a computer
code programed to process and smooth the signals into a format that could be used for
analysis purposes.
circuit of a solar cell consists of a photo-current source generated by the action of solar
radiation, a diode, a resistance in parallel and a shunt resistance (representing the
internal resistance of the material to the flow of the current), as shown in Fig. 1.
Applying Kirchhoff’s law to the representation of a solar cell as seen in Fig. 1. The
next equation can be obtained.
Where IPH is the photo-generated current, ID is the current of the diode, ISH is the
shunt resistance loss in parallel and I is the output current of the cell. The first term can
be determined from next equation and depends on the irradiance, area and temperature
in which the cell is being exposed.
G
IPH ¼ ½A JSC þ aJSC ðT TSTC Þ ð2Þ
GSTC
where G is the irradiance value under given conditions, GSTC is the irradiance value
under standard conditions, A is the area of the cell, JSC is the short-circuit current
density of the cell, aJSC is the temperature coefficient of the short-circuit current
density, T is the temperature under given conditions, TSTC is the temperature under
standard conditions.
The second term from Eq. (1) is the current of the diode which can be determined
from the Shockley equation.
VD
ID ¼ IO enVT 1 ð3Þ
Where IO is the saturation current, VD is the voltage of the diode, n is equal to the
ideality factor of the diode and VT is the thermal voltage. From Eq. (3) the diode
saturation current is obtained, which is that part of the reverse current in a semicon-
ductor diode caused by diffusion of minority carriers from the neutral regions to the
depletion region
JSC A VROC
IO ¼ VOC SH ð4Þ
e 1
nVT
Where JSC is the short-circuit current density of the cell, A is the area of the cell,
VOC is the open circuit voltage of the cell, RSH is the resistance in parallel, VT is the
thermal voltage and n is equal to the ideality factor of the diode.
The third term from Eq. (1) is shunt resistance loss current which corresponds to
the current leakage is usually suppressed in some models because the parallel resistance
(RSH) usually has a very high value with respect to the series resistance.
Data Acquisition System for Solar Panel Analysis 37
V þ RS I
IRSH ¼ ð5Þ
RSH
Where V is the output voltage of the solar cell, RS is the series resistance, RSH is
the parallel resistance and I is the resistance current in parallel.
Equations 2, 3 and 4 are substituted in Eq. 1, so the general modeling equation for
a solar cell is obtained.
V þ RS I V þR I
S
I ¼ IPH IO e nVT 1 ð6Þ
RSH
A solar panel is formed by grouping cells in series, in order to obtain the output
current (I) of a solar panel the output voltage (V) in the Eq. 6 is divided by the number
of cells (Ns) [5, 6].
!
V þR I V
NS S
NS þ RS I
I ¼ IPH IO e nVT
1 ð7Þ
RSH
3 Data Validation
To represent the mathematical equation and to model a solar panel, the Matlab and
Simulink platform tools are used, this model is visualized as a set of subsystems
connected and configured to calculate the variables needed to be analyzed.
Inside the block that represents the PV module, the values and elements from Eq. 7
are modeled in subsystems Fig. 3.
The fundamental parameters related to solar cell are short circuit current (Isc), open
circuit voltage (Voc), efficiency of solar cell and fill factor. These values are measured
using the hardware circuit built in an arduino board using voltage, current and tem-
perature sensors. Short Circuit Current is the condition when the impedance is low and
it is calculated when the voltage equals to zero.
Isc occurs at the beginning of the forward-bias sweep and is the maximum current
value in the power quadrant. For an ideal cell, this maximum current value is the total
current produced in the solar cell by photon excitation. Isc is the short-circuit current of
cell at 25 °C and an Irradiance 1000 W/m2, Iph is the light generated current directly
proportional to the solar irradiation, Io is the diode saturation current of the cell varies
with the cell temperature, Irs is the reverse saturation current of a cell at a reference
temperature and a solar irradiation, Eg is the band gap energy of the semiconductor
used in the cell and Ish stands for Parallel resistance loss current, this term of the
equation that corresponds to the current leakage is usually suppressed in some models
because the parallel resistance usually has a very high value with respect to the series
resistance [6].
Modeling the Solar Panel in Simscape. To Model the PV cell in Simscape involves
the estimation of the I-V and P-V characteristics curves to emulate the real cell under
various environmental conditions. The most popular approach is to utilize the electrical
equivalent circuit, which is primarily based on diode. The Solar Cell block in
Data Acquisition System for Solar Panel Analysis 39
SimElectronics Fig. 4. represents a single solar cell as a parallel current source, Iph, an
exponential diode, D and a shunt resistance, Rsh that are connected in series with a
resistance Rs.
The Solar panel circuit is modeled using elements from Simscape and Simulink.
Solar cell was connected with blocks such as current sensor and voltage sensor to
measure the current and voltage across solar cell. Blocks for varying Irradiance level
and Temperature, were also connected in the above model. Rest of the blocks were the
interface between major blocks and used for plotting I-V characteristics (Fig. 5).
This model circuit works with the data we acquired from the readings of our
experimental solar cell at a lab.
Parameter Estimation Methodology. Smoothing splines are well known to provide
nice curves which smooth discrete, noisy data, in this case it is proposed a system based
on parameter estimation theory which will allow the system to filter measurements
which can be considered error. In signal processing, an autoregressive (AR) model is a
representation of a type of random process; as such, it is used to describe certain time-
varying processes, the model specifies that the output variable depends linearly on its
own previous values and on a stochastic term.
The AR model is a widely used model to understand time series data. Traditionally,
the innovation noise of the AR is modeled as Gaussian. However, many time series
applications, for example, situations arising in applications of signal processing where
40 H. De la Rosa et al.
the time series are non-Gaussian and heavy-tailed, either due to intrinsic data gener-
ation mechanism or existence of outliers, the AR model with more general heavy tailed
innovations is preferred.
Another issue that frequently occurs in time series is missing values, due to system
data record failure or unexpected data loss. In the recent era of data deluge, many
applications collect and process time series data for inference, learning, parameter
estimation, and decision making. The AR model is a commonly used model to analyze
time series data, where observations taken closely in time are statistically dependent on
others. In an AR time series, each sample is a linear combination of some previous
observations with a stochastic innovation.
An AR model of order p, AR(p), is defined as.
Xp
yt ¼ u 0 þ i¼1
ui yti þ et ð8Þ
1 h Xn 2
i
f ð xÞ ¼ l ð MesEstimated v mes ð jÞ Þ ð9Þ
2 j¼1
This paper defines the evolution of the value of the current density in the proposed
function f ð xÞ, so that the minimum error is chosen, but it takes in account the prediction
of the state x and the coefficients of value for the state l and the measures estimated
(MesEstimated = MS) at the instant k.
The motivation for this cost function is that if the vectors are error-free and the true
matrix f ð xÞtrue, is assumed to be the same for all the measurements, then
MesEstimated is equal to vmes trueð jÞ for all j and the cost function is equal to zero,
which is clearly the minimum value, for f ð xÞ equal to f ð xÞtrue.
So for this method we define f ð xÞ as our current density Ix and adapted it to get the
next formula.
1 h Xn 2
i
Ix ¼ l j¼1
ð I est I mes ð j Þ Þ ð10Þ
2
Data Acquisition System for Solar Panel Analysis 41
Where IX stands for the parameter estimation in which the optimal values con-
sidered for each test are, Iest for the estimated current values, Imes is for the current date
measured and n will stand for the number of readings we get each in a minute [8].
4 Results
In this section, part of the simulation and results are represented in order to show the
performance of the proposed system, from our readings captured by our circuit, this
data is inserted in the Simscape modeling.
The readings obtained from the test are fluctuating and represent a certain percent
of noise and error; we must determine the optimal value to get the plot needed to
analyze Fig. 6.
Fig. 6. I-V and power curve plotted from variable sample reading in the laboratory
Applying the Eq. 10 in each of the sampling readings from our test the optimal
value can be determined with the minimum range of error.
As presented in the Table 1 the values obtained for ISC are calculated as the optimal
and with less percentage of error.
Fig. 7. Percentage error obtained by our optimal value compared to the highest and lowest
values of variation in the test.
This percentage error calculated using the value for short circuit current stabilizes
rapidly in the range of 1% which can be interpreted as an acceptable range; the test
results we registered have an acceptable percentage error Fig. 7. This data reading
represents the values of a single solar cell which has been modeled and plotted Fig. 8.
Fig. 8. Plotting for the I-V and power curve obtained from experimentation data for a single
solar cell.
To get the power curve we have programed a code that calculates the value using
Watt’s law, so when the plotting process in Matlab starts both curves are sent to us to
analyze.
The simscape model we designed can be modified to represent a complete solar
panel, the element we configured with our readings can be connected in series or
parallel to represent the model of Solar panel we desire Fig. 9.
Data Acquisition System for Solar Panel Analysis 43
Depending of the number and connection type of the solar panel, a subsystem
configuration can be created to represent the PV module needed to analyze Fig. 10.
This way, the PV panel designed can be analyzed in solitary, or add more panels to
represent a PV array, the curves obtained from the readings in this solar panel can be
saved for report or analysis Fig. 11.
44 H. De la Rosa et al.
Fig. 11. I-V curve and power curve for a photovoltaic panel.
5 Conclusion
This paper has analyzed the development of a method for the mathematical modeling of
PV arrays, using a simulation platform and an embedded system which simulates these
systems.
The method obtains the parameters of the I–V equation by measuring the values
generated by our solar panel, which are the open circuit voltage, short-circuit current,
and maximum output power. To accomplish our main objective which is to reduce
error and represent a clean and smooth curve for analysis and report purposes, we used
the cross validation method proposed with Wahba’s problem.
Acknowledgements. The authors gratefully thank the financial support of the Secretaria de
Investigación y Posgrado del Instituto Politécnico Nacional (SIP-IPN).
References
1. World Energy Council Home page. https://www.worldenergy.org. Accessed 2019
2. Kirmani, S., Jamil, M., Akhtar, I.: Effective low cost grid-connected solar photovoltaic system
to electrify the small scale industry/commercial building. Int. J. Renew. Energy Res. 7(2),
797–806 (2017)
3. Matsumoto, Y.: Celdas Solares De Silicio: Fundamento y Tendencia. Caos Conciencia 5, 32–
44 (2011)
4. Crabtree, G.W., Lewis, N.S.: Solar energy conversion. Phys. Today 60, 37–42 (2007)
5. Villalva, M., Gazoli, J., Filho, E.: Comprehensive approach to modeling and simulation of
photovoltaic arrays. IEEE Trans. Power Electron. 24, 1198–1208 (2009)
6. Granda-Gutiérrez, E.: Modelado y simulación de celdas y paneles solares (2013)
7. Craven, P., Wahba, G.: Smoothing noisy data with spline functions. Numer. Math. 31, 377–
403 (1978)
8. Salmerón-Quiroz, B.B., Villegas-Medina, G., Guerrero-Castellanos, J., Milhacea, E.O.,
Villalobos-Martinez, R.: A quaternion approach in the estimation of the attitude of an
industrial robot (2017)
Pattern Recognition Through Empirical
Mode Decomposition for Temperature
Time Series Between 1986 and 2019
in Mexico City Downtown for Global
Warming Assessment
1 Introduction
Nowadays, we constantly face dread events due to Climate Change (CC), like
glacier melting and intensification of drought and precipitation episodes [1,2].
Since the middle of the 20th century, scientists detected a clearly increasing
tendency in the temperature records all around the world. This information
along with paleoclimatic records during the Holocene constitute an unfailing
proof of Global Warming (GW) [3,4].
GW is defined as the increment in the global mean temperature over land
and ocean, and it has a direct effect over the whole Earth’s ecosystem [4]. In
its last report, the International Panel on Climate Change (IPCC) remarks that
for the year 2100, a rise in the mean global temperature of 2.5 ◦ C (with respect
to the mean temperature in the 1961–1980 period) is expected, driving severe
consequences for mankind as well as for nature, which threatens the survival of
life at Earth in the nearby future [4,5].
GW has unleashed a very intense research activity in CC since the second half
of the 20th century. Most of the research in the field focuses on both the study of
the anthopogenic causes and consequences of related physicochemical processes,
and on providing projections or predictions of the global mean temperature
behaviour in diverse scenarios [3].
From the point of view of natural phenomena, some of the most remarkable
investigated effects of GW are on droughts [1,6], floods [2,7], permafrost/glacier
alterations [8,9], ENSO phenomenon [10,11], hurricane activity [12,13], sea level
rise [14,15], biodiversity affectations [16,17], as well as the temperature incre-
ment itself [18,19], etc. Of special importance are the impacts of GW on human
socioeconomics. There are clear indications that GW already has importantly
affected agriculture [20,21], wine production [22,23], cattle farming [24,25], wood
assets [26,27], fishing [28,29], textile industry [30,31], transportation [32,33],
manufacturing [34,35], energy [36,37], among many others.
In 2018, an estimated 55.3% of the world’s population lived in urban settle-
ments [38], in contrast to the value of 33.616% in 1960 [39]. Moreover, by 2050
it is expected to rise to 68% (up to two thirds of mankind) [40]. Therefore, the
effects of GW on urban settlements and on their dwellers are of the greatest con-
cern to scientists and governments. In this sense, several hazards for urban envi-
ronments have been detected/forecast, for instance in interurban transportation
[41], urban-based pollution [42,43], health issues in inhabitants, [44,45], urban
design plans and policies [46,47], urban growth and land use [48,49], natural
hazards in city environments [50,51], poverty and gender aspects of population
[52,53], adaptation of current urban infrastructure [54,55], governance, policies,
resilience and adaptation [56,57], feeding challenges, [58,59], sustainability under
population growth [32,60], forced migration/mobility [61,62], as well as pro-
nounced temperature increments yielding to heat waves (heat islands) [63,64],
In this sense, it is very important to focus efforts to mitigate the effects of
GW in big cities. This is the case of Mexico City, the fifth most populated in
the world in 2018 according to United Nations [38] (and the second in America,
just below Sao Paulo, Brazil), which in the last years has faced serious events
of heat waves, precipitations, as well as forest fires and pollution peaks [65]. In
Temperature Pattern Recognition in México City and Global Warming 47
this work, we tackle the rise temperature in Mexico City Downtown by means of
computational science techniques which have proved to be successful to interpret
and forecast data arisen from both natural and social sciences, particularly those
in the field of Artificial Intelligence (AI) [66]. Artificial intelligence is a collection
of methods dealing with the simulation of human intelligence on a machine, so as
to make the machine efficient to identify and use the right piece of “Knowledge”
at a given step of solving a problem [66, section 1.2]. Despite its proved capacity
to analyse data and to solve problems, AI has been barely used in GW and
CC with respect to the huge amount of research in the area. In this sense,
relevant AI techniques to assess CC and WG have been applied, like [67], fuzzy
theory [68,69], agent-based modeling [68,70], swarm optimization [71], genetic
algoritmhs [72,73], pattern recognition [74], machine learning [75,76], cellular
automata [68,77], and artificial neural networks (ANNs) [78,79].
In this paper, currently a work in progress, we implement a pattern recog-
nition system to analyse temperature time series measured near Mexico City
Downtown. The purpose is to identify periodicities and tendencies in the data
that might suggest that temperature is being driven by GW, as well as to build
local indicators to accurately guide local strategic decision-makers to design and
implement adequate policies and actions to mitigate and counteract CC. This
work is presented as follows: In Sect. 2 we present details about the region under
study and about the analysed data, while in Sect. 3 we present the applied the-
oretical background. The obtained results are presented in Sect. 4 along with an
enriching discussion, while finally in Sects. 5 and 6 we present some final remarks,
as well as future work, respectively.
3 Methodology
In this section, we briefly describe the methods used in this paper.
Fig. 2. Temperature time series obtained from MER station, where the values gaps
have been interpolated through the kriging method.
Fig. 3. Individual view of the first six Intrinsic Mode Functions (IMFs) obtained
through the application of EMD to the temperature time series in Fig. 2.
Temperature Pattern Recognition in México City and Global Warming 51
The first extremely notable result that ought to be remarked is the very first
IMF, D1 (t), which is observed in Fig. 3a. The lowest frequency mode, D1 (t),
reveals a clear increasing tendency in the temperature from 1986 to April, 2019.
In order to have a better acquaintance of how representative is this increment,
we take the mean of the temperature records in the time series with a 30 years
baseline, i.e. splitting the data and taking the first set of records between January
1st, 1986 and January 1st, 2016. The baseline mean temperature is 16.98 ◦ C.
Actually, the tendency shown by the first mode function (D1 (t)) reached such
value in August 29th, 1997, at 10:40:59.56, so there is a clear increasing tendency
above the baseline mean temperature since mid 1997. In this sense, we can
confirm that GW has been noticeable since such date.
Furthermore, by the end of the baseline period, the temperature increase was
0.67 ◦ C while by April 30th, 2019, it was 0.72 ◦ C above the baseline mean. In
order to try to assess the future impact of GW in the region under study, we
extrapolated the data for the first frequency function (D1 (t)) in order to try
to forecast the tendency which is present in the dataset. The extrapolation is
performed by a linear algorithm. The temperatures in the meteorological station
under study (MER) as well as their values above the baseline mean temperature
were forecast for three key years according to IPCC: 2030, 2050 and 2100 [92].
Such predicted values can be observed in Table 3.
As it can be observed from the data shown in Table 3, the current tendency
expects an increment of 1.09 ◦ C above the baseline temperature for 2050, which
could increment to 1.68 ◦ C for 2100. Although this simple tendency is within
the SR15 GW scenario which considers as a goal to limit the temperature to
reach 1.5 ◦ C by 2050 [93] it entails disastrous consequences to the urban spot at
Mexico City as well as to its dwellers.
As for the periodicities identified in the following EMD modes in Fig. 3, and
which are shown in Table 2, we can make the following observations.
The second IMF, D2 (t) (see Fig. 3b) reveals a coarse frequency that cor-
responds to a mean period of 12.9960 years, which is almost two years above
the typical mean duration of the solar cycle, so it is possible the solar cycle
to partially drive this IMF. Furthermore, it is interesting to note that the first
minimum in such time series occurred at August 23th, 1987 (see Fig. 3b), while
the closest minimum of solar activity was reported almost one year before, at
August 28th, 1986 [94], which reinforces this hypothesis.
Table 4. Periods T of the thirty harmonics obtained from [94]. The periods are given
in years.
Fig. 4. IMF 22 (D22 (t)) of the EMD applied to the original temperature time series
in Fig. 2. It can be observed that due to its small enough periodicity (1.33 days), it
reproduces qualitatively well the original time series by itself.
Table 5. Comparison of the obtained periods through the application of EMD with
those reported in the literature for the solar activity [94, 95] for the coincident IMFs,
along with the relative error between both.
IMF EMD period (year) Measured period (year) Relative error (%)
D2 (t) 12.9960 10.83 20.00
D3 (t) 6.5174 5.42 20.25
D4 (t) 3.8167 3.61 5.73
D5 (t) 2.5956 2.71 4.22
D6 (t) 1.4766 1.55 4.74
D7 (t) 1.2312 1.20 2.60
D8 (t) 0.9818 0.98 0.18
D9 (t) 0.5864 0.57 2.88
D10 (t) 0.336 0.36 6.67
no match can be made. Nevertheless, they constitute themselves a good hint for
researches to investigate the existence of such very short periods in the solar
activity.
Finally, it is important to observe that the highest frequency IMF, D22 (t),
has a really small period of 0.0009 years, which is equivalent to 1.33 days. In
this sense, its resolution is high enough to almost grasp the day-night cycles. In
Fig. 4 we can observe the IMF 22. Comparing such image with the original time
series (see Fig. 2), it is clear that the 22 IMFs obtained through EMD reproduce
qualitatively well the original time series.
54 M. G. Orozco-del-Castillo et al.
5 Conclusions
In this work we presented the main frequency modes in the time series of tem-
perature records from January 1st, 1986 to April 30th, 2019, with a sample
interval of 1 h, at MER station close to Mexico City’s Downtown. The detected
periods were obtained through EMD after the series were interpolated through
kriging method. The first frequency mode revealed a clear increasing tendency
of temperature in the region under study, which surpassed the baseline period
mean since August 29th, 1997. This temperature increment, most likely due to
the direct impact of GW in the studied area, reached a value of 0.72 ◦ C above
such baseline at the end of the time series, i.e. April 30th, 2019. Furthermore,
linear extrapolation of the tendency provides increments of 0.8491 ◦ C, 1.0872 ◦ C
and 1.6826 ◦ C on the key years 2030, 2050 and 2100, according to IPCC [92,93].
On the other side, when removing the effect of GW on the time series, i.e.
D1 (t), the lower frequency components of the temperature time series are most
likely driven by the solar activity. In order to better grasp this fact, in Table 5
we present a summary of the EMD obtained periods against with those reported
in the literature. In such table, we also present the relative error between both
values. It is noticeable that the greater periods (D2 (t) and D3 (t)) have relative
errors of ≈ 20%, so it remains to study which other factors may influence these
IMFs. Nevertheless, that is out of the scope of this work.
Lower periods (D4 (t) to D10 (t)) match to some reported periods in the solar
activity with reasonable errors (≈ 5%), so they are very likely to be driven by
the sun. Finally, the lowest periods reported in Table 2 (Di (t) ∀ i > 10) are to be
investigated by researchers in order identify if they are part of the several high
frequencies present in solar activity, or if they are due to other physical factors.
6 Future Work
This work, still in progress, has three research directions to follow: (1) integrate
other data into the analysis, such as population, pollution, industry, growth,
etc., factors which could also affect the temperature increase in the metropolitan
area; (2), extend this methodology to the temperature time series arisen from
other strategically located along the MZVM so to build a more robust map of
the effects of GW in a larger region; and (3) implement AI techniques to better
extrapolate the increasing temperature tendency discovered in the first frequency
mode (see Fig. 3a) to provide more realistic forecasts.
References
1. Nguvava, M., Abiodun, B., Otieno, F.: Projecting drought characteristics over East
African basins at specific global warming levels. Atmos. Res. 228, 41–54 (2019).
Cited By 0
2. Yadollahie, M.: The flood in Iran: a consequence of the global warming? Int. J.
Occup. Environ. Med. 10(2), 54–56 (2019)
3. Suarez Gallareta, E., Hernández Gómez, J.J., Cetzal Balam, G., Orozco del
Castillo, M., Moreno Sabido, M., Silva Aguilera, R.A.: Sistema Hı́brido Basado
en Redes Neuronales Artificiales y Descomposición Modal Empı́rica para la Evalu-
ación de la Interrelación entre la Irradiancia Solar Total y el Calentamiento Global.
Res. Comput. Sci. 147(5), 319–332 (2018)
4. Interngubernamental Panel on Climate Change: IPCC: Working Group I Contri-
bution to the IPCC Fifth Assessment Report, Climate Change 2013: The Physical
Science Basis, IPCC, AR5. Interngubernamental Panel on Climate Change (2014)
5. Spratt, D., Dunlop, I.: Existential climate-related security risk: a scenario app-
roach, Melbourne, Australia, May 2019
6. Xu, L., Chen, N., Zhang, X.: Global drought trends under 1.5 and 2 ◦ C warming.
Int. J. Climatol. 39(4), 2375–2385 (2019). Cited By 0
7. Sen, Z.: Noah and Joseph effects: floods and droughts under global warming. Int.
J. Glob. Warm. 16(3), 347–364 (2018)
8. Ding, Y., Zhang, S., Zhao, L., Li, Z., Kang, S.: Global warming weakening the
inherent stability of glaciers and permafrost. Sci. Bull. 64(4), 245–253 (2019)
9. Kotlyakov, V.M., et al.: Glacier revival and advances in the period of global warm-
ing. Dokl. Earth Sci. 481(2), 1113–1118 (2018)
10. Ying, J., Huang, P., Lian, T., Chen, D.: Intermodel uncertainty in the change
of ENSO’s amplitude under global warming: role of the response of atmospheric
circulation to SST anomalies. J. Clim. 32(2), 369–383 (2019)
11. Jiang, N., Zhu, C.: Asymmetric changes of ENSO diversity modulated by the cold
tongue mode under recent global warming. Geophys. Res. Lett. 45(22), 12506–
12513 (2018)
12. González-Alemán, J., Pascale, S., Gutierrez-Fernandez, J., Murakami, H., Gaert-
ner, M., Vecchi, G.: Potential increase in hazard from Mediterranean hurricane
activity with global warming. Geophys. Res. Lett. 46(3), 1754–1764 (2019)
13. Duan, H., Chen, D., Lie, J.: The impact of global warming on hurricane intensity.
In: IOP Conference Series: Earth and Environmental Science, vol. 199, no. 2 (2018)
14. Khan, A.: Why would sea-level rise for global warming and polar ice-melt? Geosci.
Front. 10(2), 481–494 (2019)
15. Carvalho, K., Wang, S.: Characterizing the Indian Ocean sea level changes and
potential coastal flooding impacts under global warming. J. Hydrol. 569, 373–386
(2019)
16. Tang, L., Higa, M., Tanaka, N., Itsubo, N.: Assessment of global warming impact
on biodiversity using the extinction risk index in LCIA: a case study of Japanese
plant species. Int. J. Life Cycle Assess. 23(2), 314–323 (2018)
17. Yamaura, K., Sakaue, S., Washida, T.: An assessment of global warming and biodi-
versity: CGE EMEDA analyses. Environ. Econ. Policy Stud. 19(2), 405–426 (2017)
18. Nangombe, S.S., Zhou, T., Zhang, W., Zou, L., Li, D.: High-temperature extreme
events over Africa under 1.5 and 2 ◦ C of global warming. J. Geophys. Res.: Atmos.
124(8), 4413–4428 (2019)
56 M. G. Orozco-del-Castillo et al.
19. Verheyen, J., Delnat, V., Stoks, R.: Increased daily temperature fluctuations over-
rule the ability of gradual thermal evolution to offset the increased pesticide toxicity
under global warming. Environ. Sci. Technol. 53(8), 4600–4608 (2019)
20. Huo, D., et al.: Global-warming-caused changes of temperature and oxygen alter
the proteomic profile of sea cucumber Apostichopus japonicus. J. Proteomics 193,
27–43 (2019)
21. Yousefpour, R., Nabel, J., Pongratz, J.: Simulating growth-based harvest adaptive
to future climate change. Biogeosciences 16(2), 241–254 (2019)
22. Drappier, J., Thibon, C., Rabot, A., Geny-Denis, L.: Relationship between wine
composition and temperature: impact on Bordeaux wine typicity in the context of
global warming-review. Crit. Rev. Food Sci. Nutr. 59(1), 14–30 (2019)
23. Di Lena, B., Silvestroni, O., Lanari, V., Palliotti, A.: Climate change effects on cv.
Montepulciano in some wine-growing areas of the Abruzzi region (Italy). Theor.
Appl. Climatol. 136(3–4), 1145–1155 (2019)
24. Campbell, A., Becerra, T., Middendorf, G., Tomlinson, P.: Climate change beliefs,
concerns, and attitudes of beef cattle producers in the Southern Great Plains. Clim.
Change 152(1), 35–46 (2019)
25. Duru, S., Baycan, S.: Change of daily milk yield during estrous period in Holstein
cattle raised under Mediterranean climate. Trop. Anim. Health Prod. 51, 1571–
1577 (2019)
26. Geng, A., Ning, Z., Zhang, H., Yang, H.: Quantifying the climate change mitiga-
tion potential of China’s furniture sector: wood substitution benefits on emission
reduction. Ecol. Ind. 103, 363–372 (2019)
27. Salman, A., Li, Y., Bastidas-Arteaga, E.: Impact of climate change on optimal
wood pole asset management. In: Life-Cycle Analysis and Assessment in Civil
Engineering: Towards an Integrated Vision - Proceedings of the 6th International
Symposium on Life-Cycle Civil Engineering, IALCCE 2018, pp. 1727–1734 (2019)
28. Avelino, J., et al.: Survey tool for rapid assessment of socio-economic vulnerability
of fishing communities in Vietnam to climate change. Geosciences (Switzerland)
8(12), 452 (2018)
29. Sharifuzzaman, S., Hossain, M., Chowdhury, S., Sarker, S., Chowdhury, M.,
Chowdhury, M.: Elements of fishing community resilience to climate change in
the coastal zone of Bangladesh. J. Coast. Conserv. 22(6), 1167–1176 (2018)
30. Rowe, M.: Textile industry welcomes change prompted by Paris climate deal. Twist
1(77), 14–15 (2016)
31. De L’Ecluse, S.: COP21: textile industry also acts against global warming [COP
21 le textile agit aussi contre le Réchauffement]. J. du Textile 52(2275), 2–4 (2015)
32. Yang, D., Frangopol, D.: Societal risk assessment of transportation networks under
uncertainties due to climate change and population growth. Struct. Saf. 78, 33–47
(2019)
33. Markolf, S., Hoehne, C., Fraser, A., Chester, M., Underwood, B.: Transportation
resilience to climate change and extreme weather events - beyond risk and robust-
ness. Transp. Policy 74, 174–186 (2019)
34. Meinel, U., Schüle, R.: The difficulty of climate change adaptation in manufactur-
ing firms: developing an action-theoretical perspective on the causality of adaptive
inaction. Sustainability (Switzerland) 10(2), 569 (2018)
35. Carlsson Kanyama, A., Carlsson Kanyama, K., Wester, M., Snickare, L., Söderberg,
I.L.: Climate change mitigation efforts among transportation and manufacturing
companies: the current state of efforts in Sweden according to available documen-
tation. J. Clean. Prod. 196, 588–593 (2018)
Temperature Pattern Recognition in México City and Global Warming 57
36. Bastida, L., Cohen, J., Kollmann, A., Moya, A., Reichl, J.: Exploring the role
of ICT on household behavioural energy efficiency to mitigate global warming.
Renew. Sustain. Energy Rev. 103, 455–462 (2019)
37. De Cian, E., Sue Wing, I.: Global energy consumption in a warming climate.
Environ. Resour. Econ. 72(2), 365–410 (2019)
38. United Nations: The world’s cities in 2018: data booklet. Economics and Social
Affairs series (2018)
39. World Bank: Urban population (% of total)—Data (2018). https://data.
worldbank.org/indicator/sp.urb.totl.in.zs. United Nations Population Division.
World Urbanization Prospects: 2018 Revision. Accessed 9 June 2019
40. United Nations Department of Economic and Social Affairs: 68% of the world
population projected to live in urban areas by 2050, says UN—UN DESA, June
2018. https://www.un.org/development/desa/en/news/population/2018-revision-
of-world-urbanization-prospects.html. Accessed 9 June 2019
41. Landis, J., Hsu, D., Guerra, E.: Intersecting residential and transportation CO2
emissions: metropolitan climate change programs in the age of trump. J. Plan.
Educ. Res. 39(2), 206–226 (2019)
42. Trájer, A., Nagy, G., Domokos, E.: Exploration of the heterogeneous effect of cli-
mate change on ozone concentration in an urban environment. Int. J. Environ.
Health Res. 29(3), 276–289 (2019)
43. Baldermann, C., Lorenz, S.: UV radiation in Germany: influences of ozone deple-
tion and climate change and measures to protect the population. Bundesgesund-
heitsblatt - Gesundheitsforschung - Gesundheitsschutz 62(5), 639–645 (2019)
44. Orimoloye, I., Mazinyo, S., Kalumba, A., Ekundayo, O., Nel, W.: Implications of
climate variability and change on urban and human health: a review. Cities 91,
213–223 (2019)
45. Woodward, A., Baumgartner, J., Ebi, K., Gao, J., Kinney, P., Liu, Q.: Population
health impacts of China’s climate change policies. Environ. Res. 175, 178–185
(2019)
46. Aminipouri, M., et al.: Urban tree planting to maintain outdoor thermal comfort
under climate change: the case of Vancouver’s local climate zones. Build. Environ.
158, 226–236 (2019)
47. Sun, Y., Xie, S., Zhao, S.: Valuing urban green spaces in mitigating climate change:
a city-wide estimate of aboveground carbon stored in urban green spaces of China’s
capital. Glob. Change Biol. 25(5), 1717–1732 (2019)
48. Doan, V., Kusaka, H., Nguyen, T.: Roles of past, present, and future land use
and anthropogenic heat release changes on urban heat island effects in Hanoi,
Vietnam: numerical experiments with a regional climate model. Sustain. Cities
Soc. 47, 101479 (2019)
49. Chapman, S., Thatcher, M., Salazar, A., Watson, J., McAlpine, C.: The impact of
climate change and urban growth on urban climate and heat stress in a subtropical
city. Int. J. Climatol. 39(6), 3013–3030 (2019)
50. Zhang, C., Li, S., Luo, F., Huang, Z.: The global warming hiatus has faded away:
an analysis of 2014–2016 global surface air temperatures. Int. J. Climatol. 2019,
1–16 (2019)
51. Shastri, H., Ghosh, S., Paul, S., Shafizadeh-Moghadam, H., Helbich, M., Karmakar,
S.: Future urban rainfall projections considering the impacts of climate change and
urbanization with statistical-dynamical integrated approach. Clim. Dyn. 52(9–10),
6033–6051 (2019)
52. Dodman, D., Archer, D., Satterthwaite, D.: Editorial: responding to climate change
in contexts of urban poverty and informality. Environ. Urban. 31(1), 3–12 (2019)
58 M. G. Orozco-del-Castillo et al.
53. Gran Castro, J., Ramos De Robles, S.: Climate change and flood risk: vulnerability
assessment in an urban poor community in Mexico. Environ. Urban. 31(1), 75–92
(2019)
54. Xiong, L., Yan, L., Du, T., Yan, P., Li, L., Xu, W.: Impacts of climate change on
urban extreme rainfall and drainage infrastructure performance: a case study in
Wuhan City, China. Irrig. Drain. 68(2), 152–164 (2019)
55. De la Sota, C., Ruffato-Ferreira, V., Ruiz-Garcı́a, L., Alvarez, S.: Urban green
infrastructure as a strategy of climate change mitigation. A case study in Northern
Spain. Urban For. Urban Green. 40, 145–151 (2019)
56. Bellinson, R., Chu, E.: Learning pathways and the governance of innovations in
urban climate change resilience and adaptation. J. Environ. Plan. Policy Manag.
21(1), 76–89 (2019)
57. Martinez-Juarez, P., Chiabai, A., Suárez, C., Quiroga, S.: Insights on urban and
periurban adaptation strategies based on stakeholders’ perceptions on hard and
soft responses to climate change. Sustainability (Switzerland) 11(3), 647 (2019)
58. Lal, R.: Managing urban soils for food security and adaptation to climate change.
In: Vasenev, V., Dovletyarova, E., Cheng, Z., Prokof’eva, T.V., Morel, J.L.,
Ananyeva, N.D. (eds.) SUITMA 2017. SG, pp. 302–319. Springer, Cham (2019).
https://doi.org/10.1007/978-3-319-89602-1 35
59. Raimundo, I.: Food insecurity in the context of climate change in Maputo city,
Mozambique: challenges and coping strategies. ROUTLEDGE in Association with
GSE Research (2016)
60. Mahaut, V., Andrieu, H.: Relative influence of urban-development strategies and
water management on mixed (separated and combined) sewer overflows in the
context of climate change and population growth: a case study in Nantes. Sustain.
Cities Soc. 44, 171–182 (2019)
61. Januszkiewicz, K., Świtoń, M.: Climate change and population mobility - envision-
ing infrastructure to reduce disaster’s impact on cities. In: International Multidis-
ciplinary Scientific GeoConference Surveying Geology and Mining Ecology Man-
agement, SGEM, vol. 17, pp. 519–526 (2017)
62. Khavarian-Garmsir, A., Pourahmad, A., Hataminejad, H., Farhoodi, R.: Climate
change and environmental degradation and the drivers of migration in the context
of shrinking cities: a case study of Khuzestan province, Iran. Sustain. Cities Soc.
47, 101480 (2019)
63. Trihamdani, A.R., Lee, H.S., Kubota, T., Iizuka, S., Phuong, T.T.T.: Urban cli-
mate challenges in Hanoi: urban heat Islands and global warming. In: Kubota,
T., Rijal, H.B., Takaguchi, H. (eds.) Sustainable Houses and Living in the Hot-
Humid Climates of Asia, pp. 529–539. Springer, Singapore (2018). https://doi.org/
10.1007/978-981-10-8465-2 48
64. Li, G., Zhang, X., Mirzaei, P., Zhang, J., Zhao, Z.: Urban heat island effect of a
typical valley city in China: responds to the global warming and rapid urbanization.
Sustain. Cities Soc. 38, 736–745 (2018)
65. Trejo-González, A., et al.: Quantifying health impacts and economic costs of PM2.5
exposure in Mexican cities of the national urban system. Int. J. Public Health
64(4), 561–572 (2019)
66. Konar, A.: Artificial Intelligence and Soft Computing: Behavioral and Cognitive
Modeling of the Human Brain. CRC Press, Boca Raton (1999)
67. Woo, T.: Artificial intelligence-based modelling for global warming analysis incor-
porated with the nuclear energy productions. Int. J. Glob. Warm. 17(4), 389–400
(2019)
Temperature Pattern Recognition in México City and Global Warming 59
68. Lu, Q., Joyce, J., Imen, S., Chang, N.B.: Linking socioeconomic development, sea
level rise, and climate change impacts on urban growth in New York City with a
fuzzy cellular automata-based Markov chain model. Environ. Plan. B: Urban Anal.
City Sci. 46(3), 551–572 (2019)
69. Woo, T.: Global warming analysis for greenhouse gases impacts comparable to
carbon-free nuclear energy using neuro-fuzzy algorithm. Int. J. Glob. Warm. 17(2),
219–233 (2019)
70. Gubareva, M., Gomes, O.: On the edge of climate change: in a search of an adequate
agent-based methodology to model environmental dynamics. In: Sequeira, T., Reis,
L. (eds.) Climate Change and Global Development. CE, pp. 37–57. Springer, Cham
(2019). https://doi.org/10.1007/978-3-030-02662-2 3
71. Atuahene, S., Bao, Y., Yevenyo Ziggah, Y., Semwaah Gyan, P., Li, F.: Short-
term electric power forecasting using dual-stage hierarchical wavelet-particle swarm
optimization-adaptive neuro-fuzzy inference system PSO-ANFIS approach based
on climate change. Energies 11(10), 2822 (2018)
72. Mohammadi, B.: “Prediction of effective climate change indicators using statistical
downscaling approach and impact assessment on pearl millet (Pennisetum glaucum
L.) yield through genetic algorithm in Punjab, Pakistan” by Asmat Ullah, Nasrin
Salehnia, Sohrab Kolsoumi, Ashfaq Ahmad, Tasneem Khaliq. Ecol. Indic. 101,
973–974 (2019)
73. Yoon, E., et al.: Modeling spatial climate change landuse adaptation with multi-
objective genetic algorithms to improve resilience for rice yield and species richness
and to mitigate disaster risk. Environ. Res. Lett. 14(2), 024001 (2019)
74. Vaghefi, S.A., Abbaspour, N., Kamali, B., Abbaspour, K.: A toolkit for climate
change analysis and pattern recognition for extreme weather conditions - case
study: California-Baja California Peninsula. Environ. Model Softw. 96, 181–198
(2017)
75. Wu, C., Chen, Y., Peng, C., Li, Z., Hong, X.: Modeling and estimating aboveground
biomass of Dacrydium pierrei in China using machine learning with climate change.
J. Environ. Manag. 234, 167–179 (2019)
76. Crane-Droesch, A.: Machine learning methods for crop yield prediction and climate
change impact assessment in agriculture. Environ. Res. Lett. 13(11), 114003 (2018)
77. Collados-Lara, A.J., Pardo-Igúzquiza, E., Pulido-Velazquez, D.: A distributed cel-
lular automata model to simulate potential future impacts of climate change on
snow cover area. Adv. Water Resour. 124, 106–119 (2019)
78. Nile, B., Hassan, W., Alshama, G.: Analysis of the effect of climate change on
rainfall intensity and expected flooding by using ANN and SWMM programs.
ARPN J. Eng. Appl. Sci. 14(5), 974–984 (2019)
79. Aslan, Z., Erdemir, G., Feoli, E., Giorgi, F., Okcu, D.: Effects of climate change
on soil erosion risk assessed by clustering and artificial neural network. Pure Appl.
Geophys. 176(2), 937–949 (2019)
80. Quezada, R., Martinez Rossier, L., Hernandez Garduno, E., Catalan Vazquez, M.,
Quezada, E., Perez Neria, J.: Comparison of carbon monoxide levels in air expired
from pedestrian in an urban area and a rural area. Rev. del Inst. Nac. de Enferm.
Respiratorias 10(1), 13–17 (1997)
81. Matheron, G.: Principles of geostatistics. Econ. Geol. 58(8), 1246–1266 (1963)
82. Krige, D.: A statistical approach to some basic mine valuation problems on the
Witwatersrand. J. Chem. Metall. Min. Soc. S. Afr. 52, 201–215 (1952)
83. Zhang, J., Li, X., Yang, R., Liu, Q., Zhao, L., Dou, B.: An extended Kriging
method to interpolate near-surface soil moisture data measured by wireless sensor
networks. Sensors 17(6), 1390 (2017)
60 M. G. Orozco-del-Castillo et al.
84. Liu, L., Cheng, Y., Wang, X.: Genetic algorithm optimized Taylor Kriging surro-
gate model for system reliability analysis of soil slopes. Landslides 14(2), 535–546
(2017)
85. Huang, N.E., et al.: The empirical mode decomposition and the Hilbert spectrum
for nonlinear and non-stationary time series analysis. Proc. R. Soc. Lond. A: Math.
Phys. Eng. Sci. 454, 903–995 (1998)
86. Kim, D., Oh, H.S.: EMD: a package for empirical mode decomposition and Hilbert
spectrum. R J. 1(1), 40–46 (2009)
87. Huang, N.E.: Review of empirical mode decomposition. In: Wavelet Applications
VIII, vol. 4391, pp. 71–81. International Society for Optics and Photonics (2001)
88. Du, H.K., Cao, J.X., Xue, Y.J., Wang, X.J.: Seismic facies analysis based on self-
organizing map and empirical mode decomposition. J. Appl. Geophys. 112, 52–61
(2015)
89. Nunes, J.C., Bouaoune, Y., Delechelle, E., Niang, O., Bunel, P.: Image analysis by
bidimensional empirical mode decomposition. Image Vis. Comput. 21(12), 1019–
1026 (2003)
90. Andrade, A.O., Nasuto, S., Kyberd, P., Sweeney-Reed, C.M., Van Kanijn, F.: EMG
signal filtering based on empirical mode decomposition. Biomed. Signal Process.
Control 1(1), 44–55 (2006)
91. Drakakis, K.: Empirical mode decomposition of financial data. Int. Math. Forum
4, 1191–1202 (2008)
92. Intergubernamental Panel on Climate Change: Homepage (2014). http://www.
ipcc.ch/home languages main spanish.shtml. Accessed 01 Apr 2019
93. Masson-Delmotte, V., et al.: Global Warming of 1.5C. An IPCC Special Report
on the impacts of global warming of 1.5C above pre-industrial levels and related
global greenhouse gas emission pathways, in the context of strengthening the global
response to the threat of climate change, sustainable development, and efforts to
eradicate poverty. Technical report 1, IPCC, Geneva, Switzerland (2018, in Press)
94. Orozco-Del-Castillo, M., Ortiz-Alemán, J., Couder-Castañeda, C., Hernández-
Gómez, J., Solı́s-Santomé, A.: High solar activity predictions through an artificial
neural network. Int. J. Mod. Phys. C 28(06), 1750075 (2017)
95. Kane, R.: Short-term periodicities in solar indices. Sol. Phys. 227(1), 155–175
(2005)
Moderate Resolution Imaging
Spectroradiometer Products
Classification Using Deep Learning
Javier Arellano-Verdejo(B)
1 Introduction
X2 w1
X1 w2
f S
n
s= f Wi X i
Input Hidden layers Output i=1
a) b)
Fig. 1. BASIC architecture of a multilayer artificial neural network
model that best classifies the information. With this set of data and applying a
learning algorithm called “retro-propagation” algorithm based on the descending
gradient technique, the ANN generates a structure that has the ability to classify
information that is not in the original data set correctly [10].
The Convolutional Neural Networks or CNN, traditionally have been used almost
specifically in the classification of images (Fig. 2). CNN has been used success-
fully in multiple areas such as computer science, medicine, economics, music,
painting, and science, to mention just a few [5–8,11]. In CNN, neurons corre-
spond to receptive fields in a manner similar to neurons in the primary visual
cortex of the biological brain. The CNN is composed of a set of filters of one
or more dimensions used to highlight specific features of the information you
want to classify, at the end of each filter layer, a function that is responsible for
performing a non-linear mapping is applied. Unlike the ANNs, during the CNN
training process a set of specialized filters is obtained instead of wi type weights.
The main operation within CNN is called convolution. The convolution allows
highlighting characteristics or specific aspects of the image and it is precisely
this ability to extract features that have made the CNN widely used successfully
in subjects related to computer vision and classification of images.
Input C
these images were extracted is located within the Caribbean Sea of Mexico and
part of Central America. The images were obtained by applying a window with
25% overlap between images to preserve the spatial relationships between them
(see Fig. 3). The data set is composed of four classes of MODIS products: chloro-
phyll, sea surface temperature, fluorescence and the photosynthetically available
instantaneous radiation (IPAR).
to extract the representative information of the last set of filters, in the last
section of the convolutional block a two-dimensional “Max Pooling” operation,
was used. The classifying block (multilayer perceptron) is composed of an input
layer formed by 128 neurons with a RELU-like activation function and an output
layer of 4 neurons with a softmax-type activation function.
Dense 128
Softmax
Dense 4
Relu
Relu
Input
Relu
3x3 3x3 2x2
MP2D
32 64
For the implementation of the proposed neural network, the Python 3.7.0
programming language and the library Keras 2.2.4 with TensorFlow 1.10.0 as a
backend were used. TensorFlow is an Open Source library developed by Google’s
Brain Team, it is used for the numerical calculation using “flow graphing pro-
gramming”. The nodes in the graph represent mathematical operations while
the connections within the graph represent multidimensional data sets known as
“Tensors”. Tensor Flow contains a set of automatic learning algorithms as well
as other tools that make it ideal for the development of new methods. Keras is a
library that can be used by Python and provides an abstraction layer that facil-
itates access to TensorFlow to create deep learning models. The convolutional
neural network was trained and tested using a Lenovo workstation with an 8-core
Xeon EP processor with 64 GB of RAM and an Nvidia Quadro k5000 GPU with
1536 parallel processing CUDA cores using the Ubuntu 18.04 operating system.
4 Results
To carry out the training and testing of the proposed neural network, the k-fold
procedure with k = 5 was used as a cross-validation method. The cross-validation
process is used to estimate the ability of an CNN to classify information that
has never been presented to the model (generalization), that is, a limited sample
of information is used to estimate how the Model performs in general when used
to make predictions about data that was not used during model training. K-fold
is one of the most popular cross-validation methods because it results in a less
biased or less optimistic estimate of the model’s capacity than other methods,
such as a simple trial-and-error division. During the training of the proposed
network, 80% of data was used to design (training, and testing) the CNN while
the rest 20% was used for the validation. As it can be seen in Fig. 5, both the
data set and the validation data set are balanced, that is, the total number
66 J. Arellano-Verdejo
200,000
150,000
chlor
ipar
sst
100,000
50,000
chlor
ipar
sst
Training Testing
Fig. 5. Histogram of training and test data sets
of elements for the class used for each of the cases is similar. The purpose of
avoiding biases when generating the model.
The training process was carried out for 20 epochs or iterations, in each
iteration blocks (batch size) of 100 images were used. The optimization method
used during the training phase was Adam’s algorithm. The Adam optimization
algorithm (Addictive moment estimation) was presented by Kingma Diederik
in 2015 [12] and this is an extension of the stochastic gradient method, this
algorithm has been adopted in the field of deep learning for its profound impact
on the applications of computer vision and natural language processing.
Figure 6 shows the convergence graph during the training process, the black
dotted line shows the behavior of the neural network during the training phase, it
is evident, the network is able to obtain a good model of classification relatively
fast. As you can see, at the epoch 10 of the training phase, it is clear that
the neural network has learned to distinguish between different classes with an
accuracy close to 100%. The continuous line of green color shows the validation
of the neural network, that is to say, the capacity of the network to classify data
that has never seen previously in other words, the generalization capability of
the network. Thereby it is clear the overfitting is not present as it can be seen in
the plot result. Finally derived from the behavior and trend of the data during
the training process, it can be clearly distinguished that the network has not
memorized the data, instead, it has generated a robust enough model for the
successful classification of MODIS satellite products.
In Table 1, a summary with the results obtained is shown. As can be seen, the
percentage of successful classification by the neural network was 99%. During
the cross-validation phase, it was found that the architecture is able to classify
up to 99.97% of the information correctly. Also in the same table, it can be seen
MODIS Products Classification Using Deep Learning 67
100 %
90 %
Accuracy
80 %
70 %
60 %
Fig. 6. Convergence plot during the training phase (Color figure online)
that the measures of central tendency and dispersion are similar between the
training and test data sets, which implies that the results, at least from this
classical point of view, they have the same tendency and statistical dispersion.
5 Discussion
The classification of images using neural networks and deep learning is not new,
In general, the classification of images is done using three bands mainly (red,
green and blue) however, depending on the complexity of the data set, neural
networks, even convolutional networks, are not able to achieve optimal results. In
many cases, for the correct classification of information, the development of new
68 J. Arellano-Verdejo
In the present work, a convolutional neural network was presented for the
automatic classification of MODIS satellite products. After the careful elabora-
tion of a data bank with more than 250,000 images of MODIS products divided
and correctly labeled in four classes for a wide region of the Caribbean Sea of
Mexico, a CNN capable of automatically classifying these images was designed.
As a result of using the k-fold cross-validation method, a CNN with a classi-
fication capacity of 99% accuracy was obtained. The application of measures
of central tendency and dispersion showed that both the training data set and
the validation data are statistically consistent, which, together with the results
obtained, helps to conclude that the proposed network is capable of generalizing
with a high degree of confidence.
7 Future Work
As part of the future work, the authors are currently working on the design of
a new architecture of Neural Network for the automatic classification of more
than 20 MODIS products. With a database of more than 1,019,200,000 data
and more than 1,300,000 images for the Caribbean Sea, new storage, access and
information manipulation challenges are being faced, which implies, among other
things, the design of new algorithms as well as the use of from other disciplines
such as data science, Big Data, parallel programming using graphical processing
units clusters and specialized hardware units for the processing of tensors like
Google TPU technology.
References
1. Russell, S.J., Norvig, P.: Artificial Intelligence: A Modern Approach. Pearson Edu-
cation Limited, Kuala Lumpur (2016)
2. Jensen, J.R., Lulla, K.: Introductory Digital Image Processing: A Remote Sensing
Perspective. Prentice Hall, Upper Saddle River (1987)
3. Arellano-Verdejo, J., Lazcano-Hernandez, H.E., Cabanillas-Terán, N.: ERISNet:
deep neural network for Sargassum detection along the coastline of the Mexican
Caribbean. PeerJ 7, e6842 (2019). https://doi.org/10.7717/peerj.6842
4. Lazcano-Hernandez, H.E., Arellano-Verdejo, J., Hernandez-Arana, H.A., Alvarado-
Barrientos, M.S.: Spatio-temporal assessment of “Chlorophyll a” in Banco Chin-
chorro using remote sensing. Res. Comput. Sci. 147(12), 213–223 (2018)
5. Cheng, G., Yang, C., Yao, X., Guo, L., Han, J.: When deep learning meets met-
ric learning: remote sensing image scene classification via learning discriminative
CNNs. IEEE Trans. Geosci. Remote Sens. 56(5), 2811–2821 (2018)
6. Rajkomar, A., et al.: Scalable and accurate deep learning with electronic health
records. NPJ Digit. Med. 1(1), 18 (2018)
7. Poplin, R., et al.: Prediction of cardiovascular risk factors from retinal fundus
photographs via deep learning. Nat. Biomed. Eng. 2(3), 158 (2018)
8. Han, J., Zhang, D., Cheng, G., Liu, N., Xu, D.: Advanced deep-learning techniques
for salient and category-specific object detection: a survey. IEEE Signal Process.
Mag. 35(1), 84–100 (2018)
70 J. Arellano-Verdejo
9. Haykin, S.S., Elektroingenieur, K., Haykin, S.S.: Neural Networks and Learning
Machines, vol. 3. Pearson Education, Upper Saddle River (2009)
10. Michie, D., Spiegelhalter, D.J., Taylor, C.C.: Machine Learning. Neural and Sta-
tistical Classification, 13 (1994)
11. LeCun, Y., Kavukcuoglu, K., Farabet, C.: Convolutional networks and applications
in vision. In: Proceedings of 2010 IEEE International Symposium on Circuits and
Systems, pp. 253–256. IEEE (2010)
12. Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization. arXiv preprint
arXiv:1412.6980 (2014)
A Dynamic System Model Using Agile
Method Techniques Applied to Geomatics
1 Introduction
The vertiginous advance of the digital technology has propitiated the whole
transformation in the development of the activities that the persons realize daily.
The evolution in the land use of the territory is a fundamental element in our
society, since it manifests different variables that affect our daily life, for example,
accessibility to different points of interest within the city, slopes of the land, etc.
This evolution has gained interest mainly fueled by the different environmental
problems especially those in urban areas [1–3].
Thanks to the advances in the computing field and the development of impor-
tant analytical tools such as Geographic Information Systems (GIS) or simula-
tion models, the study of the changes taking place in metropolitan areas has been
promoted [4]. The analysis of the environmental alterations that result from these
changes and the development of new planning instruments, has caused that dif-
ferent disciplines, specifically the Artificial Intelligence (AI), approaches from a
computer and mathematical point of view to give alternative solutions to this
problem [5,6]. Numerous modeling tools have emerged in recent years. In the
case of urban growth, the models based on cellular automata (CA) are the most
widely used [7,8].
In this paper we present a discrete dynamic system represented by a cellular
automaton that can be used to simulate a geographical area. We test the model
by doing experiments from a digitalized map and generate a corresponding map
from a simulation. Finally, as an additional element to the model definition we
use user stories and scenario description as agile method techniques for commu-
nicating important aspects that characterize the model.
2 Dynamic Systems
The dynamic systems are systems which internal parameters (variables of the
state) follow a series of temporary rules. There are called systems because they
are described by a set of equations (system) and dynamic because its parameters
change with regard some variable that generally is the time. The study of the
dynamic systems can be divided into 3 subdisciplines:
– Applied dynamics: Modeling processes using state equations that relate past
states to future states.
– Dynamics Mathematics: Focuses on the qualitative analysis of the dynamic
model.
– Experimental Dynamics: Laboratory experiments, computer simulations of
dynamic models.
The dynamic systems can split into two big classes: those in which the
time changes continuously and in which the time passes discreetly. Dynamic
continuous-time systems are expressed with differential equations; these can be
ordinary differential equations, differential equations in derivative partial, and
differential equations with delays. On the other hand if the time is discreet the
systems are described by means of difference equations. In the following figures
some examples of dynamic systems can be observed [9,10].
The cellular automata were conceived at the beginning of the 1950’s by John
Von Neumann, a mathematical brilliant interested in investigating about the
required complexity so that a device could be auto-reproductive and the organi-
zation of this one to have a correct operation when repairing itself with part that
could have and operation badly [11]. The results were not let hope, all a current
of investigation has been developed with strong interrelation in fields like frac-
tals and dynamic systems, as well as applications in: parallel cellular computing,
simulation of dynamic systems and recognition of patterns, and simulation of
socio-economical models, for instance (Figs. 1 and 2).
In general therms, we can say that a cellular automata is defined and operates
based on the following elements [11]:
A Dynamic System Model Using Agile Methods Applied to Geomatics 73
3 Agile Methods
There are methodologies based on classical paradigms within the scope of soft-
ware development guided by models used at the time by software engineers like
linear models represented in cascade, and the evolutionary, iterative and incre-
mental represented in spiral and prototypes.
One of the most widely used standard methodologies is the Unified Process
(UP) for different application areas and different classes of organizations [15].
There are three characteristics that distinguish the UP from the rest of the
methodologies: is centered in the architecture, is iterative and incremental, and is
oriented and directed by use case. Although these characteristics distinguish UP,
do not leave aside the advantages offered by other methodologies, that is, they
take advantage of their strengths in the development of stages and phases that
are already well established in the field of software engineering, but increasing
the flexibility and giving the first guidelines for which are known as agile methods
[16].
The agile methods are part of a set of methodologies also known like light
which use similar practices based on the results, the persons and its interaction.
Exists a diversity of methodologies as the extreme programming (XP) Crys-
tal Clear, Feature Driven Development (FDD), Adaptive Software Development
(ASD) and Scrum. All these methodologies adopt the one that is known as an
agile manifesto, which values:
76 R. Rodrı́guez Zamora et al.
These four postulates are implemented using agile principles such as: satisfying
the client through an early and continuous delivery of the software, the fact that
the requirements are changing, the elements of an organization are positively
assumed. They must work together on a daily basis during the development
of the project, it is established that the best architectures, requirements and
designs emerge from teams that self-organize.
Agile methodologies seek at all times teamwork, multidisciplinary and col-
laborative, which allows to carry out projects with immediacy and flexibility,
adapted to changes, demanding at any time modify the mentality, a cultural
change based on respect, responsibility, values, competencies and skills of the
person.
An agile methodology implements a series of practices (values, aptitudes and
skills) short and repetable, that is to say, it is a question of iterating repeatedly
to be adapting changeable requisites inside the frame of the scope of the project.
It seeks to get ready entregables often, with preference to the period of time as
short as possible, to refine and to converge to an acceptable solution.
In addition to the model incremental, some other differential aspects of agile
methodologies are, without doubt, the following:
Within the set of agile methodologies is located SCRUM [17,18], which provides
a number of tools and roles for, iteratively, be able to see the progress and results
of a project.
A way to use user stories [19,20] to define in this case part of the dynamic
system model is through the technique BDD (Behavior Driven Development) in
which integrates the history a scenario description of behavior.
A Dynamic System Model Using Agile Methods Applied to Geomatics 77
4 Model Definition
The fundamental idea in CA Models is that the state of a cell at any given
time depends on the state of the cells within its neighborhood in the previous
time step, based on a set of transition rules [21]. The CA model used in this
investigation is the one proposed by R. White (see Eq. 2), is a constrained cellular
automata for high-resolution modelling of urban land-use dynamics [22,23]. The
CA model is oriented towards the representation of the attributes of a given
geographic region in a two-dimensional lattice, raster maps provides these data
format to the CA.
A raster map can be represented formally by an array of real values. This
matrix is represented as A = aij of order mxn such that 0 ≤ i ≤ m, 0 ≤ j ≤ n
where each element A = [aij ] ∈ R.
A neighborhood filter matrix 1 is required to analyze each element A = aij ,
this neighborhood is formally represented B = bii of order nxn such that 0 ≤
i ≤ n where each element B = [bii ] ∈ Z.
⎡ ⎤
bi−1j−1 bi−1j bi−1j+1
B = ⎣ bij−1 bij bij+1 ⎦ (1)
bi+1j−1 bi+1j bi+1j+1
The neighborhood filter is used to calculate the transition potential from state h
to j for each element A = [aij ]. The calculation methodology is detailed below:
78 R. Rodrı́guez Zamora et al.
⎛ ⎞
Phj = vsj aj ⎝1 + mkd Iid ⎠ + Hj (2)
k,i,d
where
Phj : is the transition potential of state h to state j.
x
v: stochastic perturbation term. v = 1 + [−ln(random)] . 0 < random < 1,
and x allows you to adjust the size of the disturbance.
sj represents the suitability of the state of the cell.
aj : euclidean distance from the cell to the nearest road.
mkd : calibration matrix, contains the weights of each cell as a function of its
state k and distance d.
1, i = k
I id =
0, i = k
where i is the index of the cell in the current neighborhood, k.
The transition potential Phj of each cell Aij is calculated only if the suitability of
the objective state sj > 0. That is, for each cell (pixel) in the map, its transition
potential will be calculated except for those in which its suitability is equal to
zero. For the neighborhood calculation, the calibration matrix mkd gives each
neighbor cell bii a weight based on its state and distance (subscript d in 2)
concerning the analyzed cell aij . The nearby neighbor cells will generally have a
higher weight, positive values are taken for an attractive effect and negative for
repulsive effect, these values tend to decrease as the distance increases between
the analyzed cell and its neighbor, this is called Distance Decay Effect. When
analyzing the neighbor cells, the Iid component helps to filter (multiplying by
1) cells with the same state.
Visually, in Fig. 6(a) we have the urban land use map, in Fig. 6(b), the neigh-
bor is set to 3 × 3 around the analyzed cell. Figure 6(c) calculates the transition
potential of each cell from its current state to a desired state, the higher is
selected. For this case we set as the higher to urban use. Figure 6(d) analyzed
cell change its value to the higher urban use. Figure 6(e) shows observation win-
dow moves to the next cell. An epoch has been completed when the last cell of
the map is calculated. A simulation may require one or more epochs. If we take
into account that this calculation must be done for each pixel of the map, we
find a problem of computational complexity O(2n ), this means, larger size of the
input maps would increase the execution time of the simulation exponentially.
To handle this complexity we used the OpenMP library.
5 Results
From the model of dynamic system using a celular automata we take as param-
eters to define the states and the evolution space a map previously digitalized
by means of a geographic information system (see Fig. 5). On the map are rep-
resented the land uses of an urban region of northwest Mexico. The Fig. 6 rep-
resents the result of the evolution of the celular automata model with the same
A Dynamic System Model Using Agile Methods Applied to Geomatics 79
land uses for the same period. As can be seen in the evolution of the automaton
there are 10 land uses. These land uses are expressed as user stories incorporating
scenarios defined based on conditions that influence the permanence or change of
state (land use) of the evaluated cell. With the above what we are looking for is
that this can facilitate the characterization and communication with those who
will read such a map, for example, an urban planner. Finally, with this model
it is possible to carry out experiments later for other eras, which will allow to
build territorial scenarios in the future to identify trends in the territorial space
that are taken to obtain the parameters of the model (Figs. 7 and 8).
80 R. Rodrı́guez Zamora et al.
6 Concluding Remark
Urban development and the territorial scenarios that derive from it are topics
of increasing interest. Phenomena such as globalization and migration that pro-
duce among other things, changes in land uses. It is important to look for these
A Dynamic System Model Using Agile Methods Applied to Geomatics 81
References
1. Plata-Rocha, W., Gomez-Delgado, M., Bosque-Sendra, J.: Simulating urban
growth scenarios using GIS and multicriteria analysis techniques: a case study
of the Madrid region, Spain. Environ. Plan. B Plan. Des. 38, 1012–1031 (2011)
2. Li, S., Juhász-Horváth, L., Pedde, S., Pintér, L., Rounsevell, M.D., Harrison, P.A.:
Integrated modelling of urban spatial development under uncertain climate futures:
a case study in Hungary. Environ. Model Softw. 96, 251–264 (2017)
3. Gounaridis, D., Chorianopoulos, I., Koukoulas, S.: Exploring prospective urban
growth trends under different economic outlooks and land-use planning scenarios:
the case of Athens. Appl. Geogr. 90, 134–144 (2018)
4. Ralha, C.G., Abreu, C.G., Coelho, C.G.C., Zaghetto, A., Macchiavello, B.,
Machado, R.B.: A multi-agent model system for land-use change simulation. Env-
iron. Model Softw. 42, 30–46 (2013)
5. Yalew, S., et al.: Land-use change modelling in the Upper Blue Nile Basin. Envi-
ronments 3(3), 21 (2016)
6. Pontius, R.: Criteria to confirm models that simulate deforestation and carbon
disturbance. Land 7(3), 105 (2018)
7. Feng, Y., Qi, Y.: Modeling patterns of land use in Chinese cities using an integrated
cellular automata model. ISPRS Int. J. Geo-Inf. 7(10), 403 (2018)
8. Aguilera Benavente, F., Plata-Rocha, W., Bosque Sendra, J.: Diseño y simulación
de escenarios de demanda de suelo urbano en ámbitos metropolitanos. Rev. Int.
Sostenibilidad, Tecnol. y Humanismo 57–80 (2009)
9. Khoshnoud, F., Dell, D.J., Chen, Y.K., Calay, R.K., de Silva, C.W., Dwhadi,
H.: The concept of self-powered dynamic systems. In: European Conference for
Aerospace Sciences (EUCASS), Munich, Germany, 1–5 July 2013. EUCASS (2013)
10. Borshchev, A.: The big book of simulation modeling: multimethod modeling with
AnyLogic. AnyLogic North America, Chicago (2013)
11. Zamora, R.R., Vergara, S.V.C.: Using de Bruijn diagrams to analyze 1d cellular
automata traffic models. In: Sloot, P.M.A., Chopard, B., Hoekstra, A.G. (eds.)
ACRI 2004. LNCS, vol. 3305, pp. 306–315. Springer, Heidelberg (2004). https://
doi.org/10.1007/978-3-540-30479-1 32
12. Beauchemin, C., Samuel, J., Tuszynski, J.: A simple cellular automaton model for
influenza A viral infections. J. Theor. Biol. 232(2), 223–234 (2005)
82 R. Rodrı́guez Zamora et al.
13. Wolfram, S.: A New Kind of Science, vol. 5. Wolfram Media, Champaign (2002)
14. Brock, Z., Karst, N., Siripong, M.: Bio-inspired computing (2005). ca.olin.edu
15. Jacobson, I., Booch, G., Rumbaugh, J.: The Unified Software Development Process.
Addison-Wesley, Boston (2000)
16. Priolo, S.: Métodos ágiles. USERSHOP (2009)
17. Abrahamsson, P., Salo, O., Ronkainen, J., Warsta, J.: Agile software development
methods: review and analysis. arXiv preprint arXiv:1709.08439 (2017)
18. Cockburn, A.: Agile Software Development. Addison-Wesley Professional, Boston
(2001)
19. Patton, J.: User Story Mapping: Discover the Whole Story, Build the Right Prod-
uct. O’Reilly Media Inc., Cambridge (2014)
20. Trkman, M., Mendling, J., Krisper, M.: Using business process models to better
understand the dependencies among user stories. Inf. Softw. Technol. 71, 58–76
(2016)
21. Barredo, J.I., Gómez, D.J.: Towards a set of IPCC SRES urban land-use scenarios:
modelling urban land-use in the Madrid region, pp. 1–16. European Commission -
DG Joint Research Centre Institute for Environment and Sustainability, Depart-
ment of Geography, University of Alcalá (2000)
22. White, R., Engelen, G.: High-resolution integrated modelling of the spatial dynam-
ics of urban and regional systems. Comput. Environ. Urban Syst. 24, 383–400
(2000)
23. White, R., Engelen, G., Uljee, I.: The use of constrained cellular automata for
high-resolution modelling of urban land-use dynamics. Environ. Plan. B Plan. Des.
24(3), 323–343 (1997)
24. Benavente, F.A., Montes, L.M.V., Lara, J.A.S., Delgado, M.G., Rocha, W.P.: Esce-
narios Y Modelos De Simulación Como Instrumento En La Planificación Territorial
Y Metropolitana. Ser. Geográfica 17, 11–28 (2011)
25. Montgomery, A.W.: Scrum framework effects on software team cohesion, collabora-
tion, and motivation: a social identity approach perspective. Creighton University
(2017)
Telematics and Electronics
Statistical Properties of Vehicle Residence
Times for Fog Computing Applications
1 Introduction
Smart Cities are envisioned to be a major component of people’s everyday lives in big
cities for the near future [1]. This technology will connect a high number of sensors
placed throughout the city to monitor highly sensitive processes in the city, such as,
water supply, gas or toxic substances leakages, pollution levels, temperature, floods,
urban wildlife (rats, dogs, cats, birds, etc.) monitoring [2], and even traffics jams [3]
and criminal acts. Sensors can be integrated into the urban infrastructure but also,
mobile devices which nowadays have many different types of sensors (gyroscopes,
acceleration, sound, video, heart rate monitoring, light, etc.) [4], can be used to gen-
erate and convey relevant data for the city operation.
For instance, urban wildlife monitoring apps can send an alert for people in a
specific point in the city, like a public park, to take sound samples searching for birds of
interest. By recognizing the individual sound characteristics of birds in the environ-
ment. In this application, only users and sensors placed in a certain region are asked to
send their data, while sensors and mobile devices outside this region are not involved in
the data recollection process. This leads to the concept of Fog Computing [5], where
devices in a certain region, inside the fog, communicate and process information that
does not need to reach the cloud computing [6], i.e., this information is locally gen-
erated, stored and transmitted.
Also, relevant in the context of Smart Cities is the development of autonomous
vehicles that will help us with our goal. For years, driver assistance, automation sys-
tems, and communications have been developed to guide us to the next generation of
mobility [7]. The complexity of these developments tries to connect vehicles with other
vehicles or with a controlled environment out of the car. Autonomous vehicles have the
great advantage that the route can be known in advance. Indeed, before driving begins,
the vehicle can calculate the adequate route based on the traffic conditions, driver’s
personal preferences, accidents in the routes, and weather conditions among many
others.
Some works talk about how to make the roads safer and efficient the vehicular
traffic through communications V2V (Vehicle-to-Vehicle) and V2I (Vehicle-to-
Infrastructure) [8], in the last concept it is necessary to use the RSUs (Road-Side
Units), i.e., devices that provide connectivity with vehicles in the range [9], this work
has a similar concept in comparison with this paper but there are differences. We
propose a fixed or mobile sensor to analyze in a determinate area, as Fog Computing, if
a car can or cannot receive data to be used as a data mule without the necessity to have
an RSU because the information can be received by a pedestrian or another mobile
device.
On the other hand, there are advances as VANET’s (Vehicular Ad hoc NETwork)
[10] where a vehicle is considered as a communication unit and vehicles can trade
information without a network or Fog Computing, this work is different than our
propose. We make an emphasis on participation and collaboration between pedestrians
and automobiles.
In this work, we make a study of the communication between vehicles and a sensor
placed in specific parts of the city. Specifically, we consider a system where a sensor,
Statistical Properties of Vehicle Residence Times 87
fixed or mobile, generate information in strategic parts of the city and take advantage of
the passage of autonomous cars. Since the routes of these vehicles are well defined, the
sensor devices can determine if the passing vehicle can be used as a data mule to
convey their data to another part of the city. Although, in this paper, the main goal is to
analyze if the interaction between vehicles and a sensor, fixed or mobile, can be able
viable or not.
To characterize these residence times, we use real vehicle trajectories from a well-
known database on SUMO (Simulation of Urban MObility) [11] simulator about the
traffic of vehicles in different parts of the city to analyze the behavior of vehicles in
different traffic conditions.
SUMO [12] is a software that can provide a vehicular scenario of a real-world
working with JOSM (Java OpenStreetMap) that provides the updated map of the city.
For this analysis, we work with an article published by the University of Luxem-
bourg [13] called “Luxembourg sumo traffic (lust) scenario: 24 h of mobility for
vehicular networking research.” where it is possible to obtain a simulation scenario in a
period of 24 h of traffic from Luxembourg. This work was compared with the Typical
Traffic option in Google Maps and showed similar traffic with the real traffic in
Luxembourg.
Then, we calculate the times that an automobile is in a street, waiting in a traffic
light or moving with a specific speed. The position and the behavior of a vehicle are
important parameters used to calculate the residence times of vehicles in a fog com-
puting application. Luxembourg is a typical European city, and we assume that the
results presented in this paper can be extended to other cities with similar conditions.
The realistic scenario was measured in a common day, during the morning the
traffic is high density, in the afternoon is medium density and, in the night, the density
is lower than other times of the day. In this scenario, there are vehicles, buses, and
pedestrians that have different behaviors.
The rest of this paper is organized as follows: Sect. 2 presents the main assump-
tions and system set-up were the residence times where measured. Then, in Sect. 3 we
explain how the characteristics of the residence times were obtained. Finally, Sect. 4
presents relevant results.
2 System Model
We can observe that both images are similar, but some routes do not exist in the
SUMO map. Then, based on both maps, we selected specific places to analyze
vehicular concurrence. These points represent parts of a smart city scenario where
information can be generated. Hence, all sensors (fixed and mobile devices) inside the
range of interest will form the Fog Computing network, i.e., data is sensed, stored,
processed and conveyed to passing vehicles to then transmit it to other parts of the city.
Also, note that this information is not relevant to other parts of the city. As such, this
information does not have to reach the cloud, effectively reducing the traffic and
resource utilization in the cloud. For instance, a WiFi-based network has a range of
about 100 m and Bluetooth about 10 m. Both these networks are typically enabled in
any commercial smartphone and sensor nodes. We also considered ranges of 50 m and
100 m for completeness purposes, then we make a radius of 25 m and 50 m around the
point to simulate a network. In Table 1, we showed the selected places.
The first place is a bridge and is an intersection between Montée de Clausen street
and under Rue Sosthéne Weis street, this location is near Casemates du Bock that is an
antique architecture to defend the city called Schlassbréck bridge. As seen in Fig. 2, the
radius is in red and, this circle simulates the network, i.e., all vehicles that enter on this
circle is considered as potential data mule to convey data to other parts of the city.
Fig. 2. The first place to analyze in SUMO map and Google map.
The second point of interest is shown in Fig. 3 and is located near Justice Palace,
this place is between Franklin Delano Roosevelt boulevard and René Konen tunnel.
Fig. 3. The second place to analyze in SUMO map and Google map.
The third point of interest is depicted in Fig. 4 and is near the Art Gallery Schorten
and is located between Emile Reuter avenue and Royal boulevard.
The fourth point of interest is shown in Fig. 5 is near a hotel called Sofitel
Luxembourg Le Grand Hotel located in d’Avranches boulevard and Rue du Labora-
toire street.
90 D. Miguel Santiago et al.
Fig. 4. The third place to analyze in SUMO map and Google map.
Fig. 5. The fourth place to analyze in SUMO map and Google map.
The last place is near the International School of Luxembourg Asbl and the location
is in Rue de Bouillon boulevard and it is shown in Fig. 6.
Fig. 6. The last place to analyze in SUMO map and Google map.
Statistical Properties of Vehicle Residence Times 91
Once the interest points are selected and the different communication ranges are placed
around it, we now can calculate the residence times of all vehicles going through the
circles described in the previous section. We analyze the information making a program
in python because SUMO generates files with XML format. These files have several
data characteristics about the vehicles but the information that we need is the position
and time of each vehicle. The position is given as cartesian coordinates and, as men-
tioned before, we consider periods of 10 min because it was difficult to analyze the
24 h. The 10 min period is equal to 600 s; during this time, we analyze the position of
each vehicle second inside the coverage data.
To find vehicles in a circumference we used the circumference equation given by:
x2 þ y2 ¼ r 2 ð1Þ
Then, we added our coverage range to localize the vehicles inside the interest points
as follows:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
50 ðx OrigxÞ2 þ ðy þ OrigiyÞ2 ð2Þ
or
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
100 ðx OrigxÞ2 þ ðy þ OrigiyÞ2 ð3Þ
Where x and y are the coordinates of each vehicle and Origx and Origy are the interest
points and all vehicles present in a radius equal or minor to 50 or 100 meters are
considered in this analysis.
In this work, we are only interested in studying the statistical characteristics of this
residence time. Then, we calculate the mean, variance and coefficient of variation of
these residence times. The main data that we are going to use is the coefficient of
variation because is a relation between the size of the mean and the standard deviation.
r
CV ¼ ð4Þ
l
5. In the case that the CV < 1, the possible distribution that can describe the residence
times is the Erlang distribution. This distribution describes the wait time until an
event occurs.
4 Numerical Results
We now present some relevant results of the residence times. In Figs. 7 and 8 we
present the histograms of the residence times in the first point of interest considering
low vehicle density with 50 and 100 m of coverage radio respectively.
For the second point of interest, we present the histograms in Figs. 9 and 10.
For the third point of interest, we present the histograms in Figs. 11 and 12.
Statistical Properties of Vehicle Residence Times 93
For the fourth point of interest, we present the histograms in Figs. 13 and 14.
Fig. 14. Histogram in Sofitel Luxembourg Le Grand Hotel with 100 m radius.
Finally, for the fifth point of interest, we present the histograms in Figs. 15 and 16.
In each figure, we show the mean (µ), the variance (r2 ) and the coefficient of
variation (CV). As we mentioned before, the CV is part of our interest since we can
choose a hypothetical distribution that matches the statistical characteristics of the
residence times. Building on this, we can see that the CV of all residence times
measured in this work are lower than 1. Hence, the Erlang distribution can be used to
model the times that the vehicles remain in the interest region.
96 D. Miguel Santiago et al.
Fig. 16. Histogram in International School of Luxembourg Asbl with 100 m radius.
In this work, we characterized the residence time of real traces of vehicles inside an
interest region in Luxembourg City. To this end, we first locate touristic locations that
are potential points where data can be gathered and disseminated for a future smart city
scenario and autonomous vehicles, although if we could select several locations maybe
we would expect similar results.
Another relevant result is that for the scenarios considered, an Erlang distribution is
a good candidate distribution to model the residence times for future Fog Computing
applications in smart city environments, we could link with the real results and test if
these results are real and not a coincidence. We must prove mathematically.
If we could analyze all data, we would expect similar results, because we consider
that vehicles have the same behavior.
These results can be used for future mathematical and simulation works to calculate
the amount of information that can be conveyed to vehicles acting as data mules to
transport information to other parts of the city in a low-cost scenario, it depends on the
future works.
References
1. Albino, V., Berardi, U., Dangelico, R.M.: Smart cities: definitions, dimensions, performance,
and initiatives. J. Urban Technol. 22(1), 3–21 (2015)
2. Khan, Z., Anjum, A., Soomro, K., Tahir, M.A.: Towards cloud based big data analytics for
smart future cities. J. Cloud Comput. 4(1), 2 (2015)
Statistical Properties of Vehicle Residence Times 97
3. Biswas, S.P., Roy, P., Patra, N., Mukherjee, A., Dey, N.: Intelligent traffic monitoring
system. In: Satapathy, S.C., Raju, K.S., Mandal, J.K., Bhateja, V. (eds.) Proceedings of the
Second International Conference on Computer and Communication Technologies. AISC,
vol. 380, pp. 535–545. Springer, New Delhi (2016). https://doi.org/10.1007/978-81-322-
2523-2_52
4. Ong, R., et al.: Traffic jams detection using flock mining. In: Gunopulos, D., Hofmann, T.,
Malerba, D., Vazirgiannis, M. (eds.) ECML PKDD 2011. LNCS (LNAI), vol. 6913,
pp. 650–653. Springer, Heidelberg (2011). https://doi.org/10.1007/978-3-642-23808-6_49
5. Stojmenovic, I., Wen, S.: The fog computing paradigm: scenarios and security issues. In:
2014 Federated Conference on Computer Science and Information Systems, pp. 1–8. IEEE
(2014)
6. Capponi, A., Fiandrino, C., Kliazovich, D., Bouvry, P., Giordano, S.: A cost-effective
distributed framework for data collection in cloud-based mobile crowd sensing architectures.
IEEE Trans. Sustain. Comput. 2(1), 3–16 (2017)
7. Sookhak, M., et al.: Remote data auditing in cloud computing environments: a survey,
taxonomy, and open issues. ACM Comput. Surv. (CSUR) 47(4), 65 (2015)
8. Aissaoui, R., Menouar, H., Dhraief, A., Filali, F., Belghith, A., Abu-Dayya, A.: Advanced
real-time traffic monitoring system based on V2X communications. In: 2014 IEEE
International Conference on Communications (ICC), pp. 2713–2718. IEEE (2014)
9. Richter, A., Friedl, H., Scholz, M.: Beyond OSM–alternative data sources and approaches
enhancing generation of road networks for traffic and driving simulations. In: SUMO2016–
Traffic, Mobility, and Logistics, pp. 23–25 (2016)
10. Ge, X., Li, Z., Li, S.: 5G software defined vehicular networks. IEEE Commun. Mag. 55(7),
87–93 (2017)
11. Kastner, K.H., Pau, P.: Experiences with SUMO in a real-life traffic monitoring system. In:
SUMO 2015–Intermodal Simulation for Intermodal Transport (2015)
12. Liu, W., Wang, X., Zhang, W., Yang, L., Peng, C.: Coordinative simulation with SUMO and
NS3 for vehicular ad hoc networks. In: 2016 22nd Asia-Pacific Conference on Commu-
nications (APCC), pp. 337–341. IEEE (2016)
13. Codeca, L., Frank, R., Engel, T.: Luxembourg sumo traffic (LuST) scenario: 24 hours of
mobility for vehicular networking research. In: 2015 IEEE Vehicular Networking
Conference (VNC), pp. 1–8. IEEE (2015)
Text and Image Transmission and Reception
Using Light from LEDs and a Light Sensor
Sergio Sandoval-Reyes(&)
Abstract. Visible Light Communication or VLC uses visible light from light
emitting diodes (LEDs) to transmit information. Using a computing device and
some hardware, the transmission of information is performed driving and
modulating the light emitted by the LEDs. In the receiver side, the information
carried by the modulated light is demodulated through a light sensor, connected
to a similar computing device for the final recovering of the information. In this
article we describe an application based on VLC using OOK (On-Off Keying)
modulation, to transmit and receive a line of text and an image from a Raspberry
Pi computer (using Python as the programming language), two yellow LEDs
and a sensor light.
1 Introduction
Visible Light Communication [1, 2] can be used to transmit audio, voice and data using
light from emitter diodes (LEDs), and light detectors at the transmitter and receiver
ends respectively (Fig. 1). VLC works in the 380 nm to 780 nm optical band which is
visible light [3–5].
To convey information this one has to be encoded, and then the light has to be
modulated and demodulated at the transmitter and receiver sides. There are several
methods to do this, some are briefly discussed in the following. Then, the received
information has to be decoded and processed to recover it fully. The success of this
recovery depends of several factors, among them: (1) The number, shape, and wave-
length of the LEDs employed; (2) The number and type of light detectors (light-sensor,
PIN-diode, reverse-biased LED, etc.) used; (3) The encoding method (RZ, NRZ,
NRZP, etc.); (4) The modulation scheme (OOK, WPM, VWPM, PPM, OFDM, etc.),
and (5) The synchronization and distance between the LEDs and the light detector.
VLC is a technology that requires line-of-sight between both emitter and receiver [6].
In this work we describe an application based on VLC using OOK (On-Off Keying)
modulation, to transmit a line of text and a color image using a Raspberry Pi 3
computer as the data source and sink (to simplify the synchronization problem between
emitter-receiver), Python as the programming language, and several modules (LEDs
and a sensor light) from littleBits, to easy the hardware implementation.
2 Related Work
Several research works on VLC technologies to transmit text and images have been
proposed. The most important are described in the following.
In this section we develop a VLC application to transmit and recover a line of text first,
and then a color image using two LEDs in the transmitter, and a light sensor in the
receiver.
We begin using in the transmitter as a data source, a Raspberry Pi 3 (RBPi)
computer, and six littleBits bit-modules [12]: power, button, proto, split, and two bright
LEDs. See Fig. 6. The RBPi using a software program written in Python will read a
line of text or an image byte by byte, and will transmit each byte using the SPI MOSI
(Serial Peripheral Interface, Master Output Slave Input) output port (Pin 19), toward the
proto module, OOK signals to drive the two LEDs.
Similarly, in the receiver we will use three littleBits components: power, light
sensor, and another proto bit. See Fig. 7. The light sensor captures the light emitted by
the two LEDs and converts it into a digital signal which is fed to the proto module,
which in turns outputs this signal and with a wire connector, feeds this signal toward
the MISO (Master Input Slave Output) input port (Pin 21) of the RBPi.
102 S. Sandoval-Reyes
Fig. 6. Raspberry Pi 3, and six modules: power, button, proto, split and two LEDS
# hola_lifi.py
# Python code to transmit and recover the line of text “hola Li-Fi” using SPI and VLC
import spidev
spi = spidev.SpiDev()
spi.open(0,0)
spi.max_speed_hz = 4000000
buffer = bytearray()
tx_data = [ord(H”), ord(“o”), ord(“l”), ord(“a”), ord(“ “), ord(“L”), ord(“i”), ord(“-“), ord(“F”),
ord(“i”)]
while True:
# Transmit the text “Hola-Li-Fi”, throught the Raspberry out pin 19 SPI-MOSI,
# and receive it in the array rx_data, throught the Raspberry input pin 21 SPI-MISO
rx_data = spi.xfer(tx_data)
# Receive every byte in rx_data and store them in a buffer to display them in the screen
for byte in rx_data:
buffer.append(chr(byte))
print buffer
Fig. 8. Python code to transmit and recover a line of text via VLC.
Transmitting an image via VLC is very similar to transmit a line of text. Here the
RBPi with a script written in Python, opens a picture file (“lenna.jpeg”) and it reads it
into a bytearray “b”. Then with a “for” loop reads byte by byte of the picture and as
before, it sends them to the SPI MOSI output port 19, using the spi.xfer () directive.
This MOSI output is fed into the input of the proto module (lower middle connector in
Fig. 7). As before, the proto module outputs and splits the OOK signals to drive the
two LEDs; the LEDs light is picked up throughout the light sensor to recover the
image. The Python program to execute the above mentioned is shown in Fig. 9.
In the code above, it is necessary to import the following libraries: “SPI”, “PIL”
(Python Image Library), and “Array”. To drive the LEDs the power module was fed
with a 9-V battery. This was necessary because the outputs of the RBPi are low-voltage
(3.3 V) and low-current (Individual pins must not pull more than 16 mA and the entire
GPIO must not source more than 50 mA), which are no good enough to drive two
bright LEDS [13]. These LEDs are simple yellow LEDS with a wavelength of 550-to-
600 nm, and consume around 16-to-20 mA each, with an aperture angle of about 120°.
See Fig. 10.
# lena_spy.py: Python code to read, transmit, and recover a picture via Visible Light
import spidev
from PIL import Image, ImageFilter
import io
from array import array
spi = spidev.SpiDev()
spi.open(0,0)
spi.max_speed_hz = 4000000
b = bytearray()
buffer = bytearray()
try:
# Load an image from Raspberry Pi
with open (“lenna.jpeg”, “rb”) as img:
f = img.read()
b = bytearray(f)
# Sending with SPI and converting the image, byte by byte into visible light
for byte in b:
rx = spi.xfer([byte])
rx_data = rx[0]
buffer.append(rx_data)
# Recovering the bytes array and back into an image
img_rec = Image.open(io.BytesIO(buffer))
img_rec.save(´lena_img_recovered.png´)
img_rec.show()
except:
print “Unable to recover image”
Fig. 9. Python code to read, transmit and recover a picture via VLC.
Fig. 7), as the light shining on the sensor gets brighter, more signal passes through it.
In DARK mode (switch down), the signal increases as it gets darker. Furthermore, the
light sensor has a sensitivity dial or slide dimmer to adjust how much light it takes to
recover the signal, and has a spectral sensitivity range from 500-to-600 nm similar to
the LEDs wavelength. See Fig. 11.
The recovered text and image from the light sensor via the SPI MISO input (pin 21)
is stored into a buffer, saved and displayed, as can be seen from the Python code lines
of Figs. 8 and 9.
Text and Image Transmission and Reception Using Light from LEDs 105
For the experiments we use as was mentioned a RaspBerry Pi 3 computer with several
LittleBits components. The whole setup is shown in Fig. 12.
Fig. 12. Setup for the transmission and reception of text and an image using VLC.
Figure 12 shows the whole set up: the Raspberry Pi 3 computer, a HP display,
keyboard and mouse, thus as the LEDs, the sensor light, and the proto modules, fed
with 9 V batteries. Figure 13 shows the recovered text and picture after the execution
of the line commands “sudo python hola_lifi.py” and “sudo python lena_spi.py”.
Figure 13.
Figure 13 also shows that the recovered text and picture was no good enough. That
was due to the presence of noise, mainly: Fluorescent light from ceiling lamps,
106 S. Sandoval-Reyes
misalignment and distance between the LEDs and the light sensor, and low sensitivity
to light from the light sensor.
4.3 Contributions
This work differs with respect to similar ones reported in [8–11], and [14–18], in the
following. In some of them only a design model is proposed without any proof of
implementation. In others a hardware setup is shown, but again, there is not any proof
of sending and receiving any image. It also required to use two microcontrollers (one
for transmission and one for reception). The main problem of using microcontrollers is
that not all of them run good enough fast and have enough RAM memory to store a
medium to large image sent by the PC. Furthermore, it requires software written in C or
Matlab to reside at both PCs to connect with the microcontrollers and to process the
picture transmission and reception. Some others, do not use visible light at all, but
infrared light. Our work differs from the above in that we only use a Raspberry Pi
computer, visible light from LEDs, littleBits modules, and simple code written in
Python as the programming language.
108 S. Sandoval-Reyes
5 Conclusions
A line of text and an image transmission and reception application using visible light
was developed using a Raspberry Pi 3, two bright LEDs and a light sensor, thus as
OOK modulation, and the Python Image Library. Both the line of text and the picture
in color, were recovered acceptably well although with a small presence of noise. It
should be noted that this noise is due to environment light, and the distance between the
LEDs and the light sensor.
We concluded that the performance of the application depends on several factors:
(1) The brightness of the LEDs: (2) The sensitivity of the light sensor with respect to
the light received from the LEDs; (3) The surrounding light (in Fig. 7 it can be seen
that the light sensor has a control for graduating the sensitivity in the presence of high
or low light); and (4) The alignment between the LEDs and the light sensor influences
the reception, and consequently the quality of the image and text recovery.
Acknowledgments. This work was supported by Instituto Politecnico Nacional Project SIP-IPN
20196133.
References
1. Haas, H.: Wireless data from every light bulb. TED: Ideas worth spreading (2011). http://
www.ted.com/talks/harald_haas_wireless_data_from_every_light_bulb
2. Tsonev, D., Videv, S., Haas, H.: Light fidelity (Li-Fi): towards all-optical networking. In:
Proceedings of SPIE (Broadband Access Communication Technologies VIII), vol. 9007, no.
2 (2013). https://doi.org/10.1117/12.2044649
3. Sherman, J.: How LED light bulbs could replace Wi-Fi. Digital Trends (2016). http://www.
digitaltrends.com/mobile/light-bulb-li-fi-wireless-internet/
4. Haas, H.: High-speed wireless networking using visible light. SPIE Newsroom. https://doi.
org/10.1117/2.1201304.004773, http://www.spie.org/newsroom/4773-high–speed-wireless-
networking-using-visible-light. Accessed March 2016
5. Vincent, J.: Li-Fi revolution: internet connections using light bulbs are 250 times faster than
broadband (2016). http://www.independent.co.uk/news/science/li-fi-revolution-internet-
connections-using-light-bulbs-are-250-times-faster-than-broadband-8909320.html
6. Location Awareness Wikipedia (2018). https://en.wikipedia.org/wiki/Location_awareness
7. Jovicic, A., Junyi, L., Richardson, T.: Visible light communication: opportunities, challenges
and the path to market (2013). https://doi.org/10.1109/mcom.2013.6685754, http://
ieeexplore.ieee.org/document/6685754/
Text and Image Transmission and Reception Using Light from LEDs 109
8. Mahendran, R.: Integrated LiFi (Light Fidelity) for smart communication through
illumination. In: International Conference on Advanced Communication Control and
Computing Technologies (ICACCCT), pp. 53–56. https://doi.org/10.1109/icaccct.2016.
7831599, https://ieeexplore.ieee.org/document/7831599/?reload=true
9. Shah, V., et al.: 2D image transmission using light fidelity technology. Int. J. Innov. Adv.
Comput. Sci. (IJIACS) 4(4), 121–126 (2015). https://www.academicscience.co.in/admin/
resources/project/paper/f201504241429887963.pdf, ISSN 2347-8616
10. Lee, K., Park, H.: Modulations for visible light communications with dimming control. IEEE
Photonics Technol. Lett. 23(16) (2011). https://doi.org/10.1109/lpt.2011.2157676, http://
ieeexplore.ieee.org/document/5773477/
11. Perfecto, G.C., Godoy, R.S., Rios, A.A.: Transmision de imagenes de Matlab a Arduino via
Puerto Serial. Boletín UPIITA-IPN, 644 CyT, No. 52 (2016). http://www.boletin.upiita.ipn.
mx/index.php/ciencia/644-cyt-numero-52/1234-transmision-de-imagenes-de-matlab-a-
arduino-via-puerto-serial
12. LittleBits. https://www.littlebits.cc/
13. Pi, R.: Raspberry Pi input and output pin voltage and current capability. Mosaic
Documentation Web. http://www.mosaic-industries.com/embedded-systems/microcontroller-
projects/raspberry-pi/gpio-pin-electrical-specifications
14. Vasuja, M., et al.: Image transmission using Li-Fi. In: IEEE Conference Publications (2018).
https://ieeexplore.ieee.org/document/8473033
15. Raju, K.S., et al.: PC to PC transfer of text, images using visible light communication (VLC).
Int. J. Adv. Eng. Manag. Sci. (IJAEMS) (2017). https://dx.doi.org/10.24001/ijaems.3.5.6
16. Hamand, A., Kuntawar, S.: An approach towards high speed communication using Li-Fi
technology. IOSR J. Electron. Commun. Eng. (2018). http://www.iosrjournals.org/iosr-jece/
papers/Vol.%2013%20Issue%204/Version-1/K1304017275.pdf
17. Chen, M., et al.: Real-time video transmission of visible light communication based on LED.
Int. J. Commun. Netw. Syst. Sci. 10(8B) (2017). https://doi.org/10.4236/ijcns.2017.108b007,
https://www.scirp.org/journal/PaperInformation.aspx?PaperID=78373
18. Costa Mari, J.: Visible light communication for Internet of things. Trabajo final de grado,
Universitat Politecnica de Catalunya (2018). https://upcommons.upc.edu/handle/2117/
117811
12–18 GHz Microwave Frequency Band
Microstrip Patch Antenna Design for the Radio
Implant Medical Devices Application
Abstract. The implantation of radio implant medical devices (IMD) into the
human body are becoming smaller, therefore and because to the high degree of
miniaturization and an efficient wireless link, antennas extremely small, which
satisfies gain, bandwidth, impedance, and low cost are demanded. This work
proposes a wide bandwidth ranging from 12 GHz to 18 GHz Microstrip Patch
Antenna, describing the design, simulation, implementation and measurement,
in order to be used on the IMD’s.
1 Introduction
with respect to the wavelength is the most important parameter influencing the
antenna’s radiation [3], and a relation between antenna size and its radiation charac-
teristics is reported in [7] concluded that geometrically small but efficient antennas are
required for high operating frequencies, and the antenna efficiency significantly
decreases if lower operating frequencies are used. Likewise, implantable antennas
operating at much higher frequencies (31.5 GHz) have also been reported in the lit-
erature [8].
On the other hand, inductive links formerly have been employed for the wireless
telemetry of IMDs [9–11], but this kind of inductive low-frequency links shows low
data rates, short communication range and unwanted emissions. Actually ITU-R
Recommendation SA.1346 (ITU-R 1998) which outlined the use of the 402.0–
405.0 MHz frequency band for medical implant communication services (MICS), in
this sense, at 402 MHz–405 MHz, the wavelength of an electromagnetic wave is
approximately 74 cm, as we know, the size of the microstrip antenna is directly pro-
portional to the wavelength at the operating frequency, consequently an antenna with
dimensions comparable to this wavelength cannot be used for an implant [12], so,
different antenna structures that need to be addressed. In this sense, due to higher
operation frequencies allow use of smaller-sized antennas and components, and without
wanting to enter a conflict with ITU-R regulations, for this purpose the designed
antenna in this work, is a just a proposal to be considered, due to wavelength corre-
sponding to this frequency band (12–18 GHz), ranges between 25 mm to 16 mm.
Only for the sole purpose to observe the dimensions of the proposed microwave
band patch antenna, this one was placed close to a RFID UHF Band meander antenna
(915 MHz) [13], (see Fig. 1).
Fig. 1. (a) RFID UHF Band meander antenna (915 MHz) and (b) Proposed microwave band
patch antenna (12–18 GHz).
Therefore, in view of the importance of their health role and because the body
directly degrades the performance of the antenna, these kind of antennas should be
safety, lightweight, extremely small, robust structure, it must be considered the inter-
action with the human body, in this sense, there are complex factors involved that have
112 S. R. Meneses González and R. T. Rodríguez Márquez
to be considered when dealing with this issue; [14] describes considerations related to
the design and performance of implantable patch antennas. In this work, first part, it is
limited to selection of the frequency band of operation, miniaturization of the antenna,
the other issues relative to biocompatibility of the antenna, will be supported on a
further work (part 2) in this area.
In this way, in this work, we propose a small slotted patch antenna operating in 12–
18 GHz Microwave Band, in order to reduce the size antenna, increase the bandwidth,
and improve the quality communication of the implantable antenna with exterior
monitor/control equipment.
To determine the performance of design parameter, as impedance, resonance fre-
quency, radiation pattern, wideband, etc., CST software (Computer Simulation Tech-
nology) has been used, in the same way, experimental tests have been applied.
The paper is organized as follows: Sect. 2 describes a brief tag antenna design
basis, discusses simulation, Sect. 3 describes measurements results, and finalizing the
present work with conclusions and references.
2 Antenna Design
simple, low cost, wide bandwidth, without the need of altering of substantial changes to
the structure, maintain the small size.
Considering a design frequency equal to 15 GHz, this being the core of the pro-
posed band, the wavelength is equal to k = 20 mm. As we know, the size of the
microstrip antenna is directly proportional to the wavelength at the operating fre-
quency, in this sense, the following expressions [16] have been used, in such a way
that, the substrate thickness h = 1.524 mm, strip width, W = 9 mm, which is
approximately equal to k=2.
Structure geometry and dimensions are shown in Figure.
Strip width:
rffiffiffiffiffiffiffiffiffiffiffiffi
1 2
W ¼ pffiffiffiffiffiffiffiffiffi ð1Þ
2fr 0 l0 r þ 1
Strip length:
1
L¼ pffiffi pffiffiffiffiffiffiffiffiffi 2DL ð2Þ
2fr e e0 l0
Effective Permittivity:
0 1
r þ 1 r 1 B 1 C
e ¼ þ @qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiA ð3Þ
2 2 1 þ 12 h
W
where:
fr , resonance frequency
er , relative permittivity
h, substrate thickness.
results gives the chance to identify the necessary dimension modifications, in order to
adjust the resonance frequency value along the 12–18 GHz. Simulated radiation pattern
(see Fig. 5) shows the vertical plane (E| vs. h), called E-Plane and the 3D simulation
pattern (See Fig. 6).
In the same way, it is possible to observe an omnidirectional performance, with no
nulls, concentrated electromagnetic power on the −90°–90° plane, that is, the front side
of the antenna, due to ground plane to its rear. This radiation pattern geometry is an
12–18 GHz Microwave Frequency Band Microstrip Patch Antenna Design 115
advantage, because of point to point communication is required for this application and
certain area close to the body skin can be covered.
3 Experimentation
In order to build the prototype, we have used Copper slotted by tracks engraved in
DUROID-type substrate with a width of 1.5 mm, which electric permittivity is er =
2.2, and a SMA connector. It should be noted that we have used and ordinary SMA
connector, due to our means in a cost-effective way. Use a better-quality and appro-
priate connector to work at this frequency band, improve the antenna efficiency (see
Fig. 7).
In the same way, we have used a Vector Network Analyzer ZVB 40 is calibrated in
the band 10 GHz–20 GHz, at short circuit, opened circuit and matching network (see
Fig. 8) in order to measure S11 parameter.
From the magnitude vs. frequency graphic achieved (see Fig. 9), it is possible to
observe that the maximum antenna efficiency is achieved at 12.86 GHz, resonance
frequency, point M2 and at approximately 13.2 GHz is the minimum efficiency. The
bandwidth is approximately equal to 6 GHz, starting in 12.11 GHz (point M1) and
ending in 18.5 GHz, that is, the designed antenna is a wideband antenna, bandwidth
enough to support high data rate.
On the other hand, literature review present other architectures antennas, in the
MICS (402.0–405.0 MHz) and ISM (433.1–434.8, 868.0–868.6 and 902.8–
928.0 MHz) bands [18]. The proposed antenna is novel, useful, innovative antenna, in
the microwave band (12–18 GHz), small dimensions and wide bandwidth.
12–18 GHz Microwave Frequency Band Microstrip Patch Antenna Design 117
3.1 Limitation
The method used to measure and calculate the gain antenna and radiation pattern,
consists of install the prototype antenna, into an anechoic chamber, acting as receiver
antenna, and a known second antenna, a commercial antenna acting as a transmitter
118 S. R. Meneses González and R. T. Rodríguez Márquez
Fig. 9. Magnitude vs. Frequency, S11 parameter (prototype antenna under experimental test).
antenna, both of them installed into the anechoic chamber, spaced 2.8 m apart, keeping
a line of sight (see Fig. 10), however, actually, we are not be able to carry out this
measurement due to our laboratory does not have capabilities to experiment at this band
frequency (transmitter and receiver equipment).
4 Conclusion
A microstrip patch antenna array has been designed, meeting the resonance frequency,
small size (millimetric), exhibiting a wide bandwidth (12–18 GHz), an appropriate
geometry of the radiation pattern, low cost, and easily construction.
There are differences between simulation and measurement results on the S11,
respect to the maximum resonance frequency, 15 GHz and 12.86 GHz, respectively,
due to certain construction defects, for instance, bad soldier, SMA connector, or adjust
the exact dimensions, however, the experimental bandwidth result was far better than
the simulated bandwidth result, due to the simulation bandwidth result is equal to
1.5 GHz and the experimental bandwidth result is approximately equal to 6 GHz.
The second part of this work consists of to continue the technical stage (gain
measurement), as soon as we get the chance, as soon as the appropriate equipment at
this band is in place, and, verify the antenna effectiveness (distance, penetration depth),
testing the antenna performance on/under human or animal skin, it will be necessary
the support biomedical research.
Finally, due the obtained results, the proposed antenna is a novel, useful and
innovative antenna for IMD’s application, in the microwave band, wide bandwidth,
small dimensions, easily construction, meeting the resonance frequencies, input
impedance requirements, an appropriate geometry of the radiation pattern, so that, we
recommend this kind of small antenna as well as the proposed frequency band to be
applied to Radio Implant Medical Devices corresponding to the health sector.
References
1. Xuyang, L.: Body Matched Antennas for Microwave Medical Applications. KIT Scientific
Publishing, Karlsruhe (2014)
2. Soontornpipit, P., Furse, C.M., Chung, Y.C.: Design of implantable microstrip antenna for
communication with medical implants. IEEE Trans. Microw. Theory Tech. 52(8), 1944–
1951 (2004). Part 2
3. Kim, J., Rahmat-Samii, Y.: Implanted antennas inside a human body: simulations, designs,
and characterizations. IEEE Trans. Microw. Theory Tech. 52(8), 1934–1943 (2004). Part 2
4. Norris, M., Richard, J.-D.: Subminiature antenna design for wireless implants. In:
Proceedings of the IET Seminar on Antennas and Propagation for Body-Centric Wireless
Communications, London, UK, April 2007, pp. 57–62 (2007)
5. Hall, P.S., Hao, Y.: Antennas and Propagation for Body-centric Wireless Communications.
Artech House, Norwood (2006)
6. Skrivervik, A.K., Zürcher, J.F., Staub, O., Mosig, J.R.: PCS antenna design: the challenge of
miniaturization. IEEE Antenna Propag. Mag. 43(4), 12–27 (2001)
7. Sánchez-Fernández, C.J., Quavado-Teruel, O., Requena-Carrión, J., Inclán-Sánchez, L.,
Rajo-Iglesias, E.: Dual-band microstrip patch antenna based on short 0 circuited ring and
spiral resonators for implantable medical devices. IET Microw. Antennas Propag. 4, 1048–
1055 (2010)
8. Ahmed, Y., Hao, Y., Parini, C.: A 31.5 GHz patch antenna design for medical implants. Int.
J. Antennas Propag. 2008, 6, Article ID 167980 (2008). Hindawi Publishing Corporation
120 S. R. Meneses González and R. T. Rodríguez Márquez
9. Freedonia: Implantable Medical Devices – Industry study with forecast for 2015 & 2020.
The Freedonia Group, Study #2852 (2012)
10. Valdastri, P., Menciassi, A., Arena, A., Caccamo, C., Dario, P.: An implantable telemetry
platform system for in vivo monitoring of physiological parameters. IEEE Trans. Inform.
Technol. Biomed. 8, 271 (2004)
11. Baghel, S., Yadav, N.K.: A review on micro electronic pill. Int. J. Res. Appl. Sci. Eng.
Technol. (IJRASET), 3(VI) (2015)
12. Kiourti, A., Nikita, K.S.: Implanted antennas in biomedical telemetry. In: Chen, Z.N., Liu,
D., Nakano, H., Qing, X., Zwick, T. (eds.) Handbook of Antenna Technologies, pp. 2613–
2652. Springer, Singapore (2016). https://doi.org/10.1007/978-981-4560-44-3_94
13. Meneses, R., Montes, L., Morales, M.: A small note about RFID technology applied to
public safety. Part 1: RFID tag antenna design. In: Conference: Power, Electronics and
Computing (ROPEC), 2013 IEEE International Autumn Meeting 2013, Ixtapa, Gro. México,
November 2013
14. Li, X.: Body Matched Antennas for Microwave Medical Applications. Forschungsberichte
aus dem Institut für Höchstfrequenztechnik und Elektronik (IHE) der Universität Karlsruhe
(TH). KIT Scientific Publishing (2015). ISSN 0942-2935
15. Madhav, B.T.P., et al.: Multiband slot aperture stacked patch antenna for wireless
communication applications. Int. J. Comput. Aided Eng. Technol. 8(4), 413–423 (2016)
16. Wong, K.: Compact and Broadband Microstrip Antennas. Wiley, Hoboken (2002)
17. CST Homepage. https://www.cst.com/Academia/Student-Edition
18. Kiourti, A., Nikita, K.S.: Performance of miniature implantable antennas for medical
telemetry at 402, 433, 868 and 915 MHz. In: Godara, B., Nikita, K.S. (eds.) MobiHealth
2012. LNICST, vol. 61, pp. 122–129. Springer, Heidelberg (2013). https://doi.org/10.1007/
978-3-642-37893-5_14
Analytics for Basic Products in Mexico
1 Introduction
Presently, the data of basket products have been studied from a wide range of per-
spectives mainly in financial and market aspects. Traditionally these approaches are
dominated for statistical approaches. However, new opportunities of analyzing and
studying these phenomenons from the perspective of data science, Big Data, and
machine learning ara available today to discover or identify insights in such data.
For another hand, according to the National Survey of Household Income and
Expenses (INEGI by its acronym in Spanish) [1]. The average household expenditure
in Mexico is invested in products of a basic basket. In light of this knowledge, it is
natural to be interested in analyzing the behavior of these products. The possible results
of the data analysis of these products have a valuable and useful character. Future price
forecasts can help commercial companies in the elaboration of control and financial
planning processes. Governments can also benefit from this type of study for the
creation of public policies.
2 Methodology
The data methodology proposed is composed of 4 phases: (1) ETL process, (2) Data
Continuity, (3) Categorical Data transformation and (4) Feature Selection. Each of one
are developed in the following sections.
Fig. 1. Time continuity of milk powder records in the dimensions of product, presentation,
brand.
The above is consistent since the products are being taken in a grouped manner,
therefore characteristics such as product and presentation are very important. When the
same feature selection algorithm is applied to certain products and certain presentations
individually, other features such as address become more important. It is shown in
Table 3.
Pm
ðyi ^yi Þ2
R2 ¼ Pi¼1
m
ð1Þ
i¼1 ðyi yi Þ 2
Looking at the previous formula it can be seen that the coefficient of determination
is based on knowing how good the predictions generated by the model are, compared
with simply having used the average of the samples as a prediction. If there is a larger
standard deviation, there will be a larger error in the lower part of the equation, which
will minimize the error produced by the model. Therefore, it is vitally important to
know if this standard deviation is due to the variation of a product over time or if it is
due to grouping, for example, different products and presentations. To clearly visualize
this situation, a detergent model created from a dataset whit different presentations are
used as an example. The presentations are as follows: bottle 1 L liquid, bottle 750 ml
liquid, bag 500 g powder, bag 900 g powder, bag 1 kg powder, bag 5 kg powder. The
presentation in bag 5 kg powder stands out in size from others, in the same way, its
price is higher. In the test stage, the model demonstrated know to differentiate between
this presentation and the others, clearly observing a distance between the forecasts,
always because of the presentation of bag 5 kg powder that generated higher prices.
The above can be seen in the Fig. 2.
The predictions shown in Fig. 2 come from a model that obtained a coefficient of
determination of 0.98.
Fig. 3. Error comparison with the model of ultra-pasteurized milk, partially skimmed, box 1 L
Lala light milk.
Fig. 4. Test for ultra-pasteurized milk, partially skimmed, box 1 L brand Lala Light in
Chedraui, Toluca, Mexico, Alfredo del Mazo 705, col. Tlacopa CP 50010
Analytics for Basic Products in Mexico 127
3 Results
A model and predictions were created of ultra-pasteurized milk, partially skimmed, box
1 L brand Lala Light. A brief description of the dataset is presented in Table 4.
Table 4. Description ultra-pasteurized milk, partially skimmed, box 1 L brand Lala Light.
Number of records 39939
Mean 14.2270
Standard deviation 0.9732
Minimum 8.0000
25% 13.5000
50% 14.3000
75% 14.9000
Maximum 25.9000
The characteristics selected in this model are the following: date, year, commercial
name, address, municipality, state, commercial chain and month. The result of the
parameter selection can be seen in the Table 5.
Table 5. Results of the selection of parameters for ultra-pasteurized milk, partially skimmed,
box 1 L brand Lala Light.
Alpha Network size Ranking
0.001 (527, 527, 527) 1
0.1 (8) 2
0.01 (527, 527) 3
0.000001 (527, 527, 527) 4
The training period was from January-2011 to June-2015. The yield obtained was
0.1722 of the coefficient of determination and an average absolute error of 0.5131, both
in the test stage, which covered from June 2015 to April 2016.
Now, an example of the predictions generated during the test stage for an address,
that is, an establishment, in this case of the Chedraui commercial chain in the state of
Toluca. See Fig. 4.
Using this methodology, it was discovered that the prices of basic products in this
database vary by address, prices are not kept by commercial chains. Even establish-
ments of the same commercial chain in the same state and municipality, present their
own prices, an example can be seen in Table 6. In this case, there is a price difference
of more than 15%.
128 P. Millán and F. Mata
Table 6. The same detergent product offered at Walmart in two establishments at different
prices on the same date.
Product Presentation Brand Price Date Commercial Address State Municipality
chain
Dish Bag 900 g Salvo 16.91 2011- Walmart Calzada de Distrito Gustavo A.
detergent powder 04-08 Guadalupe 431 Federal Madero
(green)
Dish Bag 900 g Salvo 19.50 2011- Walmart Othon de Distrito Gustavo A.
detergent powder 04-08 Mendizabal Ote Federal Madero
(green) 343
It is interesting how this variation occurs even in nearby locations, like in this case,
two establishments belonging to the Walmart commercial chain at approximately 3 km
away. It is depicted in Fig. 5.
Analytics for Basic Products in Mexico 129
We conclude t is possible not only to know the behavior of the products but also to
generate models of prediction of the prices of these products in a particular place and
some period.
It is possible by combining the techniques of exploration and machine learning on
time and location domains. It is applied to categorical attributes such as product,
presentation, brand, and commercial chain, address, among others.
Nevertheless, It is necessary to work on other approaches that allow the transfor-
mation of qualitative or categorical data to numerical data. Considering the conditions:
do not insert bias and not increase the number of columns.
The use of Feature selections algorithm allowed identifying datasets with better
performance than the original ones to produce price predictions. However, it is pro-
posed as future work to add new variables that can later be analyzed with other
algorithms, to discover possible external variables to this dataset that influence the
variation of prices products in Mexico.
References
1. INEGI Homepage. Encuesta Nacional de Ingresos y Gastos de los Hogares (ENIGH) (2014).
https://www.inegi.org.mx/programas/enigh/tradicional/2014/default.html. Accessed May
2019
2. PROFECO Homepage. Quién es Quién en los precios. https://datos.gob.mx/busca/dataset/
quien-es-quien-en-los-precios. Accessed January 2018
3. Hinton, G.E.: Connectionist learning procedures. In: Artificial Intelligence, pp. 185–234
(1990)
4. Glorot, X., Bengio, Y.: Understanding the difficulty of training deep feedforward neural
networks. In: Thirteenth International Conference on Artificial Intelligence and Statistics,
Sardinia, Italy (2010)
5. He, K., Zhang, X., Ren, S., Sun, J.: Delving deep into rectifiers: surpassing human-level
performance on ImageNet classification. In: Proceedings of the IEEE International
Conference on Computer Vision, pp. 1026–1034 (2015)
6. Kingma, D.P., Ba, J.: Adam: a method for stochastic optimization (2014)
7. Pedregosa, F., Varoquaux, G., Gramfort, A., et al.: Scikit-learn: machine learning in Python.
J. Mach. Learn. Res. 12, 2825–2830 (2011)
8. Rumelhart, D.E., Hinton, G.E., Williams, R.J.: Learning representations by back-
propagating errors. Cogn. Model. 5(3), 1 (1988)
9. LeCun, Y.A., Bottou, L., Orr, G.B., Müller, K.-R.: Efficient BackProp. In: Montavon, G.,
Orr, G.B., Müller, K.-R. (eds.) Neural Networks: Tricks of the Trade. LNCS, vol. 7700,
pp. 9–48. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-35289-8_3
10. Pedregosa, F.: «Scikit Learn» [En línea]. https://scikit-learn.org/stable/modules/neural\_
networks\_supervised.html\#regression. Accessed July 2019
11. Kingma, B.J.: Adam: a method for stochastic optimization. In: 3rd International Conference
for Learning Representations, San Diego (2014)
Design and Development of Photovoltaic
Power Meter
1 Introduction
Till now, the greatest amount of electricity used by man is extracted from natural fossils
fuels. The energy generation by fossils fuels has concerned due to its pollution on eco-
system and limited amount available. So that it is now necessary to look for alternating
energy sources that can supply energy without producing harmful gases for long
period. Among different alternative sources of energy, solar energy is sustainable and
renewable energy, which can be converted directly from sun energy to electrical energy
by photovoltaic (PV) cells and modules [1]. The photovoltaic (PV) cell is an electronic
device, which transforms the visible light into direct current electrical energy. The
output power depends on area of cells, incident light level, efficiency but it is reduced
for temperature increase. Due to the photovoltaic cell is a low voltage device (0.56 V)
then photovoltaic modules are built that are integrated by several electrically connected
cells in series to increase the voltage.
The basic module is designed to couple with an automotive type battery, its output
voltage in the open circuit of the order of 17 to 19 V so that it can supply power to the
battery. Its output power is variable from 20 W and depends of cell size.
The manufacturers are fabricating PV cells and modules of mono crystalline and
polycrystalline silicon principally although many other semiconductor materials can be
used [1]. The basic PV module is integrated by 36 PV cells connected in series con-
nected to automotive battery of 12 V, but with least time PV modules with higher output
power and voltages are being manufactured [2]. The PV cell or module is electronic
device that converts the light energy from solar energy into the electrical energy or direct
current (DC). The PV cell is a semiconductor diode the more common is done with
silicon and phosphorus although there are many semiconductors materials to make PV
cells [2]. The output current of photovoltaic module is represented by Eq. 1.
h qðVc þ Ic Rs Þ i I R þV
c s c
Ic ¼ I0 e KT 1 þ þ Ig ð1Þ
Rf
From Eq. 2, Po is the output power, Vo is the output voltage, Ig is the generated
current, q is the electron charge, Io is the output current, K is the Boltzmann constant,
T is the temperature, Rs is the series resistance [3]. When both are solved changing the
voltage Vc from open circuit condition to short circuit condition, all the values of power
and voltage are obtained, in order to get the current-voltage (I-V) and power-voltage
(P-V) curves of the PV module (see Fig. 1).
The characteristic I-V and PV curves (see Fig. 1) of the photovoltaic module are
very important because they have the information of their electrical variables. The
maximum output power is used when a photovoltaic system is designed. But the power
of the module is variable because it depends mainly on the environmental conditions of
irradiance level, wind speed and module temperature. The modules are electrically
characterized by the manufacturer in standard conditions (i.e. solar power of
1000 W/m2, temperature of 25 °C) where in real working conditions the photovoltaic
module is subjected to environmental changes all time, then it is necessary the PV
module output maximum power must be measured on installation place.
According to the graph in Fig. 1, the I-V curve of solar module is nonlinear, and
then it is not easy to locate the maximum power point (MPP). So, it is necessary to
132 J. Vega Pérez et al.
performance of solar cells because the temperature of the solar cell decreases the solar
cell performance.
Through sampling and calculation, the maximum Power Point is tracked success-
fully by determining the step size according to the present and previous values of
voltage and power variation to strike a balance between tracking time and tracking
accuracy. Peng [8]. To solve the problem, variable-step size MPPT algorithms have
been reported and they concluded the transient response of PI type P&O is the fastest
among all. However, the complexity of this method is also the highest.
Santos et al. [9] proposed a traditional Perturbation and Observation to track the
maximum power point (P&O) MPPT technique with trade-off between step-size,
tracking-time and tracking-accuracy. The traditional P&O method sampled the voltage
and current, and the power can be obtained by multiplying the two values. Farhat et al.
[10]. Developed un algorithm based on the architecture Multi-layer Perceptron
(MLP) to track real time, the maximum power point (MPPT). Maximum Power Point is
a function of irradiance and temperature. In contrast, the PV systems must be designed
to operate at their peak output power to extract and maintain the peak power of the PV
panel even when climatic conditions are unfavorable. The obtained results show that
the use of artificial neural networks present a precise estimate of the maximum power
of PPV compared to the measured values.
Ramana et al. [11] described an MPPT algorithm to track the global peak out of
many local peaks of PV module. The output of the PV array mainly depends on the
irradiance and temperature where the non-uniform irradiance conditions occurs on PV
module due to passing clouds, bird dropping, shadows etc. The MPPT algorithm tracks
the global pick out of the many local peaks. The input to the PV array is irradiance and
temperature of four modules and the output is the duty ratio. Pulse Width Modulator
receives the duty ratio as input and generates pulses to the MOSFET of boost converter.
Battery is used as load.
Wang et al. [12] defined an improved particle swarm optimization algorithm is
proposed to track the maximum power point of the PV array when the PV array is
partially shaded by object occlusion. The model of PV array is established, and sim-
ulation is carried out to verify the effectiveness of the algorithm. The method can avoid
the local maximum and track the global maximum power point. Badis et al. [13]
reported the conventional methods such as Perturb and Observe (P&O) and the
Incremental Conductance and fails to extract the global MPP of the PV panel if the PV
generator is partially shaded. They propose an Evolutionary Algorithms (AE), trough
the Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) is better. They
developed an algorithm based in neural networks using MatLab software. Moreover, it
is proven that PSO process mitigates more efficiently power losses caused by PS with
lesser fluctuation and higher steady final output power.
2 Proposal
The prototype proposed is integrated by electronic circuit, two sensors for current and
voltage. After that, they were used as two amplifier steps with integrated circuit LM308
The output signals of this amplifier are sent to analogical-digital converter. An elec-
tronic power circuit was designed [14] to drain up to 10A which is coupled to the
photovoltaic module, with the combination of analogic integrated circuits like TL081
and power transistor as MJ5529. A circuit generator was designed [15] to generate the
random time to do the measurement. Tl081 integrated circuit, transistors as BC547, an
electrolytic capacitor and several resistances were used. An electronic ramp voltage
generator circuit was also designed [16] to activate the power circuit coupled to the PV
module, using analogic integrated circuit MC308 combined with capacitors and
resistance to get a time reference in which the electronic system can do the measure-
ment This circuit was calibrated to get the range of 1 to 5 s so that the laptop can
capture, the storage and processing of the electrical signals found during the mea-
surement. In Fig. 3 is showed the block diagram of this electronic circuit. Also, the
Arduino control target was used to link the electronics circuit with the laptop. For this,
it is necessary to design the appropriate software based in Matlab.
4 Software Design
For accessing the voltage and current signals, the Arduino electronic card was used
between the developed electronic circuit and the PC and the software was designed
guessing the following logical of the diagram on Fig. 4. The voltage signals were
processed using Matlab software and an Arduino hardware was use in other to capture
the electrical signal of current and voltage or PV module and transfer to laptop or
computer where the electrical signals and determinate the characteristic curve or PV
module.
5 Developing
All the analogies and integrated digital circuits, where signal transistors, resistances,
and capacitors are installed, were assembled by a printed circuit board. The power
transistor was installed on a heat sink because it works with 15 A of the current. The
control target was used to link the electronic circuit designed with the computer. Both,
electronics and control target are assembled into the cabinet. The electronic circuit
developed is shown in Fig. 5. Also, a cabinet was used for the installation of the
electric controls, as well as a digital display to show the values of current, voltage and
power that are utilizing to measure efficiency in the photovoltaic module. Photovoltaic
power meter prototype developed is shown in Fig. 6.
6 Preliminaries Measurements
The maximum power, the I-V and P-V curves were measured with same level of
incident solar power and room temperature. Several photovoltaic modules of different
nominal powers were measured with the prototype developed. The measurements have
done in a normal day and the incident solar power was changing around from 600 to
860 W/m2. With the help of lux meter intensity (Solar power meter TES 1333R
Datalogging), two PV modules of 20 W manufactured in CINVESTAV Mexico were
used. The first measurement was done with one module and the second one was done
with two modules connected in parallel. The short circuit current (Isc), open circuit
voltage (Voc), maximum power (Pm), the maximum current (Im) and the Maximum
voltage (Vm) of photovoltaic module were measured. The experimental results are
shown in the Table 1.
Table 1. Shows the various parameters of PV module for the calculation of efficiency
PVm (W) Light power (W/m2) Iscc (A) Voc (V) Im (A) Vm (V) Pm (W)
20 600 1.12 18.45 0.91 12.8 11.52
20 700 1.24 18.51 1.15 12.00 14.04
20 860 1.25 18.64 0.91 12.7 17.24
40 790 2.31 18.75 2.15 12.9 27.55
40 800 2.36 18.60 2.24 13. 30.52
40 820 2.4 18.82 2.35 13.6 31.42
40 850 2.42 19.25 2.3 14.20 32.30
Also, the I-V curve was measured. The previous results are showed in the Fig. 7.
Here, the power voltage curve of PV module was plotted with the same incident
solar power was 700 W/m2. The previous results are showed in the Fig. 8.
138 J. Vega Pérez et al.
Based on the experimental results of Table 1, it was found that the developed electronic
prototype works correctly and can measure the maximum power of the photovoltaic
module, and its current coordinates for varying levels of incident solar power. it is
possible to see how voltage and power variables of the photovoltaic module were
changed it was also confirmed by connecting two modules in parallel the current and
power variables increased while the voltage was almost the same value as that of a only
module.
In addition, it was observed that there was a greater change in current as a function
of the incident light power, however, the voltage change was minimal, which corre-
sponds to Eqs. 1 and 2 because the current changes exponentially while the voltage
changes logarithmically. From the voltage-current curve reported in Fig. 2, it could be
verified how the prototype developed can properly capture and process the current and
voltage signals of the photovoltaic module, store the information and then measure the
corresponding curve.
In the mentioned curve it was observed that the experimentally obtained curve
starts from zero with the horizontal axis of voltage this is due to the fact that the
measurement process of the photovoltaic module starts from the open circuit condition.
On the other hand the curve does not touch the vertical axis of current that is because
the electronic power circuit coupled to the photovoltaic module during the measure-
ment has a minimum resistance of the order of 0.3 V, which causes a voltage drop of
the order 0.6 V which prevents the curve from reaching the current axis. In relation to
the voltage power curve of Fig. 3, it is verified that the computer program developed in
Matlab which controls the measurement process works correctly because it is equal to
the theoretical curve of Fig. 1, which verifies that the voltage current processing was
right because it determines the power for all captured points.
However, it can be seen that the curve starts from the horizontal axis of voltage,
which indicates that the measurement starts in the open circuit condition of the pho-
tovoltaic module but does not reach the crossing point or zero of the axes, therefore it
reaches the short circuit condition, also due to the saturation of the power circuit of the
prototype developed. The development of a method of measuring the maximum power
of the photovoltaic module and its characteristic curves I-V and P-V propose is dif-
ferent of the proposed method. the purposed circuit is activated with a ramp voltage
signal in contrast with the reported methods which use technics like oscillating voltage
signal, pulse-width modulation, power signal derivation among others. Therefore, the
measurement with the proposed prototype is more stable and the proposed computer
program is simple. The proposed prototype can measure whatever photovoltaic module
with any incident light power.
8 Conclusions
According with the experimental results, the electronic system developed works cor-
rectly and its experimental results are agreement with the theoretical method used in the
design.
140 J. Vega Pérez et al.
Acknowledgments. The authors are thankful to National Polytechnic Institute from México,
and also grateful to the project No. 20195869 SIP-IPN by their financial support.
References
1. Vega-Perez, J., Ponomaryov, V., Nino-de-Rivera, L.: Electronic system for power
optimization of photovoltaic generator. Int. J. Electromagn. Waves Electron. Syst. 7–8(8),
12–15 (2003)
2. Wanzeller, M., Cunha Alves, R., Fonseca Neto, J., dos Santos Fonseca, W.: Current control
loop for tracking of maximum power point supplied for photovoltaic array. IEEE Trans.
Instrum. Meas. 53(4), 1304–1310 (2004)
3. Munji, M., Okullo, W., Van Dyk, E., Voster, F.: Local devise parameter extraction of a
concentrator photovoltaic cell under solar spot illumination. Sol. Energy Mater. Sol. Cell 94,
2129–2136 (2010)
4. Ashish, P., Nivedita, D., Ashok, K.: High-performance algorithms for drift avoidance and
fast tracking in solar MPPT system. IEEE Trans. Energy Convers. 23(2), 681–689 (2008)
5. Khatib, T., Ghareeb, A., Tamimi, M., Jaber, M., Jaradat, S.: A new offline method for
extracting I-V characteristic curve for photovoltaic modules using artificial neural networks.
Sol. Energy 173, 462–469 (2018)
6. Danandeh, A., Mousavi, S.M.: Comparative and comprehensive review of maximum power
point tracking methods for PV cells. Renew. Sustain. Energy Rev. 82, 2743–2763 (2018)
7. Solimana, A., Hassana, H., Ookawara, S.: An experimental study of the performance of the
solar cell with heat sink cooling system. Energy Proc. 162, 127–135 (2019)
8. Peng, B., Chen, J., Liu, Y., Chiu, Y.: Comparison between three different types of variable
step-size P&O MPPT technique. In: International Conference on Computer Information
Systems and Industrial Applications (2015)
9. Santos, T., Galhardo, A., Perturbation, A.: Observation routine used to control a power
converter. In: 2014 Sixth World Congress on Nature and Biologically Inspired Computing
(NaBIC), pp. 78–83 (2014)
10. Farhat, S., Alaoui, R., Kahaji, A., Bouhouch, L.: Estimating the photovoltaic MPPT by
artificial neural network. In: Proceedings of International Renewable and Sustainable Energy
Conference (IRSEC), pp. 49–53 (2013)
11. Ramana, V., Jena, D.: Maximum power point tracking of PV array under nonuniform
irradiance using artificial neural network. In: Proceedings of IEEE International Conference
on Signal Processing, Informatics, Communication and Energy Systems (SPICES), pp. 1–5
(2015)
Design and Development of Photovoltaic Power Meter 141
12. Wang, Y., Bian, N.: Research of MPPT control method based on PSO algorithm. In:
Proceedings of the 4th International Conference on Computer Science and Network
Technology (ICCSNT), pp. 698–701 (2015)
13. Badis, A., Mansouri, M., Sakly, A.: PSO and GA-based maximum power point tracking for
partially shaded photovoltaic systems. In: Proceedings of 2016 7th International Renewable
Energy Congress (IREC), pp. 1–6 (2016)
14. Rashid, M.: Power Electronics Handbook. Prentice Hall, Upper Saddle River (1999)
15. Malik, N.R.: Electronic Circuits: Analysis, Simulation, and Design. Prentice Hall, Upper
Saddle River (1998)
16. Sedra, A.S., Smith, K.C.: Microelectronics Circuits. Oxford University Press, Oxford (1999)
Artificial Intelligence and Machine
Learning
Clustering Methodology in Mixed Data Sets
Abstract. One of the most challenging tasks of data analysis is finding clusters
in mixed data sets, as they have numerical and categorical variables, and lack a
labeled variable to serve as a guide. These clusters could serve to summarize all
the variables of a data set into one and be able to find information more easily
than generating summarizations for each variable. In this research thesis, a
methodology of clustering on mixed data sets is proposed, which yields better
results than the methods applied in the state of the art.
1 Introduction
At today the human activity generates data every time, the problem is what to do with
them? How to find valuable information? As an interconnected technology society, we
already have stable data structures, industrial algorithms, repositories of almost all
topics, distributed computing to deal with large volumes but the problem remains, how
to process them?
This paper begins with the assumption of the existence of a hidden phenomenon in
these data sets that we will call a pattern [1]. The mathematic on which the data
analysis is based is focused on finding this phenomenon. Due to the complexity of the
high dimensions of the phenomena that describe our data sets, it is sometimes difficult
with simple statistics to find this pattern. That is why science never stops working and
keep looking for new ways to find information about the pattern hidden in our data
because no method has the absolute truth.
Then a specific methodology is proposed to process, through unsupervised clas-
sification, mixed data sets through a series of steps, which lead us to satisfactory results
as will be seen below, and which raises a new way of creating clusters beyond applying
algorithms directly. The steps are quantified with stable metrics, and especially this
methodology is intended to be usable in real problems.
The case study presented in this paper is the safe houses where the crime of
kidnapping was committed, the complexity here was to integrate and process different
mixed data sets from official surveys, to find valuable information that a specialist on
issues of public safety you may be interested.
qualitative respectively; when a variable is numerical, its domain is of the real type,
therefore, its values can be continuous or discrete, finite or infinite. Its elements can be
part of any rational subset, integer, irrational or transcendent, then
domainðXnumerical Þ ¼ R; when a variable is categorical, its domain has a set of finite
symbols, its values can be numbers or characters that represent some category. These
categorical can be of two types; nominal, when there is no order in the symbols, and
ordinals, when, there is an order in the symbols, then
domain Xcategorical ¼ fs1 ; s2 ; . . .; sn g:
3 Proposed Methodology
The most common way to attack the problem of unsupervised classification in mixed
data sets is through algorithms applied directly to the data, but in this research work,
the hypothesis is through a specific methodology for this type of data set (see Fig. 1),
could help in the search for clusters or groups in the data. Therefore, this thesis presents
a methodology to find groupings of observations in mixed data sets through the pre-
processing and application of an array arrangement of unsupervised partition classifi-
cation algorithms.
Steps 1 and 2, source and mixed data set, are subjective since the data set can come
from a specific experiment and the scientist who is analyzing the data, and they can
have an accurate knowledge of the phenomenon that is wanted to find or may also be in
exploration of the phenomenon.
148 J. G. González León and M. F. Mata Rivera
3.1 Preprocessing
This step consists of 4 sub-phases, load, clean, fill and transform, which have the
sequential sense to ensure the different actions: understanding the information, quality,
integrity and homogeneity in the data.
The mixed data set must be loaded into the computer’s memory to be able to
analyze it, and regardless of the operating system. The cleaning of a data set is done so
Clustering Methodology in Mixed Data Sets 149
that the values correspond to the range of values that it should. One of the problems to
which this step gives solution is to deal with erroneous, inconsistent, obsolete and/or
redundant data. For this, it is proposed, with the help of metadata and/or prior
knowledge of the subject, to search and count the unique values of each variable and
compare the original domain with the current domain, in this way you can also easily
detect outliers and extremes
Once the anomalies in the data have been cleared, and if it is the case that there are
missing values, either spaces or any symbol that indicates a lack of data (NA, NaN, or
NULL) this missing data can be filled. Although there are several conventional
methods [4], among them, eliminating the rows that contain the missing values, or
replacing the missing value with the average, you can also use predictive methods [5]
based on supervised machine learning such as Linear Regression or Random Forest to
complete this lack of values.
The last sub-phase of the preprocessing is a transformation of the data, this in order
to normalize the domain in which the values of the variables are. As already mentioned
before, a mixed data set has numerical and categorical variables, for each type there is a
different transformation:
• Numerical: this transformation is made through the min-max normalization tech-
nique [6] with the minimum and maximum operations, respectively, which scales
all values to a range between [0, 1], according to the following Eq. 2:
xi minðXÞ
xi :¼ ð2Þ
maxðXÞ minðXÞ
In where X is the vector of the numerical variable to transform, and xi is the ith
value of vector X.
• Categorical: this transformation replaces the symbolic values of the categories of a
variable, the numerical values of its cumulative probability, and is based on the
proposal of [7] through the following method:
a. For each column of categorical data in the data set, the categories are sorted in
descending order by frequency.
b. Knowing the total observations, we find the probability pxi of each possible
categorical value xi through Eq. 3:
f xi
px i ¼ ð3Þ
N
In where xi is the ith value of the variable x; f xi is the frequency of observations of
the ith value of the variable x; and N is the total number of observations.
c. The categorical value is replaced by the value of its cumulative probability axi
through Eq. 4:
150 J. G. González León and M. F. Mata Rivera
X
axi :¼ i
px i ð4Þ
example, is to find perhaps what are the “true groups” or “true classification” [14] by
comparing the partition found against the classification original of the dependent
variable.
The evaluation of the found partitions of a mixed data set without labeling variable,
with unsupervised classification techniques, is called “internal evaluation” or “internal
measure” and is used to evaluate the quality of the partitions that are generated. Testing
different values of k [15] and measuring the quality of the partition, which found the
unsupervised classification algorithm that was used, through the similarity within the
groups and the separation between them, you can find this evaluation, commonly
through an internal evaluation index that is expressed by reason
ðintra-group)/(inter-group) [16]. For example, in [17] 27 internal evaluation indices are
proposed, whose value is the maximization or minimization of this ratio; the best
partition generated will then be the one that can have the least compactness within its
groups, and the highest among the separability among all.
Ensemble Algorithm
An ensemble of unsupervised classification algorithms and a heuristic that evaluate
different values of k and different unsupervised classification algorithms are proposed
in the following pseudocode:
Consensus Function
Once you have the validation matrix vm ½kn ; am ; pm ; em , a decision function f(vm Þ will
decide which is the best partition, by maximizing or minimizing the value of the
internal evaluation index em for each proposed group kn , in the following pseudocode
is showed a consensus function that maximizes:
152 J. G. González León and M. F. Mata Rivera
Figure 2 shows these security houses found on the map of Mexico City in the
2016–2017 period, in total 21 security houses were found distributed in 10 of the 16
municipalities.
The data found from the safe houses when being extracted from news do not come
with some extra information, beyond the address, that would allow us to be able to
identify the reason for the phenomenon, so it was devised to integrate different sets of
data that had characteristics of the geography where these houses were found, in order
to discover information.
Clustering Methodology in Mixed Data Sets 153
According to the methodology developed in this thesis, these data must come from
an official source. Therefore, it was chosen to take open data from regular household
surveys of the Instituto Nacional de Estadística y Geografía (INEGI). Each of these
surveys is a set of mixed data, as there are questions with binary, or categorical
answers, as well as numerical data such as age, or amounts of money.
After completing the preprocessing phase proposed in the methodology, the mixed
data sets from the resulting surveys are shown in Table 3. Once the preprocessed mixed
data sets are done, in this case 5: HOGARES, SDEMT, TMOD, TPER_VIC1,
VIVIENDA, the unsupervised classification can start.
3.4 Implementation
To begin with the measurement of the clustering tendency of mixed data sets, Hopkins
statistics were proposed [18] since this test that determines whether a data set has a
random structure, that is, measures the probability that the data set comes from a
uniform distribution of data. If the data sets do not come from a uniform distribution of
data, it may contain well segmented groups.
In [19] they review the unsupervised classification algorithms used over 20 years,
including partition algorithms, so it was proposed to use the k-means k-modes and k-
medoids algorithms. These algorithms have the particularity that the main parameter is
to indicate the k groups, so, according to this methodology, an internal evaluation index
is needed to find this number k.
The proposed internal evaluation index is the Dunn index [20] since it measures the
separation between groups, seeking the maximization of this inter-cluster distance, and
cohesion within the groups, seeking the minimization of intra-cluster distance. The
decision function will then be the maximization of this Dunn index [21] the larger the
Dunn index, the better the partition found. The index implementation could be bene-
ficial to find the optimal number of the k groups needed for the assembly algorithms
since this number of k could vary between 1, the case where the whole mixed data set is
a group, and N, where each observation is a different group.
4 Results
Earlier in the last chapter it was already mentioned, that the integration of information
to the case study “safe houses” is explored to describe this phenomenon, with an
unsupervised classification approach. It was also mentioned what is the source and
mixed data sets chosen for analysis, 5 surveys with different topics.
After processing the mixed data sets, the result of the clustering trend measurement
can be seen in Table 4.
Table 4. Result of the clustering tendency in mixed data sets from INEGI surveys.
Mixed data set Clustering tendency: Hopkins statistics
Without methodology With methodology
HOGARES 0.0461 0.3746
SDEMT 0.00814 0.2422
TMOD_VIC 0.0464 0.3759
TPERC_VIC1 0.1925 0.4908
VIVIENDA 0.1664 0.3552
Once it was measured how many significant groups, through the grouping trend,
can be found in these data sets, the unsupervised classification was performed with the
help of the algorithm assembly. Table 5 shows the result of the evaluation of the Dunn
index of the assembly of unsupervised classification algorithms proposed in the
Clustering Methodology in Mixed Data Sets 155
methodology, against the algorithms that were found are used in the state-of-the-art k-
medoids and k-prototypes.
Table 5. Result of internal evaluation in mixed data sets from INEGI surveys.
Mixed data set Internal evaluation: Maximum Dunn Index
Without methodology With methodology
k-medoids k-prototypes k-means k-medoids k-modes
HOGARES 0.1040 1.94e−05 0.3872 0.0598 0.2326
Clusters 10 9 4 2 2
SDEMT 0.1810 7.0460e−06 0.6779 0.0515 0.6779
Clusters 2 2 2 10 2
TMOD_VIC 0.0271 0.0238 0.0334 0.0162 0.0331
Clusters 3 3 6 2 3
TPERC_VIC1 0.0078 0.0083 0.2458 0.1959 0.2884
Clusters 4 4 2 10 9
VIVIENDA 0.0819 0.0064 0.1892 0.1091 0.1302
Clusters 2 3 3 3 2
Fig. 3. Visualization of the groups found of the mixed data set HOGARES.
156 J. G. González León and M. F. Mata Rivera
Table 6. Sample of 5 variables of the groups found in the mixed data set HOGARES.
Survey: HOGARES
Variables acc_alim1 acc_alim3 acc_alim6 alim17_2 alim17_4
Cluster Meaning
Worry Limited Number of Number of Number of
meal is variety of days they ate days they ate days they ate
over food egg meat fruits
1 No No 2 2 7
2 Yes Yes 3 2 2
3 No No 2 7 7
4 No No 3 3 7
In Fig. 4 the different groups can be displayed, as well as the map of Mexico City,
the delegations belonging to each group found from the SDEMT survey, and Table 7
shows a sample of 5 variables that characterize each group found.
Fig. 4. Visualization of the groups found of the mixed data set SDEMT.
Table 7. Sample of 5 variables of the groups found in the mixed data set SDEMT.
Survey: SDEMT
Variables Sex eda cs_p13_1 cs_p17 e_con
Cluster Meaning
Sex Age What degree did the Do you currently Marital
school approve? attend school? status
1 Woman 21 Elementary Yes Single
2 Man 40 Undergraduate No Married
In Fig. 5 the different groups can be displayed, as well as the map of Mexico City,
the delegations belonging to each group found from the TMOD_VIC survey, and
Table 8 shows a sample of 3 variables that characterize each group found.
Clustering Methodology in Mixed Data Sets 157
Fig. 5. Visualization of the groups found of the mixed data set TMOD_VIC.
Table 8. Sample of 3 variables of the groups found in the mixed data set TMOD_VIC.
Survey: TMOD_VIC
Variables BP1_1 BP1_5 BP1_3C
Cluster Meaning
What month did the Where place did the In which municipality did
crime happen? crime happen? the crime occur?
1 May On the street Álvaro Obregón
2 February On the street Iztapalapa
3 December On the street Iztapalapa
4 December In his house Álvaro Obregón
5 August On the street Benito Juárez
6 December In his house Cuauhtémoc
In Fig. 6 the different groups can be displayed, as well as the map of Mexico City,
the delegations belonging to each group found from the TPER_VIC1 survey, and
Table 9 shows a sample of 3 variables that characterize each group found.
In Fig. 7 the different groups can be displayed, as well as the map of Mexico City,
the delegations belonging to each group found from the VIVIENDAS survey, and
Table 10 shows a sample of 5 variables that characterize each group found.
158 J. G. González León and M. F. Mata Rivera
Fig. 6. Visualization of the groups found of the mixed data set TPER_VIC1.
Table 9. Sample of 3 variables of the groups found in the mixed data set TPER_VIC1.
Survey: TPER_VIC1
Variables AP4_10_01 AP4_10_04 AP4_10_06
Cluster Meaning
For fear of being a victim For fear of being a For fear of being a
of a crime, did you stop victim of a crime, did victim of a crime, did
going out at night? you stop taking a taxi? you stop carrying cash?
1 Yes No Yes
2 No No No
3 Yes No Yes
4 No No No
5 No No Yes
6 No No No
7 No No Yes
8 No No No
9 Yes Yes Yes
Fig. 7. Visualization of the groups found of the mixed data set VIVIENDAS.
Clustering Methodology in Mixed Data Sets 159
Table 10. Sample of 5 variables of the groups found in the mixed data set VIVIENDAS.
Survey: VIVIENDAS
Variables antiguedad tipo_viv num_cuarto cuart_dorm auto_pick
Cluster Meaning
Housing Type of Number of Bedroom Own car
age housing rooms availability
1 27 Independent 5 2 Yes
house
2 25 Independent 3 1 No
house
3 32 Department 4 2 Yes
5 Discussion
With only the information from the mixed data sets from the INEGI surveys in Table 3,
we could begin to generate hypotheses about how many groups these data sets could
have. For example, from the SDEMT table that contains more observations, we might
think that there are many more groups than the others, or that the HOGARES table that
contains fewer observations, there are few groups. So, when we apply the methodol-
ogy, we realize that only two groups were found in the SDEMT survey, while four
groups were found in the HOGARES survey. There is not yet a rule that, depending on
the number of observations or variables, so many groups will be obtained, which is
why the unsupervised classification exists and makes sense, to propose this number of
partitions in the data that at first glance cannot be defined. Also, with the basic
information of the groups for each mixed data set, summaries could be made with their
variables. but summary tables should be made for each variable that contains the mixed
data set, this can take a long time and many tables, so it might not be useful.
One advantage of finding clusters in the data is to simplify this task to only one
table, as for example, in the case of Table 6 of the first HOGARES survey analyzed. It
is the case similar to wanting to see on a map the characteristics of each delegation, the
same number of maps would have to be made to be able to visualize all the variables, as
is the case of Fig. 3 of the HOGARES survey, where the map is seen with the dele-
gations belonging to each group found, it is easier to read and interpret the variables of
a mixed data set, based on the groups found with this methodology.
Then the interpretation for each of the data sets, based on the question: which
delegations do not have safe houses? You can respond by summarizing the groups
found. Even with the methodology supported by the different measurements that are
made, questions remain that perhaps only an expert on the subject could answer.
Questions like if the groups found have any interest value? Or in what range of k
clusters should the heuristics of the algorithm assembly be run?
160 J. G. González León and M. F. Mata Rivera
References
1. Ströing, P.: Scientific Phenomena and Patterns in Data. Ludwig-Maximilians-Universität,
München (2018)
2. Zaki, M.J., Meira, W.: Data Mining and Analysis. Cambridge University Press, Cambridge
(2014)
3. Bramer, M.: Principles of Data Mining. Springer, London (2016). https://doi.org/10.1007/
978-1-4471-7307-6
4. Soley-Bori, M.: Dealing with missing data: key assumptions and methods for applied
analysis, vol. 23. Boston University (2013)
5. Yadav, M., Roychoudhury, B.: Handling missing values: a study of popular imputation
packages in R. Knowl.-Based Syst. 160, 104–118 (2018)
6. Larose, D., Larose, C.: Discovering Knowledge in Data: An Introduction to Data Mining,
2nd edn. Wiley, Hoboken (2014)
7. Patki, N., Wedge, R., Veeramachaneni, K.: The synthetic data vault. In: IEEE International
Conference on Data Science and Advanced Analytics (DSAA), pp. 399–410. IEEE (2016)
8. Adolfsson, A., Ackerman, M., Brownstein, N.: To cluster, or not to cluster: an analysis of
clusterability methods. Pattern Recogn. 88, 13–26 (2019)
9. McCue, C.: Public-safety-specific evaluation. In: Data Mining and Predictive Analysis:
Intelligence Gathering and Crime Analysis, pp. 157–183. Butterworth-Heinemann (2015)
10. Wu, X., Ma, T., Cao, J., Tian, Y., Alabdulkarim, A.: A comparative study of clustering
ensemble algorithms. Comput. Electr. Eng. 68, 603–615 (2018)
11. Jukes, E.: Encyclopedia of machine learning and data mining (2nd edition). Ref. Rev. 32, 3–
4 (2018)
Clustering Methodology in Mixed Data Sets 161
12. Loshin, D.: Knowledge discovery and data mining for predictive analytics. In: Business
Intelligence. The Savvy Manager’s Guide MK Series on Business Intelligence, 2nd edn.,
pp. 271–286 (2013)
13. Tharwat, A.: Classification assessment methods. Appl. Comput. Inform. (2018)
14. Hennig, C.: What are the true clusters? Pattern Recogn. Lett. 64, 53–62 (2015)
15. Gurrutxaga, I., Muguerza, J., Arbelaitz, O., Pérez, J., Martín, J.: Towards a standard
methodology to evaluate internal cluster validity indices. Pattern Recogn. Lett. 32, 505–515
(2011)
16. Jauhiainen, J., Kärkkäinen, S.: Comparison of internal clustering validation indices for
prototype-based clustering. Algorithms 10, 105 (2017)
17. Desgraupes, B.: Clustering Indices. University of Paris Ouest-Lab Modal’X, vol. 1, pp. 34
(2013)
18. Han, J., Kamber, M., Pei, J.: Cluster analysis: basic concepts and methods. In: Data Mining,
pp. 443–495 (2012)
19. Benabdellah, A., Benghabrit, A., Bouhaddou, I.: A survey of clustering algorithms for an
industrial context. Proc. Comput. Sci. 148, 291–302 (2019)
20. Rodriguez, M., Comin, C., Casanova, D., Bruno, O., Amancio, D., Costa, L., Rodrigues, F.:
Clustering algorithms: a comparative approach. PLoS One 14, e0210236 (2019)
21. Yang, Y.: Temporal Data Mining via Unsupervised Ensemble Learning. Elsevier Science,
Amsterdam (2016)
American Sign Language
Electromiographic Alphabet
Sign Translator
1 Introduction
3 Electromiographic Sensors
Electromyography (EMG) and nerve conduction study (NCS), are tests that
use electrodes to detect, translate, and record the electrical signals in muscles
164 E.-A. Catalan-Salgado et al.
and nerve cells, generally used by specialist in order to determine the causes
of diseases that affect muscles and nerves, but lately have been used to make
automatic ASL translation.
Between the commercial grade devices, the Myo Arm band excels over others,
it have a EMG sensors and Inertial Movement Unit (IMU), at a relative low price,
the wearable is shown in Fig. 2.
4 Method
When we made a sign, the position of our fingers, hand and wrist changes. Thus
undoubtedly changes the tension of our forearm muscles, changing also the emg
signals. So theoretically the combination of the eight signal values in time t can
represent a sign, but can’t be directly used due to their oscillating nature, so a
processing is needed in order to be useful as a muscle activation measure.
As a muscle activation measure for a given time t, a standard deviation σ of
each sensor s is calculated for the period [t−n, t], where n is a period size. This
is calculated as follows:
t
x̄st = (1)
i=t−n
t s
i=t−n (xi− x̄st )2
σts = (2)
n−1
With this we can get the following row vector in a given time t:
Xt = σt1 , σt2 , σt3 , σt4 , σt5 , σt6 , σt7 , σt8 (3)
For sign recognition purposes the time is not needed, so we can take it out
from vector showed in 3, and let it as follows:
X = σ1 , σ2 , σ3 , σ4 , σ5 , σ6 , σ7 , σ8 (4)
166 E.-A. Catalan-Salgado et al.
4.1 Classifier
For simplicity a KNN classifier was used for testing, due to is a easy algorithm
and doesn’t have parameters that can alter or bias to the final classification,
allowing a general behaviour. In this case, we used the value 1 for K parameter.
ASL Electromiographic Sign Translator 167
X a b c d e f g h i k l m n o p q r s t u v w x y %
a 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100
b 0 4 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 40
c 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 20
d 0 0 0 5 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50
e 0 0 0 0 9 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90
f 0 6 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 40
g 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100
h 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100
X a b c d e f g h i k l m n o p q r s t u v w x y %
i 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100
k 0 0 0 0 2 0 0 0 0 0 0 0 0 2 0 0 6 0 0 0 0 0 0 0 0
l 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 100
m 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 2 0 0 0 0 0 0 80
n 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 100
o 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 100
p 0 0 0 0 0 0 1 0 0 0 1 0 0 0 6 2 0 0 0 0 0 0 0 0 60
q 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 9 0 0 0 0 0 0 0 0 90
X a b c d e f g h i k l m n o p q r s t u v w x y %
r 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 100
s 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 3 0 0 0 0 0 70
t 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 90
u 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 100
v 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 100
w 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 90
x 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 100
y 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 100
Movement Unit of the Myo Arm Band. Another way to improve this research is
the use of two arm band, one for each forearm, in order to be able to recognize
ideograms made with two hands. In fact, the research of low cost features would
benefit all the sign recognition disciplines.
References
1. ASL: Ethnologue Definition, American Sign Language (ASL) (2019). https://www.
ethnologue.com/language/ase
2. Banerjee, A.E.A.: Generation of movement explanations for testing gesture based
co-operative learning applications. In: Proceedings - 2019 IEEE International Con-
ference on Artificial Intelligence Testing (2019)
3. Chong, T.W., Lee, B.G.: American sign language recognition using leap motion
controller with machine learning approach. Sensors (Switzerland) 18, 3554 (2018)
4. Fatmi, R., Rashad, S., Integlia, R.: Comparing ANN, SVM, and HMM based
machine learning methods for American sign language recognition using wear-
able motion sensors. In: 2019 IEEE 9th Annual Computing and Communication
Workshop and Conference (2019)
5. INEGI: Disability in Mexico. INEGI (2014)
6. Lewis, M.P., Simons, G.F., Fennig, C.D.: Ethnologue: Languages of the World,
17th edn. SIL International, Dallas (2013)
7. OMS: World report on disability. OMS (2014)
8. ONU: impaired people rights conference. ONU (2017)
9. Pansare J.R., Ingle, M.: Vision-based approach for American sign language recogni-
tion using edge orientation histogram. In: 2016 International Conference on Image,
Vision and Computing, ICIVC 2016 (2016)
10. Torung V.N.T., Yang, C.K., Tran, Q.: A translator for American sign language to
text and speech. In: 2016 IEEE 5th Global Conference on Consumer Electronics,
GCCE 2016 (2016)
11. Vanco, M., Minarik, I., Rozinaj, G.: Evaluation of static hand gesture algorithms.
In: International Conference on Systems, Signals, and Image Processing (2014)
12. Sandler, W., Lillo-Martin, D.: Sign Language and Linguistic Universals. Cambridge
University Press, Cambridge (2006)
13. Wikipedia: American Sign Language. https://en.wikipedia.org/wiki/American
Sign Language
14. Wikipedia: Sign Language. https://en.wikipedia.org/wiki/Sign language
Artificial Intelligence as a Competitive
Advantage in the Manufacturing Area
Instituto Politécnico Nacional, UPIITA, G.A. Madero, 07340 Mexico City, Mexico
juvenalmv69@gmail.com, hurtadoupiicsa@yahoo.com, felnieto73@gmail.com
1 Introduction
The Oriental companies that apply the Lean concept (3rd stage), have two funda-
mental guidelines and without them it cannot operate properly [11]. The First
one, a hierarchical structure is horizontal (democratic), which implies that all
the members of the company have administrative and operational functions, so
that decision-making is the result of a consensus.
The Second one, its based on first sell and then produce, that means,
what the company is looking for is not to work blindly, the programming of
the orders is the result of real orders; for this purpose it is necessary to bring
transparency to the entire production process, making visible the stoppages in
real time; achieving this comply with the agreed with customers.
Western companies [11]. Been preferred to maintain a vertical structure
(Authoritarian), so the work environment is very stressful and they seek to pro-
duce and then sell; but when the first premise is not fulfilled, it is generated that
c Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 171–180, 2019.
https://doi.org/10.1007/978-3-030-33229-7_15
172 J. Mendoza Valencia et al.
the established work program is not achieved, generating delays, poor quality
products and penalties; the above is the result of the low participation of workers
in making decisions, as well as excessive rotation of them at different levels.
In the different research works in manufacturing companies in Mexico, it
has been found that their managers are more focused on achieving short-term
results, such as compliance with the volume of production, leaving aside quality,
variety and time of delivery.
The problem is that managers do not understand the philosophy of lean
manufacturing, so by not complying with the premises indicated above, their
companies do not have the possibility of joining a global production chain, which
requires them to comply with the times of delivery, international prices and
quality of its products.
As companies of Japanese origin have arrived, they have worked with their
suppliers in the teaching and understanding of their work methods, which has
allowed a greater number of CEO, managers and workers to understand these
forms of production.
25 years have passed since the signing of the free trade agreement between
the United States, Canada and Mexico, it is expected that in this second stage
governments will require their companies as a requirement to operate the imple-
mentation of Lean Manufacturing, which will put their companies of manufac-
turing at the same level as the Oriental companies.
With the above [10], the foundations are laid for equitable competition
between the different countries, which means that they are able to move to indus-
try 4.0. They have 5 years for the installation of 5G communication technology,
which will allow the development of the IIoT industry (Internet of industrial
things).
2 Methodology
that can bring creative ideas to the company, If this is taken into account, it
will substantially change the way it operates, being part of the solution of the
production process.
the design order is given immediately, the execution order is given, which makes
them ideal for new products, metal parts with hollow interiors, parts that are out
of the market, construction of spare parts in advance, in general manufacturing
of any pieces with irregular shapes that the only way is that this type of device,
which, once they leave the printer can begin with the assembly of the other
parts.
A problem in the manufacturing lines is that to make a product it is necessary
to have a car body that contains all the parts that it will carry, for which
manufacturers use different techniques to build these, such as injection processes
for the car of electronic products, car body building stamping, die-cutting with
oxyacetylene for metal structures, etc., which consume a lot of time for its design,
construction and installation, which makes any mistake or design change very
expensive.
On the other hand, if we use 3D printing, it would allow us to build the entire
car body of the product, make the necessary adjustments, change the design, in
an economical and fast way; if 3D printing goes hand in hand with artificial
intelligence [4], we would obtain additional potential, having the ability to see
what the market demands in real time, changes in production volume, changes
in design, order and cancellations, resulting in a decrease in costs, which will
give the company a competitive advantage.
A very revolutionary application is represented by the digital twins companies
[7], that which consists of a physical part (the product) and the virtual part
(Artificial Intelligence), within the virtual part there is a set of designs that
correspond to the different models that it has The company, at the time the
customer makes the order for any of them, the system gives a code which is
associated with the following data: customer, model and place of delivery; This
production order is transferred to the corresponding plant to define the delivery
time, in turn the plant assigns the start-finish time and the manufacturing line;
When the scheduled date arrives, the body or cabinet is manufactured in a 3D
print, a motherboard with the code corresponding to the personalization of the
product is placed.
This motherboard is part of the embedded system that will be the one that
will control the different production phases, the most important sensors in man-
ufacturing are RFID [8] (which defines the work cells) and GPS geolocation (for
external logistics), To fulfill this purpose, the communication must be wireless,
the ERP control system with Artificial Intelligence, would track the product in
real time, thus eliminating that of working blindly.
When the product leaves the production line [6], the system will define the
route and the means of transport required; The GPS geolocation sensor will keep
you informed of the time and distance traveled, in this way you can check the
total time consumed, as well as those unforeseen for future improvements in the
same process, so that in external logistics blind work is eliminated.
Once the product reaches the customer’s hands, Artificial Intelligence will
be able to monitor [14] the behavior of the product, the habits of use, the
energy consumption, the general performance of the product, for this it will
Artificial Intelligence as a Competitive Advantage in the Manufacturing Area 177
use the different sensors that have the same such as: temperature, vision, lis-
tening, speaking, vibration, GPS, RFID, etc., so that with this information the
buyer can be indicated about the required maintenance, software updates, avail-
able applications and product life time (which is depending on mechanical use),
so that suitable precautions may be taken. As the manufacturer can sell thou-
sands, hundreds or millions of products, Artificial Intelligence along with Big
Data, Analytics and the cloud, can make a comparison of the use of all its prod-
ucts, the failures, the communication problems presented, the spare parts used,
satisfaction surveys, comments on social networks of the product, sales areas,
characteristics of the population, religion professed, culture and comments on
the company, This way you can make a very approximate profile of the think-
ing of customers, to know what you like or dislike, improvements and possible
models for different ages of the population.
All of the above can not be possible today because of many countries, as well
as companies suffer from 5G communication (24 to 100 GHz transmission range),
which can work with massive information and transmit it in milliseconds; The
most extreme case is the ability to receive information from 100 devices in a
square meter, which is inferred from having a hyperconnected world.
3 Results
Companies are constantly changing due to the emergence of soft and hard tech-
nologies, which cause them competitive disadvantages, these have to be elim-
inated in order to remain in the market, for this purpose they should not be
lost in the multiple technological alternatives that are they present, for which
it is important to identify their strengths and weaknesses, to have the right
human talent and the most efficient technological option to create the competi-
tive advantage.
It is important to keep in mind that artificial intelligence, intelligent
machines, robotics, digital twins and 3D Printing are in the phase of growth
in the curve of the technological life cycle, so it has to start to be implemented
in the productive processes, with the purpose of knowing their capacities and
deficiencies, as well as defining the technological profile of the collaborators, the
training programs, the life plan, as well as the most appropriate organizational
design for the fulfillment of the objectives and goals of the company.
We must also keep in mind that we have other technologies such as, cloud
computing, big data, mobile internet, augmented reality and virtual reality, that
allow us to increase automation and efficiency in production processes, which are
complementary to the previous ones and which together will allow us to solve
complex problems of the everyday world, as would be the case with smart cities.
Which generate a large number of solutions to problems in the fields of: health,
safety, transport, housing, energy, water and food, among others.
The purpose of the 4.0 revolution is to change the current paradigms, of how
we understand technology and its future implications, to have a more inclusive
world, where ethics plays an important role in defining its regulation to maintain
a harmony between humans and machines.
178 J. Mendoza Valencia et al.
Data analytics and the use of the cloud will be part of a complex and intel-
ligent process that allows for reliable and well-processed data that helps genetic
algorithms work with quality information, generating reliable solutions for the
current requirements of the company.
For the production of a set of different and personalized goods, a real-time
transmission of information is necessary, as well as the control of the process, in
order to comply in a timely manner with what the market demands, and also
consider that the consumer does not follow any rules at the time of establishing
their requirements, this causes that you constantly have to be planning, to adapt
the production to these needs.
It is very important that companies determine the time that consumes the
different custom goods, so that in their computer platforms customers obtain the
delivery time and the price, in a few seconds, which will imply a communication
with the entire value chain (suppliers, manufacturing and distribution), in this
way comply with the agreement.
By having a more informed client, it becomes more demanding, for the reason
that in a few minutes you can visit different Web pages of each of the companies
that interest you, making it necessary to have your own vision that distinguishes
it from the rest of the competition, otherwise the chances of remembering it will
be minimal.
4 Discussion
We are getting to the point that if we want to be competitive it is important
to use smart manufacturing, cloud-based manufacturing, cognitive manufactur-
ing, etc. Which have a very close relationship with artificial intelligence, we are
allowed to design and develop intelligent systems that contribute to the creation
of machines, which have the ability to collect and analyze data, and communicate
with other systems that have the characteristic of solving complex problems.
We are in a social transformation at this time, which is reflected in a constant
change in the way of doing and seeing the world, the new generations of young
people are more qualified for work in smart factories, they are accustomed to
change, so they are more familiar with these new technological innovations, as
well as their use, application and new solutions, so they will continue being a
factor of change.
Many of the ways in which the company works will tend to disappear and
will be replaced by intelligent methods, which will be constantly changing, for
which they must have a life of their own, to deal with unexpected changes in the
production processes, models business and agile project management because it
will demand capacity and speed of response to the demand for new solutions.
With respect to manufacturing companies they have to be workplaces that
allow workers to have a good work culture, which will be reflected in the develop-
ment of their potential, with organizational structures that have as their purpose
innovation, caring for the planet and respect for the different ways of thinking,
all this within the framework of tolerance and respect for others.
Artificial Intelligence as a Competitive Advantage in the Manufacturing Area 179
We have 5 years to define how manufacturing companies 4.0 will be, in what
they have to do with their organizational structure, this is very important, since
it defines the rights and obligations of all those involved in the manufacture of
products, as well as with the plant distribution, which until now is online, which
implies the sequencing of activities. With new technologies such as: artificial
intelligence, robotics, 3D printing, Big Data, the cloud and intelligent machines,
we must think of other ways of organizing production, technological possibilities
allow.
5 Conclusions
Although much of the current literature indicates that the future of humanity
will be of abundance and mass unemployment, where machines will have control
of all daily activities, it should be remembered that at the time a technologi-
cal development is used in general, the people learn to see its advantages and
disadvantages, if this happens we can say that there is a real understanding of
it, so it will take another technological revolution to solve the new needs of the
population.
We must also bear in mind that at present people only use 10% of our brain
mass, and that as we use it more, it will necessarily have to increase the percent-
age that is not used, so we can be in a future beings that can travel in the galaxy
and see how it is and the machines will continue to be the tools that allow us to
achieve this purpose and generating the right conditions, so that human beings
can inhabit, eat and live on these planets.
The use of disruptive technologies, such as the Artificial Intelligence, is only
paving the ground for the aforementioned to happen and for human beings to
abandon the phase of scarcity of resources, hunger, injustices, etc. for a new one
where the search for happiness, finding our true selves and facing any unknown
challenges ahead will be the ideals of this new era.
References
1. Aggour, K., Gupta, V., Ruscitto, D.: Artificial intelligence/machine learning in
manufacturing and inspection: a GE perpective. MRS Bull. 44, 545–558 (2019).
Materials research society. www.mrs.org
2. Bartodziej, C.: The Concept Industry 4.0, 1st English edn. Springer, Wiesbaden
(2017). https://doi.org/10.1007/978-3-658-16502-4
3. Brau, S.: Lean 4.0 Manufacturing, 1st English edn. (2016)
4. Bukkapatnam, S., Afrin, K., Dave, D., Kumara, S.: Machine learning and AI for
long-term fault prognosis in complex manufacturing systems. CIRP Ann. - Manuf.
Technol. 68, 459–462 (2019)
180 J. Mendoza Valencia et al.
5. Ford, M.: Rise of the Robots, 1st English edn. Basic Books, New York (2015)
6. Hu, L., Miau, Y., Wu, G.: iRobot-Factory: an intelligent robot factory based on
cognitive. Future Gener. Comput. Syst. 90, 569–577 (2019)
7. Joyanes, l.: Industria 4.0, 1a edición en español. Alfaomega, México (2018)
8. Knapciková, L., Balog, M.: Industry 4.0: Trends in Management of Intelligent
Manufacturing System, 1st English edn. Springer, Cham (2019). https://doi.org/
10.1007/978-3-030-14011-3
9. Leflar, J.: Practical TPM, 1st English edn. Productivity, Portland
10. López, J.: La gran transición, 1a edición en español. Fondo de cultura económica,
México (2018)
11. Monden, Y.: Toyota Production System. 4th English edn. CRC Press, Boca Raton
(2012)
12. Moreira, L., Li, W., Fitzpatrick, M.: Supervision controller for real-time surface
quality assurance in CNC machining using artificial intelligence. Comput. Ind. Eng.
127, 158–168 (2019)
13. Ostrosi, E., Fougeres, A.: Intelligent virtual manufacturing cell formation in cloud-
based design and manufacturing. Eng. Appl. Artif. Intell. 76, 80–95 (2018)
14. Qin, J., Chian, L.: Advances and opportunities in machine learning for process
data analytics. Comput. Chem. Eng. 126, 465–473 (2019)
15. Ustundag, A., Cevikcan, E.: Industry 4.0: Managing the Digital Transformation, 1st
English edn. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-57870-5
Software Engineering and Education
Mobile Application for the Support
in the Learning of the Alphabet, Verbs
and Pronouns of the Mexican Sign
Language Based on Augmented Reality
1 Introduction
An alternative to address the child learning process for Mexican Sign Language
is through the technology we currently have, since the fast advancements of
technological developments allows to find wide possibilities in the learning envi-
ronment, such as the use of mobile devices and augmented reality. There are ben-
efits of using mobile devices in the learning process, such as: improves student’s
autonomy, facilitates feedback and helps to understand topics with multimedia
elements (by augmented reality, for example).
This learning way is good for topics like: Problems solving, to acquire skills,
languages, continuous learning, exploratory and transversal competences, games,
science and model representation. This learning method is called Mobile Learn-
ing and it is a new strategy for education. In this learning method, the materials
are important: videos, conceptual maps, images, models, conferences, simula-
tions, test, etc. The contents should be optimal and suitable for mobile devices.
c Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 183–191, 2019.
https://doi.org/10.1007/978-3-030-33229-7_16
184 A. Gordillo-Ramı́rez et al.
2 Learning Environment
The learning environment is based on the use of a mobile application, whose
development focuses on the concepts of Mobile Learning, Gamification and Aug-
mented Reality, as described in Fig. 1. The application is designed for children
in first year of primary school, and pretend to be a supporting tool for the learn-
ing of the alphabet, verbs and pronouns, that are the base for more complex
structures in the Mexican Sign Language context.
Each child must have access to a mobile device with android operating sys-
tem, which according to the statistics obtained from [7], for May 2019 has 82.28 %
of use in mobile devices in Mexico.
Each learning element (alphabet, verbs or pronouns) has a lesson, which is
carried out by the child and supervised by the teacher in the classroom or by
one of his parents in other place. Each lesson is based on the use of augmented
reality technology through printed markers. Once the child completes a lesson,
the parent or teacher can record the evaluation corresponding to the child’s
performance. There is also a game for the alphabet, which is not intended to
make an evaluation, the aim is to motivate the child to practice the elements of
the corresponding lesson of alphabet. The approach for the mobile application
development, are described as follow.
Mobile Application for the Support in the Learning of the Alphabet 185
between modules. Finally we carried out the users acceptance tests for the
mobile application, in the learning environment.
4 Access Control
Before a child start the mobile application on a device, one of the parents or
the teacher must be registered, and then log in and activate the corresponding
activity, either a lesson and its subsequent evaluation, or a game. The account
Information and Log in, is showed in Fig. 3.
5 Augmented Reality
The augmented reality in the mobile application is based on the use of the
ARToolKit API, in his 5.3.3 version, for more detail see [8]. The general operation
of this module is showed in Fig. 4.
Each element is described as follow:
– Marker. Has the design showed in Fig. 5, that contains the dimensions and
sections for each marker, and represents the relationship between signified
and signifier, in the Mexican Sign Language context. In the example, the bee
image (Abeja in Spanish) is the signified, and the image with the letter A and
the corresponding Mexican Sign Language representation, are the signifier.
This concepts presents in the marker, are crucial for language acquisition
process.
Mobile Application for the Support in the Learning of the Alphabet 187
Fig. 6. Feature map image: edge points in green crosses and textured areas in red
squares (Color figure online)
with all feature map images stored in the mobile application, referenced by
the markers.dat file, and corresponding to all recognizable markers.
– Search 3D Model. If the input marker is identified, the module should
search the corresponding 3D Model, that is referenced by the objects.dat file.
This model is the 3D graphical representation of the Mexican Sign Language
signifier corresponding in the identified input marker.
– 3D Model on Marker. In this stage, the module identifies the input marker
ubication, and show on the mobile device screen, the 3D Model over the input
marker. At this point, the user can move the input marker to left or right,
and the 3D Model reacts to this movements showing the corresponding view.
To accelerate the augmented reality processing, OpenSceneGraph (OSG).
6 Lessons
The lessons corresponding to the alphabet, pronouns and verbs in the application
are described below, in the order that which they are executed:
3. Verbs. The last lesson is verbs, we have knowledge of Spanish Language has
a lot of verbs, but first we were taking into account that in Spanish, the
LSM verbs do not indicate mode or time, also in LSM, the verb “Ser o estar”
(To be) is very rarely used, for these reasons we selected some verbs related
to the school and then we carefully chosen some of them that have easier
movements. When we selected the five verbs, we designed the markers only
with signifier represented by the word and sign because the marker is too
small to include much information. In this time, we have be careful with the
position and structure of the hands because some verbs can be confused with
other sign. The child performance is evaluated in the mobile application by
the parent or teacher. An example is showed in Fig. 9.
190 A. Gordillo-Ramı́rez et al.
7 Game
This game consists on asking for 20 random letters to child, each letter at time,
so they have to place the correspondent marker in front of the camera, allowing
the recognition of the marker and verifying if it match with the asked letter,
therefore the child can interact with the markers and the mobile application.
To implement the game mechanics we had to add a list containing the letters
and it’s corresponding marker index, the application randomly extracts a letter
and asks for it in the screen, also, a 3D model of a picture frame is displayed
to provide feedback, so the user knows the application is detecting a marker,
besides, we added a GIF image of the letter’s sign as assistance to recognize
the correspondent marker. We implemented a reward system based on trophies
represented by candies, if the child place the 20 letters correctly in a time less or
equal to two minutes, three candies will be provided, if the letters are placed in
a time higher than 2 min and less or equal to 3 min, the child will be rewarded
with two candies, and finally, if the child place the letters in a time higher than
3 min, the reward will be one candy. An example is showed in Fig. 10.
References
1. Ramı́rez Montoya, M.: Dispositivos de mobile learning para ambientes virtuales:
implicaciones en el diseño y la enseñanza. Educación de la Universidad Virtual del
Tecnológico de Monterrey. Monterrey, México (2008)
2. Aurélio Galvão, M., Roberto Zorzal, E.: Augmented Reality Applied to Health
Education. In: 2013 XV Symposium on Virtual and Augmented Reality. IEEE
(2013). https://doi.org/10.1109/SVR.2013.54
3. Vinumol, K.P., Chowdhury, A., Kambam, R., Muralidharan, V.: Augmented reality
based interactive text book: an assistive technology for students with learning dis-
ability. In: 2013 XV Symposium on Virtual and Augmented Reality. IEEE (2013).
https://doi.org/10.1109/SVR.2013.26
4. Rabelo Nazareth, D., Dos Santos Alencar, M., José Francisco de Magalhães Netto:
ELRA - teaching brazilian sign language using augmented reality. In: 2014 XVI
Symposium on Virtual and Augmented Reality. IEEE (2014). https://doi.org/10.
1109/SVR.2014.37
5. Cadeñanes Garnica, J.J., Arrieta, M.A.G.: Augmented reality sign language teach-
ing model for deaf children. In: Omatu, S., Bersini, H., Corchado, J.M., Rodrı́guez,
S., Pawlewski, P., Bucciarelli, E. (eds.) Distributed Computing and Artificial Intel-
ligence, 11th International Conference. AISC, vol. 290, pp. 351–358. Springer,
Cham (2014). https://doi.org/10.1007/978-3-319-07593-8 41
6. Al-Megren, S., Almutairi, A.: Assessing the effectiveness of an augmented reality
application for the literacy development of arabic children with hearing impair-
ments. In: Rau, P.-L.P. (ed.) CCD 2018. LNCS, vol. 10912, pp. 3–18. Springer,
Cham (2018). https://doi.org/10.1007/978-3-319-92252-2 1
7. Mobile Operating System Market Share in Mexico - May 2019 Homepage. Stat-
Counter. http://gs.statcounter.com/os-market-share/mobile/mexico. May 2019
8. ArToolkit Homepage. http://www.hitl.washington.edu/artoolkit/. May 2019
9. Mobile D. Electronics Agile Homepage. http://virtual.vtt.fi/virtual/agile/mobiled.
html (2019)
10. Neumann, U., You, S.: Natural feature tracking for augmented reality. In: IEEE
Transactions on Multimedia, vol. 1, No. 1, March 1999. https://doi.org/10.1109/
6046.748171
Augmented Reality with Swift in ARkit
and Their Applications to Teach Geometry
1 Introduction
In the past two decades, we can see several technologies of virtual reality and aug-
mented reality such as OpenGL, VRML [1], X3D that is web technology [2]. These
technologies have a limited impact on educations because of the availability of com-
puter equipment. Twenty years ago the reduction cost of computer technologies per-
mits the democratization of access to the Internet. The reality now is different mobile
devices are part of the fourth revolution [3]. Two technological firms Apple and Google
are developing two different and independent implementations called ArKit and
ArCore respectively to exploit the augmented reality market, which is becoming
standards [4]. Education is one of the greatest beneficiaries thanks several students have
access to mobile devices. It is possible to teach several mathematics topics with the
help of augmented reality. Specifically, geometry that is one of the main topics in
engineering at Universities because develops our abilities of spatial reasoning.
Examples of applications we can find in GI Science where we need to teach Cartesians
to polar transformations, robotic planning the student will need concepts such as vector,
planes in order to describe the states of an agent, etc. Swift is a programming language
born in 2014 and open to the Open Source community in 2015. The ArKit API was
released in 2017 [5] and multidevice integration in the fall of 2018 [6]. On WWDC
2019 Apple introduced new advances on Augmented Reality, ARKIT 3 and Reality
Composer. The present very interesting novelties such as: People Occlusion, Motion
Capture Multi-Face Tracking, Simultaneous front and Back camera, Collaborative
Sessions, y and improvement on image recognition. People Occlusion allows to insert
animation in front of people or hide some part of their bodies. Motion Captures scan
the body of a person to track their movement and generate a 3D model. Front a Back
camera are now available to captures images in order developers can program more
interactive applications. Face tracking allows to sense the face of a person so we can
develop more interesting applications, these features is presents in all devices with iOS
13 or higher [7]. Reality composer is a full framework designed to integrate all
applications not just. For iOS but for MACOS, integrates Animations and audios, can
record and play [8] (Fig. 1).
2 Didactic Strategy
Virtual reality and augmented reality allow us to visualize elements and environments
as never seen before. Although education has been enriched by these tools, there are
few documented cases of its use to promote the learning of abstract concepts of
mathematics and exact sciences. But these environments can be useful to create
learning objects for analytic geometry under a playful approach. In Fig. 2 we can see a
Virtual environment develop in Java 3D, that we have developed in class.
194 E. E. Loza Pacheco et al.
In contrast a we can see, the same game in Augmented reality using Arkit in Swift
in Fig. 3. The environment can be combined with elements in reality that teacher can
be used to add additional information, with the help of another element such as
blackboard or projector.
3 ArKit Implementation
Arkit implementation is simple and reduces the complexity that can be found in many
programming languages such as Java using the API Java 3D [10], or VRML [1]. For
example in Java, the programmer has to write large sums of code, in VRML the code
Augmented Reality with Swift in ARkit and Their Applications 195
has no expressivity. Instead Swift keeps important concepts such as the definition of a
root object and all objects heritage of it. As the majorities of all Graphics programming
languages, As we can see in Fig. 2 shows the design of the virtual world is the same as
in VRML. Where the “transform” node has the regular transformation actions such:
translation, rotation and scale and geometric attributes. Similar to VRML in Java 3D
we can see the same structure (see Fig. 4).
Fig. 4. (a) Semantic net of a scene for VRML and (b) Java 3D structure.
196 E. E. Loza Pacheco et al.
However Swift can create scenes from programmatically as we can see in the
previous images. Or Visually by adding objects in art.scrassets View. Here we can
modify objects such: color, transparency, transformation (translation, scale, rotation).
As we can see in Fig. 5.
The programmatical we can see ARKit divides the properties of the graphic object
in material, shape, and appearance in one leaf of the root object of the scene and the
transformation properties in another like scale, translation, and rotation. Also, we can
be found the same object definitions of geometry objects such, textures, transparency
and illumination. In Fig. 6(a) we can see a fragment of the code of a part of a scene in
Java 3D versus the Swift implementation.
Augmented Reality with Swift in ARkit and Their Applications 197
Fig. 6. (a) Java code in Java 3D and (b) Swift code in Swift ARKit.
Using ArKit in Geometry. The API allows us to create an AR scene and combine
them with real-world images. Additionally, as is common in AR implementations, AR
Kit has an implementation that permits image recognition. For this purpose, we create a
project in XCode (New iOS Project/Augmented Reality App). So we can see a envi-
ronment like this un Fig. 7. Then we add a new AR resource Group, so we can add the
set of images we would like to recognize (New AR Resource Group), the result is
shown in Fig. 8. The images are the place in the real world where Swift will place the
AR scene [11, 12].
198 E. E. Loza Pacheco et al.
After that we need to enable the application the possibility to recognize images, as
is seen in line 43 of Fig. 9 in the function viewwillAppear(…), we will need to create
the object that can discover the images. Also, we need to specify the Folder name
where the images are stored (The image should have good resolution). Then we just
need to add different images and program a different behavior for everyone.
Augmented Reality with Swift in ARkit and Their Applications 199
Having built the Environment, we need to select the size of our screen in order to
develop for that size. In Fig. 10(a), we can see the different size XCode allow us to use.
In Fig. 10(b) We can select the compatible technology we would use.
Finally, with these few lines of code, it is possible to develop an application that
works with hybrid objects. Where part of the implementation is in virtual scenes of AR
and others in a physic document. This gave us the opportunity to integrate mobile
devices and conventional educational resources. For example, whether we need to
explain planes and surfaces of Geometry. Part of the learning process can have a
theoretical background, and the practice can be the human-machine interaction with the
virtual world. In Fig. 11 we can see a small learning object applied to this matter.
200 E. E. Loza Pacheco et al.
Fig. 10. (a) Different iOS devices (b) Different version of devices.
Augmented Reality with Swift in ARkit and Their Applications 201
4 Conclusion
Modern technologies of AR are becoming very popular despite the fact there are new.
In the case of ARKit of Apple, has been eased the form we develop AR Applications.
Also, with the decreasing cost of mobile technologies. Today we have a great
opportunity to integrate AR to a variety of applications in areas where the society can
be benefited. The areas of applications are wide such as education, GIS, robotic
planning, engineering, spatial reasoning, etc.
References
1. Carey, R., Bell, G.: The Annotated VRML 2.0 Reference Manual, vol. 15. Addison-Wesley,
Reading (1997)
2. Web3d.org Homepage. What is X3D?| Web3D Consortium. http://www.web3d.org/x3d/.
Accessed June 2018
3. Toffler, A., Alvin, T.: The Third Wave, vol. 484 (1980)
4. Stallings, W.: Network and Internetwork Security: Principles and Practice, vol. 1. Prentice
Hall, New Jersey (1995)
5. Apple Inc. Homepage, ARKit. https://developer.apple.com/arkit/. Accessed June 2019
202 E. E. Loza Pacheco et al.
Keywords: Data mining Data cube Educational data mining Big data
1 Introduction
Over the last years, the term data mining has been increasingly used in scientific
literature and implies the use of different methods and tools to analyze large amount of
data. Also, data mining term refers to data sets that are so massive or complex in nature
so that traditional data processing methods and methodologies are inadequate to
effectively analyze them [1, 2]. Data mining and big data has been successfully applied
to different fields of human endeavor, including environmental and pollutant studies
[3–6], disease detection and m-health [7–10], IoT and smart cities [11–13], but it does
not only apply to these fields; these are just some of the recent fields of application of
these techniques, they also apply to economic studies such as price prediction, mar-
keting, or banking operations to name a few.
The data warehouse were created in response to the need to gather all the infor-
mation of an organization in a single database, for analysts and administrators in order
to find valuable data or verify situations that are related to different types of decisions
where it is also necessary to reduce the time to obtain it [14–16]. The generation of a
warehouse requires applying processes of extraction, integration, transformation and
cleaning data. The design and construction of a data warehouse is a complex and
delicate task [16–18], since it must be outlined the multidimensional model to be used,
as well as defining the optimal set of views to materialize. Finally, the last process
consists of determining the necessary tools for data visualization.
This paper focuses on the data organization process from a warehouse using a
software tool that automatically designs a multidimensional model and carries out the
materialization of the hypercube to subsequently apply the algorithm of selecting the
optimal set of views to materialize (Harinarayan method) [19]. The tool is also able to
determine the frequent patterns present in the data taking into account a certain support
and confidence, calculating also the association rules. The results of execution of the
software tool developed are analyzed and finally the conclusions are presented.
2 Literature Review
The use of data mining in educational fields is knowing as Educational Data Mining,
which emerges as a paradigm oriented to design models, methods and algorithms for
exploring data from educational settings [20]. In this section are analyzed some fre-
quently used educational data mining tools and techniques.
In [21] a review of works was presented; identifying statistics-visualization and
web mining as a couple of techniques to classify. Related to visualization, four works
are referenced and web mining is split into three tasks:
1. Clustering, classification and outlier detection
2. Association rules and sequential pattern
3. Text mining
Some features that are commonly present in friendly educational data mining tools
for non-technical users, are the standardization of data mining methods and data, the
integration of data mining functionalities and the design of techniques devoted to
educational data mining.
In [20] an interpretation of educational data mining review is present, it shows a
detailed analysis about:
1. Disciplines involved in data mining such as probability, machine learning, statistics,
soft computing, artificial intelligence and natural language, where probability,
machine learning and statistic are the principal approach used.
2. Data mining models, where two models are analyzed: descriptive models (unsu-
pervised learning functions to produce patterns) and predictive models (supervised
learning functions to estimate unknown or future values).
3. Data mining tasks such as clustering, association rules, correlation analysis, produce
descriptive models, classification, regression and categorization generate predictive
models. The most typical tasks are classification and clustering.
4. Data mining methods and techniques such as Bayes theorem, decision tree,
instance-based learning and hidden Markov model are the most popular methods.
Logistic, linear regression, frequencies and hierarchical clustering are the most
popular techniques.
Computer Tool Prototype for the Selections of Views 205
5. Algorithms, equations and frames used for data mining are implemented to mine the
source data. The most popular are K-means, expectation maximization and Naïve-
Bayes are the most used algorithms. Statistical equations are the most used equa-
tions. Several version of Bayesian networks are the most popular frames.
In [22] some of the tools for research and practice in educational data mining are
analyzed, discussing where relevant tools are used in data science communities, but
tools for creating structural equations model and multilevel models are not covered.
The tools analyzed are resumed as follows:
6. Two tools for manipulation, cleaning and formatting the data are discuss (Microsoft
Excel and EDM Workbench).
7. Six tools for determining the test can be conducted, what models can be con-
structed, what relationships can be mapped and explorer and how ca be validating
the findings are analyzed (RapidMiner, Weka, KEEL, KNIME, Orange and SPSS).
8. Tree visualizations tools was analyzed (Tableau, d3js and InfoVis).
In all previous work nothing is mentioned about a tool which shows visually the
implementation of an algorithm for the selection of the optimal set of views to mate-
rialize from a multidimensional cube, neither it is not taken in account the views
materialization cost. Our proposal is a tool that allows defining operations with cubes on
different dimensions from a list of business questions that can be performed on them.
The algorithm was implemented with the use of programming structures known as a
linked list, which is filled from a data source file and emptied into various views within
a database, forming what is known as the data cube. In a complementary way and to be
able to adequately construct the structure of the data cube a modified B tree is
implemented since each node has several children, it is necessary to take account that a
node, in turn, can also have several parents. In the application interface named the file
selection screen from which the user has to introduce the source file name in CSV
format. If the user does not know the path or source file name, it can be selected from a
file chooser, and once the source file is selected, the application can read the data
source chosen in CSV format.
The data cube is used to represent the facts; in this case, the learning and under-
standing of data cube generation in data mining discipline. A data cube generally is a
multi-dimensional concept; it can be from one to n dimensions. In this study, each
dimension represents a separate measure, whereas the cells in the cube represent the
facts of interest (how to generate the cube). Sometimes cubes hold a few values with
the rest being empty, e.g., undefined, sometimes most or all cube coordinates hold a cell
value of this type.
The tool was developed using Java programming language and was designed to
read any text file sequentially and interpret it appropriately if it contains the data in a
format where the column headings are in the first row separated by commas, as well as
each data within tuple. The program is responsible for calculating the views on the
MySQL database and manage a cached memory file in the form of a simple linked list
206 E. Moreno Galván and E. A. Carmona García
and a modified B tree, where each node of the tree constitutes a cuboid or view,
generating both a logical and a visual cube of data, where the calculation of data is
carried out the cardinality of each dimension and its using an algorithm [19, 23] to
determine the views to materialize. The first tests of the execution were made on a data
cube created in a synthetic way; that is, the data was created by a computer program
and the measurements were generated with random values.
Once the necessary data have been loaded, we work on the obtained database, so the
schema of a cube is composed of Dimensions (D1, D2, … , DN) that corresponds to the
attributes of the database, obtained from the column headings reading of the file that
contains the data store. The cube generated (Grill L = D1 D2 … DN) contains
as nodes the cartesian product of dimensions D1, D2, … , DN, where each node is
called cuboid and represents potentially useful aggregations [19, 23].
Once the structure is generated, in order to answer any business questions, we must
take advantage of the order in the structure of lattice that store the disk views in its
nodes, besides it may a good option to materialize or pre-calculate all the cuboids,
however the limitation of the size on disc it important to know. On the other hand, it is
recommended to materialize the total base (maximum detail or apex) since with it can
be answer any question and then move to the views with less cost, resulting into less
time in obtaining the answer.
The formulation of business questions can be carried out through expressions of the
structure query language (SQL), based on data obtained from the reading of the data file.
The results of answering expressions or queries are known as view. In addition, oper-
ators can be created in order to work with views or data cubes. So, in a query as follows:
Where:
– Field: is a subset of attributes within the database called dimensions of the form
D = {d1, d2, … , dn}
– Op(field): is an operation in a numeric type dimension, such as COUNT, SUM,
MAX, MIN.
The views are all those business questions that make use of the GROUP BY clause
of at least one field and can be executed using the apex view. The cost of a view C(v) is
the cost of using that view to evaluated the average question and is proportional to the
size of the view v. So, the construction of the lattice is done as follows:
A ! B si A B
Computer Tool Prototype for the Selections of Views 207
It has been implemented the Greedy algorithm [19] in order to get an efficient views
generation, the algorithm works as it follows: Let C(v) be the cost of a view v, which is
the cost of using that view to evaluate a business question. Due to the disk size, there is
a limit of k views (plus the total or top) to use. After selecting some set S of views, it
has to be defined the benefit of the view v relative to S, denoted by B (v, S). The only
views that benefit from having v materialized are those that can be calculated from v,
including v itself, we call each of them w. So the total benefit is the sum of all the
benefits of the w views. Figure 1 shows the resulting pseudocode.
Figure 2 shows the Compute cube implementation with average by top-k H-tree-
based implemented for the efficient views generation.
Where:
• Let C(v) be the cost of sight v
• Due the size of the disc, there is a limit of k views to use
• After selecting some set S of views (which of course includes the apex), the benefit
of the view v relative to S is defined, denoted by B(v, S)
• Calculate B(v, S), the benefit of materializing the view v when the views of the set S
have already materialized
• The only views that benefit from having v materialized are those that can be
calculated from v, including v itself. Its denotated by w
• For each w v, define Bw
• The view w can be calculated from v
• The lowest cost view is selected and denoted as u
• The benefit of Bw is C(v) < C(u), Bw = C(u) – C(v); otherwise Bw = 0
• If v has lowest
P cost than u, the benefit is the difference
• Bðv; SÞ ¼ Bw .
wv
In plain terms, the benefit of materializing v is calculated by looking at how the cost
of evaluating other views, including itself, can be improved.
5 Results
The results obtained in the tool developed when loading the generated synthetic data
are presented below. In Fig. 3 the result of loading the data in the application is
observed.
Subsequently, the views that can be generated from data cube in which the names
of the dimensions are replaced by an alias is represented in Fig. 4.
Fig. 4. Generation of data cube and its possible views. Source: Self-made.
Having all possible views to generate, we proceed to make the analysis of those that
are suitable to materialize through the use of the proposed algorithms. Each view
represents a set of fixed dimensional data in relation to a variable dimensional data. The
frequent patterns are highlighted in the interface and determined by stablishing the
required values of minimum support as Fig. 5 shows.
Finally, the existing association rules are determined by stablishing the required
values of minimum confidence as Fig. 6 shows.
Fig. 6. Associations rules finding with min support = 3 and min conf = 80%. Source: Self-
made.
The data growth is a factor that is always present in any data warehouse and the
analysis of data cubes is an area that is still being studied, and there are many types of
algorithms that address the problem in several ways. However, although the multidi-
mensional design of data benefits the way in which the data cube is materialized, the
null values within the measurement data, provide a greater benefit and a much higher
performance before the queries.
In this article, different test is presented, with combinations of four and eight
dimensions with different degrees of cardinality and associated measure. The results
were presented with respect to the processing time. As a general conclusion, it can be
inferred that the best scenario occurs when the dimension with the lowest cardinality is
in the outermost area of the data cube and the dimension with the highest cardinality is
in the innermost area. Visual learning has shown great benefits in teaching environ-
ments, the tool being visual provides the student with the ability to abstract and
organize information, allowing him to relate the complexity of data analysis in a natural
way and faster that conceptually as has been done traditionally.
Computer Tool Prototype for the Selections of Views 211
References
1. Bellazzi, R., Zupan, B.: Predictive data mining in clinical medicine: current issues and
guidelines. Int. J. Med. Inform. 77(2), 81–97 (2008). https://doi.org/10.1007/978-3-540-
30576-7183
2. Wong, Z.S.Y., Zhou, J., Zhang, Q.: Artificial intelligence for infectious disease big data
analytics. Infect. Dis. Heal. 24(1), 44–48 (2019). https://doi.org/10.1016/j.idh.2018.10.002
3. Wang, J., Boesch, R., Li, Q.X.: A case study of air quality - pesticides and odorous
phytochemicals on Kauai, Hawaii, USA. Chemosphere 189, 143–152 (2017). https://doi.org/
10.1016/j.chemosphere.2017.09.045
4. Wang, Q., Wang, J., He, M.Z., Kinney, P.L., Li, T.: A county-level estimate of PM2.5
related chronic mortality risk in China based on multi-model exposure data. Environ. Int.
110, 105–112 (2018). https://doi.org/10.1016/j.envint.2017.10.015
5. Uni, D., Katra, I.: Airborne dust absorption by semi-arid forests reduces PM pollution in
nearby urban environments. Sci. Total Environ. 598, 984–992 (2017). https://doi.org/10.
1016/j.scitotenv.2017.04.162
6. Bari, M.A., Kindzierski, W.B.: Ambient fine particulate matter (PM2.5) in Canadian oil
sands communities: levels, sources and potential human health risk. Sci. Total Environ. 595,
828–838 (2017). https://doi.org/10.1016/j.scitotenv.2017.04.023
7. Vijayakumar, P., Ganesh, S.M., Deborah, L.J., Rawal, B.S.: A new SmartSMS protocol for
secure SMS communication in m-health environment. Comput. Electr. Eng. 65, 265–281
(2018). https://doi.org/10.1016/j.compeleceng.2016.11.016
8. Kazemi, Y., Mirroshandel, S.A.: A novel method for predicting kidney stone type using
ensemble learning. Artif. Intell. Med. 84, 117–126 (2018). https://doi.org/10.1155/2018/
6798042
9. Echeverría, M., Jimenez-Molina, A., Ríos, S.A.: A semantic framework for continuous u-
health services provisioning. Proc. Comput. Sci. 60(1), 603–612 (2015). https://doi.org/10.
1016/j.procs.2015.08.187
10. Acharya, U.R., et al.: Data mining framework for breast lesion classification in shear wave
ultrasound: a hybrid feature paradigm. Biomed. Sig. Process. Control 33, 400–410 (2017).
https://doi.org/10.1016/j.bspc.2016.11.004
11. Malik, K.R., Sam, Y., Hussain, M., Abuarqoub, A.: A methodology for real-time data
sustainability in smart city: towards inferencing and analytics for big-data. Sustain. Cities
Soc. 39(April), 548–556 (2018). https://doi.org/10.1016/j.scs.2017.11.031
212 E. Moreno Galván and E. A. Carmona García
12. Ye, Y., Zhao, Y., Shang, J., Zhang, L.: A hybrid IT framework for identifying high-quality
physicians using big data analytics. Int. J. Inf. Manag. 47(January), 65–75 (2019). https://
doi.org/10.1016/j.ijinfomgt.2019.01.005
13. Allam, Z., Dhunny, Z.A.: On big data, artificial intelligence and smart cities. Cities 89
(January), 80–91 (2019). https://doi.org/10.1016/j.cities.2019.01.032
14. Zhang, L., Wen, J.: A systematic feature selection procedure for short-term data-driven
building energy forecasting model development. Energy Build. 183, 428–442 (2019). https://
doi.org/10.1016/j.enbuild.2018.11.010
15. Wang, F., Liang, J.: An efficient feature selection algorithm for hybrid data. Neurocomputing
193, 33–41 (2016). https://doi.org/10.1016/j.neucom.2016.01.056
16. Lin, Y., Wang, H., Zhang, S., Li, J., Gao, H.: Efficient quality-driven source selection from
massive data sources. J. Syst. Softw. 118, 221–233 (2016). https://doi.org/10.1016/j.jss.
2016.05.026
17. Manbari, Z., AkhlaghianTab, F., Salavati, C.: Hybrid fast unsupervised feature selection for
high-dimensional data. Expert Syst. Appl. 124, 97–118 (2019). https://doi.org/10.1016/j.
eswa.2019.01.016
18. Mahdiyah, U., Irawan, M.I., Imah, E.M.: Integrating data selection and extreme learning
machine for imbalanced data. Proc. Comput. Sci. 59(ICCSCI), 221–229 (2015). https://doi.
org/10.1016/j.procs.2015.07.561
19. Harinarayan, V., Rajaraman, A., Ullman, J.D.: Implementing data cubes efficiently.
SIGMOD (1996). https://doi.org/10.1145/235968.233333
20. Peña-Ayala, A.: Educational data mining: a survey and a data mining-based analysis of
recent works. Expert Syst. Appl. 41(4 PART 1), 1432–1462 (2014). https://doi.org/10.1016/
j.eswa.2013.08.042
21. Romero, C., Ventura, S.: Educational data mining: a survey from 1995 to 2005. Expert Syst.
Appl. 33(1), 135–146 (2007). https://doi.org/10.1016/j.eswa.2006.04.005
22. Slater, S., Joksimovic, S., Kovanovic, V., Baker, R., Gasevic, D.: Tools for educational data
mining: a review. J. Educ. Behav. Stat. 42(1), 85–106 (2017). https://doi.org/10.3102/
1076998616666808
23. Han, J., Pei, J., Dong, G., Wang, K.: Efficient computation of Iceberg cubes with complex
measures, In: Proceedings of the ACM SIGMOD International Conference on Management
of Data, vol. 30, no. 2, pp. 1–12 (2005). http://www.scopus.com/inward/citedby.url?scp=
0034825777&partnerID=8YFLogxK
Intelligent Learning Ecosystem in M-Learning
Systems
1 Introduction
The current information society poses an environment in which ICTs play a strong role
in all areas, including education [2]. This demands the establishment of pedagogical
models aimed at both the design of educational materials in digital format and their
mode of application [3] In this sense, higher education institutions have been forced to
adopt strategies to respond to new trends in information and communication by com-
peting with a greater advantage among their peers; by offering alternatives to access
knowledge different from traditional forms, which require the convergence of teacher
and student time-space. The use of information and communication technologies
(ICT) offers a wide variety of tools in the educational process, consequently, by
themselves, they do not guarantee appropriate learning, which is why the importance of
pedagogical strategies that support the learning of learning is evident. Collaborative way
Identifying student learning at any time is important, on the one hand, because of prior
learning conditions the construction of new learning and, on the other hand, because of
monitoring, both by the same apprentice and by whoever oriental is a fundamental
condition for its control and direction. This dimension of research requires the appli-
cation of measurement methods and statistical procedures [4].
The use of intelligent learning systems with analytical methods in online education
focuses on the collection, analysis, and reporting of data on student online actions [5].
The learning analysis has aroused the interest of academic institutions and teachers by
displaying information that was not available before, allowing them to better under-
stand how students learn and, therefore, to take actions with information to support the
process [6]. Advances in technology in recent years have changed student learning
behaviors and teaching methods with new environments and ubiquitous resources
methodologies. Consequently, our goal is to provide a theoretical and analytical
understanding of the pedagogical model and its implications in teaching and learning.
The analysis of the pedagogical model and didactic sequence (thematic of basic sci-
ences in oral pathology anatomy) is implemented by generating interactions with
students in the new m-learning environments, which integrate students into learning
outcomes common to all and desirable. The m-learning education allows both teachers
and students to connect through their mobile devices generating communicative
interaction.
This article reveals how technological mediation is generated in learning, from a
trans and interdisciplinary perspective, by establishing relationships between knowl-
edge. This is possible because the authors root their discourse in an epistemic frame-
work and an objective that makes all knowledge common, through reflexive and
dialogical analysis.
The technology has offered a line of advancement to the attention in intelligent learning
and the creation and efficient use of intelligent learning environments mediated with
ubiquitous technology.
their activities and are assisted by their teachers in technology-mediated learning and
web-based learning, towards mobile learning and especially “ubiquitous context-aware
learning” [13].
sharing. Although formal education differs from learning in the workplace, the stress on
students caused by fulfilling their curriculum on time while staying at the same level as
the group does not support creativity and innovation [24]. Therefore, to the extent that
the tools of learning analysis do not recognize the value of experimentation and risk-
taking, they intensify a vision of learning based on the efficiency in which failure is
penalized.
Digital data has become a key element in management techniques that are
“evidence-based.” Educational institutions are subject to a logic similar to that of
contemporary organizations, which are based on the use of data and information.
A good example of this trend can be found in the university and the school, in which
the emphasis on the indicators has been questioned, therefore hiding good practices in
teaching and learning. From this perspective, the most critical voices They affirm “the
analysis of learning, especially the academic analysis, intensifies the management
culture in education” [25], some sectors of the academic community affirm that the
analyzes can improve the understanding and performance of the students [26] For
teachers, the possibility of accessing the data generated by the students allows them to
reflect on the ins design and management of the courses they teach. In this case, the
learning analysis is presented as a tool that promotes the knowledge and reflection of
educators about some aspects of their professional practice.
Learning analysis is part of a trend based on decision-making informed by data.
From this perspective, automatic collection and analysis of student behavior data are
assumed to be relevant and reliable, or at least more reliable than subjective percep-
tions. The trust dedicated to computer algorithms is not exclusive to learning analysis,
and similar attitudes towards data in business, medical care, social services, sports, etc.
can be found. Although stakeholders in education recognize that learning analysis
enriches teaching and learning, we could question the extent to which learning analysis
is affecting the credibility we give to personal impressions. In a data-driven society, can
we rely on subjective and qualitative data collected through individual experiences?
The learning analysis modifies certain aspects of the role of teachers, especially in
online education. Here, we could say that the confidence in the data of the learning
analysis is closely related to the emergence of educational programs online. As Mazza
and [27] point out, in e-learning courses, students face related challenges such as
loneliness, technical problems or loss of motivation [28]. In these cases, the lack of
visual guidelines from teachers that help them recognize when students are poorly
motivated, anxious or overwhelmed is compensated through learning analysis. In
mixed learning scenarios, the learning analysis affects the teacher’s ability to perceive
group feedback, as there is a growing tendency to rely on the information collected
through the back channels during large conferences. The ultimate goal of these efforts
is to improve adaptation and improve teaching. But as McLuhan and McLuhan [29]
pointed out, the simultaneous effect is the disappearance of certain practices. In this
case, the praxis that is being relegated is the ability of certain teachers to detect
individual and group behaviors.
218 D. C. Burbano Gonzalez et al.
2.5 Interactions
The pedagogical design currently proposed by the traditional education system requires
the adequacy of the design of competent and individualistic producing minds that
accommodate an industrialized social context. This learning structure based on the
mass production paradigm seeks to retake the “feed-feed” model within the educational
processes; a model that uses different pedagogical practices of collaborative interaction
that involve technological and virtual communication scenarios [30].
This type of learning scenarios generally considered communication devices are
centralized in the development of participatory cognitive structures within the social
and educational environment, with the support of convergent interactive technologies
to articulated languages for different purposes [31]. This technological accompaniment
from the constant interactivity of individuals within a currently techno-mediated
environment [32], arises amid the need to create new collaborative relationships of
dissimilar transmissions.
The manifestation of these transmissions based on interaction relationships forms a
connective knowledge based on autonomy, diversity, openness and
connectivity/interactivity [31]. For Badillo, these interactions arise from the dialogue
and argumentation of individuals, which allows the reciprocal exchange of different
opinions, perspectives, and reconstruction of meanings, where new technologies are
fundamental in collective interaction, differentiators in the reorganization of mental
functions; they assign the meaningful concept of learning and enable the basic principle
of knowledge transfer to real situations [31].
learning strategies appropriate to this context can help educators facilitate the mobile
learning process and reach their educational goals. Likewise, when these materials are
properly adapted to their educational formats, the student can take full advantage of
spaces that until now were not part of the learning time [36].
It is necessary to analyze the options for adapting training content (which are
generally considered robust) to mobility scenarios, that is, transferring learning
methods to a mobile context-based on flexible and interactive styles that are not dif-
ficult to grasp [37]. The use of technology alone does not directly include improve-
ments in training. Cañizares points out through the author [38], artificial intelligence
theorist and learning expert, the skepticism about a large part of the courses offered
online; It considers that there is no sample of changes in the educational models and
they are conditioned to offer the same materials as always. Schank is committed to
learning based on experience. As a student of the human mind, he states that mental
processes evolve greatly from experience, given that he faces real situations. Contrary
to what happens with conventional learning, based on questions and answers outside
the real context and far from the individual’s praxis [36].
Education should examine how educational resources are designed and delivered
and take into account the needs and characteristics of the new student generation. For
example, in the delivery of improved technology to the student, what is the purpose of
the tool at the educational level? Does it fit the current training needs? The current
generation of students uses this technology in continuous transformation, such as
mobile devices, which require the support of teachers as well as information and timely
feedback. It must be taken into account that these changes occur according to the
training needs supported by current technologies that arise in a society with high
information demand [39]. The educational system must rescue informal learning
practices adapted to the new connective and flexible world in the use of mobile
technology, which is aimed at the autonomous and collaborative learning of students.
Without a doubt, this is a knowledge society that must be willing to take on the
challenges associated with communicative and informational technologies [40].
3 Research Design
Information and communication technologies (ICT) represent each, new challenges and
possibilities in the educational field. The term educational technology acquires a new
meaning that refers to the incorporation of ICT in teaching and learning processes. This
emerging field called technological mediations in education supposes its epistemo-
logical status, supported by positioned epistemic tendencies, which reveal cognitive
configurations that focus its interest in categories such as learning and collaboration,
privileging interactivity, under the assumption that it gives meaning to experience and
applicability of the contents learned to real-life situations.
Following the reflections of [41], it is possible to overcome the false dichotomy
between quantitative research and qualitative research in the social and human sci-
ences, for which he postulates an integrative vision called “alternative research”. “The
two types of techniques need each other most of the time … every existence or
220 D. C. Burbano Gonzalez et al.
4 Research Method
The Research-Action (I-A) Lewin’s work in the period immediately after World War II.
Lewin identified four phases in I - A (planning, acting, observing and reflecting) and
imagined it based on the principles that could lead “gradually towards independence,
equality, and cooperation” [42]. Throughout these years, the I-A method has been
configured based on numerous contributions from different geographical and ideo-
logical contexts. The great diversity of conceptions that currently exist around I-A, both
from a theoretical and experiential perspective, makes it less than impossible to arrive
at a unique conceptualization. However, there are many common features in which
most authors are coincidental. In the first place, it is worth highlighting the prepon-
derant nature of the action, as defining this research method. This dimension is spec-
ified in the active role assumed by the subjects who participate in the research, which
takes as a start the problems arising from educational practice, reflecting on them,
breaking in this way with the separatist dichotomy theory/practice. “Action research is
a form of research carried out by practitioners on their practices” [43].
5 Hypothesis
In the present investigation, the following instruments will be used for the collection of
the information: script of questions for focus group (qualitative analysis) and the
questionnaire (pre-test) in the mobile application (quantitative analysis).
In line with internationally accepted standards to ensure the ethical treatment of people
involved in scientific research, an informed consent form was developed. The document
will explain to the research participants in the focus groups the nature of the study, its
objectives, as well as the right not to participate or to suspend their participation at any
time and the confidentiality of the information provided to the participants.
In the present study the classic content analysis will be used, it implies the creation of
small pieces of data (Chunks) and the subsequent assignment of a code to each one of
them. However, instead of creating a theme from the codes, these are placed in similar
groups and then accounted for. Using the three-element coding framework there are
only three ways to use classical content analysis with focus group data: a. the analysis
can establish if each participant uses a given code; b. the analyst can determine if each
group used a given code; c. the analyst can identify all instances in which a certain code
appears [46]. Researchers should not only provide information on the frequency of
each code (that is, quantitative information) but also complement it with a description
of each code (qualitative information); which will create a mixed modality of content
analysis [47].
Fig. 1. Pre test and post test behavior. Source own elaboration
The behavior of the graph demonstrates the changes obtained once technological
mediation has been applied in periodontics residents, due to the fact that a high fre-
quency of technological and social interaction is evident. The average score that
marked the difference between pre and post was 7.86.
Technique: T-Student Test for Related Simples
The t-student test was applied, to calculate and know the values of the mean, standard
deviation and the standard mean error of the sample, as shown in Table 1.
The results show a difference of the average between the pre and post-test at a 7.867
higher in the result of the evaluation of the interactions that the post-test throws. For the
standard deviation, the dispersion of the data is more significant in the post-test with a
score of 5,244 compared to 4,837 that produced the pre-test; This is due to atypical data
presented within the post-test results.
Regarding Pearson’s linear correlation (Table 1), in this case, it is 0.100. As this
value of significance is greater than the value of alpha (0.05), it can be statistically
deduced that the scores in both registration conditions do not correlate, that is, there are
differences in their results.
224 D. C. Burbano Gonzalez et al.
“saturation”, thus the cyclical analysis of the information comparing data to form
categories, relating them to each other and resorting to new data and informants to
confirm, modify or reject them, special attention was given to even a saturation point,
that is, the new cases will not provide new information, which indicated that at that
time the analytical process should conclude.
Dialogue. In terms of dialogue, the virtual courses analyzed, from the perspective of
the students, have an appropriate level of dialogue, understood as a bidirectional,
horizontal, respectful and learning-centered relationship, between students and teachers
and between the students themselves. The intensity of the dialogue is variable, although
none of the courses presented a major difficulty that prevented the learning process.
The availability of technological tools allows an acceptable level of dialogue.
Structure. Students perceive that the courses have an appropriate level of structure:
course program, reading material, support material, links and activities scheduled
according to a schedule. However, sometimes the lack of structure is a factor that
negatively affects students’ perception of teacher performance and the degree of per-
ceived learning.
Direct instruction is exercised by teachers asynchronously, through a “mediated
dialogue”, as previously seen; They use written materials available on the platform, as
well as announcements and messages to intervene in the presentation of content. How-
ever, students must assume their understanding of the material by reading and evacuating
doubts at the time required, mainly with the use of the question forums and email.
13 Conclusions
The main contribution of this research is the approach of three elements directly
related to the positive perception of pedagogical presence by students, namely: feed-
back, academy, and socialization. In general, both students and teachers consider the
contribution that peer interaction makes learning as important.
References
1. Paramo, P.: La investigacion en ciencias sociales: estrategias de investigacion. Universidad
piloto de Colombia, Bogota, p. 332 (2003)
2. Galvis, E., Sánchez, J.: A critical review of knowledge management in software process
reference models. J. Inf. Syst. Technol. Manag. 10 (2013). https://doi.org/10.4301/S1807-
17752013000200008
3. Rosanigo, Z., Bramati, P.: Objetos de aprendizaje. XIII Workshop de Investigadores en
Ciencias de la Computación, pp. 574–869 (2011). http://sedici.unlp.edu.ar/bitstream/handle/
10915/19934/Documento_completo.pdf?sequence=1. Accessed 7 Sept 2013
4. Maldonado, L.: Virtualidad y autonomía pedagogía para la equidad. ICONK Editorial,
Colombia (2012)
5. Lockyer, L., Heathcote, E., Dawson, S.: Informing pedagogical action: aligning learning
analytics with learning design. Am. Behav. Sci. 57(10), 1439–1459 (2013)
6. Dawson, S., Gašević, D., Siemens, G., Joksimovic, S.: Current state and future trends: a
citation network analysis of the learning analytics field. In: Proceedings of the Fourth
International Conference on Learning Analytics and Knowledge, March 2014, pp. 231–240.
ACM (2014)
7. Hwang, G., Chang, H.: A formative assessment-based mobile learning approach to
improving the learning attitudes and achievements of students. Comput. Edu. 56, 1023–1031
(2011)
8. Zapata-Ros, M.: Teorías y modelos sobre el aprendizaje en entornos conectados y ubicuos.
Bases para un nuevo modelo teórico a partir de una visión crítica del “conectivismo”. Teoría
de la Educación. Educación y Cultura en la Sociedad de la Información, vol. 16 (2015)
9. Sharples, M., Arnedillo-Sánchez, I., Milrad, M., Vavoula, G.: Mobile learning. In: Balacheff,
N., Ludvigsen, S., de Jong, T., Lazonder, A., Barnes, S. (eds.) Technology-Enhanced
Learning, pp. 233–249. Springer, Dordrecht (2009). https://doi.org/10.1007/978-1-4020-
9827-7_14
10. Hwang, G., Furuchi, T., Naganuma, A.: The ubiquitin-conjugating enzymes, Ubc4 and
Cdc34, mediate cadmium resistance in budding yeast through different mechanisms. Life
Sci. 82(1182–5), 23–24 (2008)
11. Pazzani, M., Billsus, D.: Content-based recommendation systems. In: Brusilovsky, P.,
Kobsa, A., Nejdl, W. (eds.) The Adaptive Web. LNCS, vol. 4321, pp. 325–341. Springer,
Berlin (2007). https://doi.org/10.1007/978-3-540-72079-9_10
12. Chen, Y., Cheng, L., Chuang, C.: A group recommendation system with consideration of
interactions among group members. Expert Syst. Appl. 34(3), 2082–2090 (2008)
13. Liu, G., Hwang, G.: A key step to understanding paradigm shifts in e-learning: towards
context-aware ubiquitous learning. Br. J. Edu. Technol. 41, E1–E9 (2010)
14. Siemens, G., Long, P.: Penetrating the fog: analytics in learning and education. Educause
Rev. 46(5), 30 (2011)
15. De Liddo, A., Shum, S., Quinto, I., Bachler, M., Cannavacciuolo, L.: Discourse-centric
learning analytics. In: Proceedings of the 1st International Conference on Learning Analytics
and Knowledge, pp. 23–33 (2011). ACM
228 D. C. Burbano Gonzalez et al.
16. Ferguson, R., Shum, S.: Learning analytics to identify exploratory dialogue within
synchronous text chat. In: Proceedings of the 1st International Conference on Learning
Analytics and Knowledge, pp. 99–103. ACM (2011)
17. Drachsler, H., Bogers, T., Vuorikari, R., Verbert, K., Duval, E., Manouselis, N.: Issues and
considerations regarding sharable data sets for recommender systems in technology
enhanced learning. Procedia Comput. Sci. 1(2), 2849–2858 (2010)
18. Verbert, K., Drachsler, H., Manouselis, N., Wolpers, M., Vuorikari, R., Duval, E.: Dataset-
driven research for improving recommender systems for learning. In: Proceedings of the 1st
International Conference on Learning Analytics and Knowledge, pp. 44–53. ACM (2014)
19. Crick, R., Broadfoot, P., Claxton, G.: Developing an effective lifelong learning inventory:
the ELLI project. Assess. Edu.: Principles Policy Pract. 11(3), 247–272 (2004)
20. Kruse, A., Pongsajapan, R.: Student-centered learning analytics. CNDLS Thought Papers,
1–9 (2012)
21. Driscoll, M.: Psychology of Learning for Instruction. Allyn & Bacon, Needham Heights
(2000)
22. Van Harmelen, M., Workman, D.: Analytics for learning and teaching. CETIS Analytics Ser.
1(3), 1–40 (2012)
23. Combs, A.: Myths in Education: Beliefs that Hinder Progress and their Alternatives. Allyn
and Bacon, Boston (1979)
24. Wesley, D.: A critical analysis on the evolution of e-learning. Int. J. E-learn. 1(4), 41–48
(2002)
25. Selwyn, N.: Data entry: Towards the critical study of digital data and education. Learn.
Media Technol. 40, 64–82 (2014)
26. Graf, S., Ives, C., Rahman, N., Ferri, A.: A tool for accessing and analysing students’
behaviour data in learning systems. In: Proceedings of the 1st International Conference on
Learning Analytics and Knowledge, pp. 174–179. ACM (2011)
27. Mazza, R., Dimitrova, V.: Visualising student tracking data to support instructors in web-
based distance education. In: Proceedings of the 13th International World Wide Web
conference on Alternate track papers & posters, pp. 154–161. ACM (2004)
28. McLuhan, M.: Understanding Media. The Extensions of $man. Sphere Books, London
(1964)
29. McLuhan, M., McLuhan, E.: Laws of Media: The New Science, vol. 1. University of
Toronto Press, Toronto (1988)
30. Aparici, R., Silva, M.: Pedagogía de la interactividad. Comunicar: Revista Científica
Iberoamericana 19(38), 51–58 (2012). https://doi.org/10.3916/C38-2011-02-05
31. Badillo, E.: Tensiones Comunicativas Emergentes En Estrategias Colaborativo. Red de
Revistas Científicas de América Latina, El Caribe, España Y Portugal, vol. 9, pp. 188–201.
http://www.redalyc.org/articulo.oa?id=265428385012
32. Ortiz, R., Ramirez, A.: Ellos vienen con el chip incorporado. Editorial, Bogotá
33. Gros, B., Maina, M.: The Future of Ubiquitous Learning. https://doi.org/10.1007/978-3-662-
47724-3. Barcelona
34. Ramírez, M.: Modelos y estrategias de enseñanza para ambientes innovadores. In: Modelos
y Estrategias de Enseñanza para ambientes innovadores, pp. 1–55 (2013). https://www.
editorialdigitaltec.com/materialadicional/ID254_RamirezMontoya_
Modelosyestrategiasdeensenanza.cap1.pdf
35. Jeng, Y., Wu, T., Huang, Y., Tan, Q., Yang, S.: The add-on impact of mobile applications in
learning strategies: a review study. Edu. Technol. Soc. 13(3), 3–11 (2010)
36. Del Campo Cañizares, E.: M-Learning y aprendizaje informal en la educación superior
mediante dispositivos móviles. Ilu, vol. 18, pp. 231–242. https://doi.org/10.5209/rev_HICS.
2013.v18.44239
Intelligent Learning Ecosystem in M-Learning Systems 229
37. Nielsen, J.: Usability 101: Introduction to Usability. Nielsen Norman Group. https://doi.org/
10.1145/1268577.1268585. Articles
38. Schank, R.C., Fano, A., Bell, B., Jona, M.: The design of goal-based scenarios. J. Learn. Sci.
3(4), 305–345 (1994). https://doi.org/10.1207/s15327809jls0304_2
39. Vidal, M., Gavilondo, X., Rodríguez, A., Cuéllar, A.: Aprendizaje móvil. Revista Cubana de
Educacion Medica Superior 29(3), 669–679 (2015). https://doi.org/10.7238/rusc.v12i1.1944
40. Ally, M., Prieto, J.: What is the future of mobile learning in education? RUSC. Revista de
Universidad Y Sociedad Del Conocimiento 11(1), 142 (2014). https://doi.org/10.7238/rusc.
v11i1.2033
41. Paramo, P.: La investigacion en ciencias sociales: estrategias de investigacion, p. 332.
Universidad piloto de Colombia, Bogota (2013)
42. Lewin, K.: La investigación-acción y los problemas de las minorías. In: Salazar, M.C. (1992)
43. Carr, W., Kemmis, S.: Teoría crítica de la enseñanza. La investigación-acción en la
formación del profesorado. Martínez Roca, Barcelona (1988)
44. Lopez, L., Hernandez, J.: Estadística descriptiva, test y ejercicios, 2ª ed. Ediciones
Académicas. (EDIASA) edn, pp. 1–33. Madrid, España (2016)
45. Cofffey, A., Atkinson, P.: Encontrar el sentido a los datos cualitativos. Universidad de
Alicante, Alicante (2003)
46. Morgan, D.: Focus Groups as Qualitative Research. SAGE, Thousand Oaks (1997). https://
doi.org/10.4135/9781412984287
47. Onwuegbuzie, A., Teddlie, C.: A framework for analyzing data in mixed methods research.
In: Tashakkori, A., Teddlie, C. (eds.) Handbook of mixed methods in social and behavioral
research, pp. 351–383. SAGE, Thousand Oaks (2003)
Application of Cyberphysical Systems Through
Logistics as a Mediating Component
in Learning Processes
1 Introduction
Logistics has been applied since the first era of human beings, being conceived as a
component in development, necessary for survival processes, a time when tribes were
in charge of storing, transporting and distributing raw materials or food. Later, in the
Middle Ages, the first technological advances were made: wheelbarrows, boats and
horse-drawn carriages, using them in processes applied to distribution, in order to
supply points of commerce in each community.
Then, the era of the industrial revolution arose, which integrated manual and
mechanical processes, leading to the manufacture of steam engines [1]. The industry
sought to improve the quality of life of human beings with the intention of optimizing
production processes. In the 1950s, the term “integral logistics” came into being,
defined by [2] as “the control of the flow of materials from the source of supply to place
the product at the point of sale, in accordance with customer requirements and with two
basic conditioning factors: a. Maximum speed in the flow of the product and b.
2 Methodology
The development of this research used a holistic approach, based on the methodology
proposed by Jacqueline Hurtado de Barrera, the project line is implemented, with seven
necessary phases. As it can be seen in Fig. 1, entitled “Research methodology”.
For each phase presented in the methodology a description was made: Phase 1:
Exploration based on antecedents, Phase 2: Description framed in technical and general
concepts; Phase 3: Describe the process, characteristics and aspects in Logistics; Phase
4: Diagnosis, recognize how the business sector is innovating; Phase 5: Description of
the technological resource; Phase 6: Reflection on the relationship of technology in
society and Phase 7: Conclusions.
232 A. F. Córdoba U. et al.
Additionally, a serious game evaluation was carried out in order to establish dif-
ferences between the results obtained from the two analysis groups; the first one of the
Mechatronic Engineering program (GEVIM) and the second one of Industrial Engi-
neering (GEVII), each one of the groups with 15 students who belong to the semester
of VI (Manufacturing Processes) and VIII (Logistics and Distribution) of the programs.
Each student was evaluated at each of the three levels through the score obtained and
the results were compared in order to establish the characteristics in relation to the
topics of 1. Organization Process 2. Production Process 3. Transportation process.
3 Conceptual Framework
One of the biggest challenges that faces the logistics industry today is the availability of
labor. It is not easy for companies around the world to find enough high-quality
employees to move products from customers to customers. Two conflicting factors
make this especially difficult: the first is the growing need for more logistics workers
and this is being driven by the e-commerce revolution and its need for more package
shipments; the second is a decrease in the size of the available workforce due to
declining population levels in the western world.
Robotics applied to logistics and collaborative work - robots - human or robots -
robots - performs work in predetermined spaces [6], fulfilling market strategies: 1.
Movement of resources 2. Support to staff 3. Inventories of materials and 4. Distri-
bution of goods [10].
The movement of resources implies increasing the effectiveness in relation to time,
this indicates that the robots that transport goods must have capacity to detect the
obstacles that represent a decision at the time of carrying out their work, these devices
have the capacity to geo-locate themselves by means of sensors and incorporated
cameras, to distribute goods or control the storage system, an example that currently bet
on this type of technologies are Amazon, Google, Vodafone, Fedex and Alibaba,
seeking to reduce their management of processes in less time.
Application of Cyberphysical Systems Through Logistics 233
4 Technological Resource
In the development of the application it was decided to develop a serious game, in order
to appropriate knowledge, executing possible tasks that are carried out in a company.
For this purpose, the program Construct 2 was used as a development tool; creating
HTML5 applications that allow the implementation of 2D games, contributing to the
integration of learning processes for people who are not experts in the field of Logistics.
The game development process is shown in Fig. 2 named “Construct 2 Work Interface”.
The development of the game was focused in strategy mode with the tilemap
method, by means of different layers, where type layers is specified. The first layer
shows the main menu composed by the title of the game and three options, the second
option sets credits, and in addition to that you can select the exit button from the men.
The second layer constitutes the beginning of the game, demonstrating the main
objective aligned to interact within it, then press the accept button to continue with the
manipulation of the game. The game provides help information, explaining its intention
to execute a process at the shortest time, where each map corresponds to a logistics
process: 1. Organization Process 2. Production Process 3. Process of Transportation;
the fourth layer belongs to a menu of maps, evidencing three options that indicate each
logistic process, finally the option to return with the button return is offered. As it can
be seen in Table 3 named “ LogiBot Layers”.
236 A. F. Córdoba U. et al.
Map 3: Transportation
5 Discussion
The transformation in the industrial sector reflects changes from the way of managing
processes to the innovation of technological equipment used in production lines,
including cyberphysical systems; countries such as Japan attributed by its technological
scope, plans to develop robots on a large scale, wanting to fulfill tasks in any sector: 1.
Commercial 2. Industrial 3. Business and 4. Sports. One of the most important
advantages is the reduction of expenses at the moment of avoiding hiring employees; in
spite of advances obtained in technological field, allowing an efficiency and
improvement for the control of times, emphasizing that the robots have favorable
attributes, even better in performance, efficiency or force, making use of tools as the
artificial intelligence or autonomous robots in different contexts. The benefits lie in the
hyper-connectivity and automation of processes, properties highlighted under the rise
of the industrial revolution 4.0 that allows a better effectiveness at a productive level,
also it is evidenced by the transition from Digital User to Digital Citizen. In this context
a transformation takes place, this expands new forms of interaction, enabling intelligent
robotic systems, components such as: 1. Artificial Intelligence 2. Autonomous Vehicles
3. Internet of Things 4. Networked Systems 5. Big Data 6. Smart Cities 7. Drone
System and ultimately 8. Intelligent learning. The above is entitled cyberphysical
238 A. F. Córdoba U. et al.
systems [14]. For the reader’s further illustration, a graph was used, showing the
contrast captured over time in the industrial revolutions from 1.0 to the current 4.0.
This is presented in Fig. 3 named “Evolution of the Industrial Revolutions”.
In Education, refer to Intelligent Learning, authors such as Ros points out a type of
adaptive and contextualized learning with mechanisms of detection, response and
recommendation, supported by learning environments with technology, according to
the individual needs of students, where they practice analysis of learning behaviors,
performance trajectories, and online and real-world contexts. The possibility of inte-
grating artificial intelligence, making use of intelligent learning or Smart Learning in
robots applied to logistics, provides aid to learning, in relation to the fields of tech-
nology and robotics. Finally, it is concluded that social processes in relation to tech-
nology are associated with the training of Engineers in Fields of Mechatronics and
Industrial Engineering.
Taken from:
Taken from:
http://www.abconveyors.com.mx/
https://www.xdrones.es/levantamientos-
Soluciones-Roboticas/
topograficos-con-drones-fotogrametria-aerea/
5. Logistics in 3D printing
Taken from:
https://ultimaker.com/en/blog/52750-6-overlooked-benefits-of-3d-printing-for-your-
supply-chain
240 A. F. Córdoba U. et al.
cognitive learning. On the other hand, this trend is related to knowledge of Big Data to
contribute to Data Learning that seeks to establish decisions with stored data to find the
right method. At present, educational projects are presented that take the characteristics
of students such as grades, stratum, income, class schedules, and that generate plat-
forms in academic processes.
The students of each of the programs interacted with the game having only one
opportunity in each of the levels and the results obtained by them were recorded. It is
important to highlight that the score obtained by each of the participants takes into
account the fulfillment of the objectives and the time spent in the development of each
map. In Table 6 named “results by maps - game use” you can see each of the ratings
obtained by the students.
The data were analyzed descriptively in order to understand the behavior of the data
obtained for each of the analysis groups, then in Table 7 named “descriptive statistical
results for collected data”.
As it can be seen from the Table, the mean values for map 1 do not differ sub-
stantially. The opposite occurs in relation to maps 2 and 3 of each of the groups where
the calculated mean values have differences of 284.79 and 178.87 correspondingly in
favor of the results obtained by the Industrial Engineering analysis group. In relation to
the deviation of the data a very similar behavior is presented showing uniformity in the
groups of evaluation that can be seen in the Fig. 4 denominated “Radial graph for
abilities of the groups of analysis”.
Table 8. Shapiro-Wilk test for normality analysis of the knowledge test performed. Source own
elaboration
Group Statistics gl * P-Value
GEVIM1 0.944 15 0.438
GEVIM2 0.948 15 0.494
GEVIM3 0.932 5 0.293
GEIIM1 0.881 5 0,051
GEIIM1 0.893 15 0.073
GEIIM1 0.928 15 0.253
As can be seen in Fig. 4, the results obtained present a greater difference for the
concepts of production, this is mainly because the students of Mechatronics Engi-
neering do not adopt this concept within their academic curriculum. The data collected
were subjected to normality tests to subsequently perform inferential analysis of the
data in Table 8 called “Shapiro-Wilk test for normality analysis of the knowledge test
performed.” You can see the values obtained for the test statistic, for this case we
proceeded to perform the “Shapiro-Wilk test” [21] highlighted by its frequent use in
case studies and with better validity at the statistical level, it is applied in groups with
samples less than 50 individuals. The normality test raises a null hypothesis (Ho) re-
lated to the origin of the data from a normal distribution and an alternative hypothesis
(Ha) that states, as the data deviates significantly from a normal distribution; To val-
idate the hypothesis, the bilateral asymptotic significance of the test (p-Value) is taken
into account.
As can be seen in the values obtained for each group analyzed, the P-Value
obtained is greater than 0.05 (P > 0.05) which allows to accept the null hypothesis
(Ho) and affirm that the data collected comes from a Normal distribution, these results
obtained from the normality analysis allowed us to apply a parametric analysis of the
data through the T-Student test.
To analyze the differences between the groups, the T-Student test was applied. The
t-test directed to independent samples that allow comparing means of two groups,
indicating if there is statistical evidence where the associated population means are
significantly different using bilateral asymptotic significance. (P-Value). For this case it
was defined as null hypothesis (Ho) the non-existence of differences between the means
of the analytical skills and alternative hypothesis (Ha) the existence of difference
between the means of the analysis skills, They were made to the different skills, in
order to verify the existence of differences between the results obtained for each of the
maps related to the logistics processes.
Application of Cyberphysical Systems Through Logistics 243
Table 9. T-Student test for analysis of difference of means. Source own elaboration
T-test for equality of means
Ability t gl Sig. Mean Typ. error of the
(bilateral) difference difference
Organization 1,434 28 0.163 83.53 58.23
Production 6,261 28 0,000 284.8 45.49
Transport 4,371 28 0,000 178.8 40.91
As it can be seen in the Table 9, the value of bilateral asymptotic significance is less
than 0.05 (P < 0.05) for the Production and Transport cases, which leads to rejecting
the null hypothesis Ho and accepting the alternative hypothesis Ha, allowing to
establish that the scores obtained by the different groups generate differences depending
on the academic program. In the case of organizational ability, it can be observed that
the bilateral asymptotic significance is greater than 0.05 (P > 0.05), which allows
accepting the null hypothesis Ho, establishing that this competition in the organiza-
tional field does not present a difference between the two groups.
7 Conclusions
The adaptation of technological tools when applying a serious game allows to establish
didactic methodologies to evaluate the technical and technological competencies
related to logistics processes, the results obtained from the application of the t-Student
test allowed to establish statistically significant differences obtaining a P-value less than
0,05 (P < 0,05), allowing to conclude that the Mechatronics and Industrial Engineering
groups presented significant differences related to Logistics and transportation skills, in
relation to the organizational skill, did not appear statistical.
The use of a serious game related to robotics applied to logistics processes signifi-
cantly improved the Learning processes within the subject of manufacturing and logistics
processes in the Mechatronics and Industrial Engineering program of the Comfacauca
University on the other hand. The combination of technological and communication
tools Smart Factory intends to maintain an appropriate solution for efficiency, punctu-
ality, productivity and quality in the services of the company, allowing to favor mean-
ingful and collaborative learning in professionals that enhanced their logistics.
The use of Robotics, in industrial environments, in Revolution 4.0, integrates human
capabilities with potentials of robots such as repetitiveness, speed, strength, security,
allowing to improve logistics processes, about to efficiency and science. efficiency
increasing up to 20% in pre-established work in the activities developed by the operators.
References
1. Barros, T.: La Industria 4.0: Aplicaciones e Implicaciones, Universidad de Sevilla, pp. 1–52
(2017). http://bibing.us.es/proyectos/abreproy/91146/fichero/La+Industria+4.0+Aplicaciones
+e+Implicaciones.pdf
244 A. F. Córdoba U. et al.
2. Anaya, J.: Logística integral: La gestión operativa de la empresa, pp. 25–27. Libros
profesionales. ESIC Editorial, Madrid (2015)
3. Sarria, I., Lozano, B., Tenorio, W.: MÉTODO JAT JUSTO A TIEMPO, pp. 1–96.
Universidad Libre Seccional, Cali (2015)
4. Tipping, A., Prümm, D., Kauschke, P., Smith, J.: Shifting patterns. PWC 11 (2016). http://
www.ncbi.nlm.nih.gov/pubmed/8974216
5. Azmi, I., Hamid, N.A., Nasarudin, M., Hussin, M., Ibtishamiah, N.: Logistics and supply
chain management: the importance of integration for business processes. J. Emerging Econ.
Islamic Res. 5(4), 73–80 (2017). http://www.jeeir.com
6. Sanz, J.: Robots industriales colaborativos: Una nueva forma de trabajo. Segurdad y salud en
el trabajo 95, 6–10 (2018)
7. Bensusán, G.: Los mecanismos de fijación de los salarios mínimos en México en una
perspectiva comparativa: el marco institucional y los actores sociales. In: Mancera, M.A.
(Coordinador), Del salario mínimo al salario digno, pp. 205–253. Consejo Económico y
Social de la Ciudad de México, México (2015)
8. Roig, M.V.: Logística y cadena de suministro en la nueva era digital. Revista de economía,
empresa y sociedad (2018)
9. Chaparro-Peláez, J., Agudo-Peregrin, A.F., Pascual-Miguel, F.J.: Conjoint analysis of
drivers and inhibitors of E-commerce adoption. J. Bus. Res. 69(4), 1277–1282 (2016)
10. Rayo, J.: La logística en el sector industrial. Logistics Assistant en Técnicas Reunidas 12,
64–69 (2016)
11. Wu, Y., Ge, D.: Key technologies of warehousing robot for intelligent logistics. In: The First
International Symposium on Management and Social Sciences (ISMSS 2019). Atlantis Press
(2019)
12. Patiño, J.G., et al.: La cuarta revolución industrial. Ingenierías USBMed 10(1), 1 (2019)
13. Báez Sánchez, X.E.: Efectividad del “Robot Milo” en el desarrollo de habilidades sociales y
comunicación en niños de 5 a 7 años con trastorno del espectro del autismo de grado 1.
Bachelor’s thesis, Quito (2018)
14. Barrera, P.: Diseño de un en el área de la logística organizacional sistémico proyecto
educativo mediante las tic. Universidad Militar Nueva Granada 7(2), 1–21 (2017)
15. Amortegui, D., Betancurt, J., Soler, A.: Diseño de un sistema autonomo para fumigacion
aerea (2014)
16. Mendez, J.: Logística colaborativa como herramienta para mejorar el nivel de servicio y
disminuir los costos de distribución. Universidad militar nueva Granada, pp. 1–20 (2018).
https://repository.unimilitar.edu.co/bitstream/handle/10654/17745/
MendezBohorquezJulioAlexander2018.pdf?sequence=1&isAllowed=y
17. García, M.S.: Uso y aplicaciones de drones en minería (2017). http://bio-digestores.blogspot.
com.co/2012/06/uso-y-aplicaciones.html
18. Lopez, J.: Fabricación aditiva y transformación logística: la impresión 3D. Oikonomics 9,
58–69 (2018). http://oikonomics.uoc.edu
19. Telefónica, F.: Sociedad Digital en España 2018. Fundación Telefónica (2019)
20. del Val Román, J.L.: La Transformación Digital de la Industria Española. In: Conferencia de
Directores y Decanos de Ingeniería Informática Ind, p. 10 (2016). https://doi.org/10.1080/
14015430802688385
21. Dean, A., Morris, M., Stufken, J., Bingham, D.: Handbook of Design and Analysis of
Experiments, 1st edn, pp. 140–148. Chapman and Hall/CRC, Boco Raton (2015)
Internet of Things
IoT Botnets
1 Introduction
2 Botnets Background
The botnet term is a combination of two main words, Bot as a Robot abbreviation and
Net for Network; so, it can be defined as a “network of infected hosts called bots, which
are controlled by a human operator, better known as the Botmaster.”
Botnets recruit vulnerable hosts by using a type of malware that exploits vulner-
abilities; an infrastructure is created between the now infected hosts and the Botmaster
who takes the control of these hosts remotely using a Command and Control server
(C&C), through the sending of commands to perform malicious activities, such as
DDoS, spam, or information theft.
In the context of IoT, botnet is a network of compromised IoT devices, such as
cameras, modems, DVR, sensors and other devices that use the IP protocol with the
characteristic to transmit data over the Internet, infected with malware. Such networks
allow an attacker to control the devices performing malicious tasks, as well as prop-
agate their malware [5].
Fig. 1. (a) Centralized [9], (b) Decentralized [9] and Hybrid [9]
As mentioned in Sect. 2.1, for an IoT botnet the main botnet stages are the com-
munication between the botmaster and the bots due to the botmaster does not have
direct communication with the bots, it makes use of the C&C server.
250 P. Beltrán-García et al.
According to Vormayr et al. [9] botnets continuously recruit new bots by infecting
hosts for launching a specific attack, this infection stage occurs in two different ways as
exposed in Table 3.
These two infection ways are applied to IoT devices due to, i.e., infected cameras
by exploiting existing vulnerabilities such as default credentials; or like tablets where
the user infects their device by a click allowing the binary bot download, these being
helpful to understand botnet communication and therefore install detection
mechanisms.
During the operation stage, the sequence of exchanging messages depends on the
botnet specific task, IoT botnets examples are in Sect. 3.
Studying the exchanging messages could be possible to collect information packets
and determine the type of botnet used so that detection mechanisms can be used.
The use of a botnet is for several purposes such as malicious activities, better known as
attacks. In a DDoS attack, the network bandwidth is consumed by the compromised
IoT devices injecting malicious packets into the network targeting a particular server;
this means it floods the traffic with service request and processed by a server [29, 30].
Specifically, the most executed IoT botnets flooding attacks are:
• TCP SYN Flood: exploits a known weakness in the TCP connection sequence, the
requester sends multiple SYN requests, but either does not respond to the host’s
SYN-ACK response or sends the SYN requests from a spoofed IP address. Either
way, the host system continues to wait for acknowledgment for each of the requests,
making binding resources until no new connections, and ultimately resulting in a
denial of service.
• UDP Flood: a large number of User Datagram Protocol (UDP) packets are sent to a
targeted server to overwhelm that device’s ability to process and respond.
• HTTP Flood: the attacker exploits seemingly-legitimate HTTP GET or POST
requests to attack a web server or application.
• ICMP Flood: overwhelms the target resource with ICMP Echo Request (ping)
packets, generally sending packets as fast as possible without waiting for replies.
• ACK/ACK PUSH Flood: receives no legitimate ACK packets that do not belong
to any of the sessions on the server’s list of transmissions. The server under attack
then wastes all its system resources (RAM, processor, etc.) trying to define where
the packets belong. The results in productivity loss and partial server unavailability.
• TCP XMAS Flood: Send a very explicitly crafted TCP packet to a device on the
network with FIN, URG and PUSH flags set.
• DNS Flood: Attackers send validly but spoofed DNS request packets at a very high
packet rate and from a large group of source IP addresses. The DNS server can be
overwhelmed by the number of requests.
A single IoT device as bot is not a threat, but the recruitment of several bots can
attached flooding attacks, due to these devices can send connections requests affecting
the availability of services. By determining the type of a botnet attack, it is possible to
apply a network-anomaly detection technique.
4 Related Work
As briefly mentioned in Sect. 3, IoT botnets are performing many attacks. In this
section, an IoT botnet comparison is done at Table 5 based on model architecture,
attacks, target, operation, and propagation [6, 13–23, 28].
IoT Botnets 253
Table 5. (continued)
Botnet Archit. Attacks Target Infection Operation
model
additional
malicious files,
also receive a
command from
the IRC server to
perform the
attack
The Moon (2014) P2P Bring down Linksys, Active: Reverse proxy:
websites and ASUS, Exploiting a download an
servers by MikroTik, command additional
overwhelming and D-Link vulnerability in proxy module
them with routers the POST that opens a
requests; request SOCKS5 proxy
obfuscate parameter on infected
information devices, who
online knows the web
page and the
parameters to
which the
redirection of
the uplink node
must be
accessed, and
the port no
longer opens
directly on the
infected node
Bashlite (2014) IRC SYN, UDP, Linux-based Active: Brute Upload Data: In
HTTP & ACK IoT devices force with the malware´s
Flood such as default binary has the
cameras and credential of IP address of
DVRs devices with the C&C server.
open Telnet It also has the
ports IP addresses
hard-coded into
it
Mirai (2016) Centralized Generates Closed-circuit Active: 10 Upload Data:
floods of television, predefined loading the
GRE IP, cameras, attack vectors malware to the
GRE ETH, routers, and Dictionary vulnerable IoT
SYN, ACK, DVR attack based on devices
STOMP, DNS, 62 entries detected, scan
UDP, & HTTP the network for
new victims
while waiting
for instructions
from the C&C.
Scan traffic uses
random
(continued)
IoT Botnets 255
Table 5. (continued)
Botnet Archit. Attacks Target Infection Operation
model
parameters to
avoid
identification
and
fingerprinting
Linux/IRCTelnet/New IRC SYN, ACK- Routers, Active: Brute Upload Data: In
Aidra (2016) PUSH, & UDP DVR, and IP force and code the malware´s
Flood cameras injection, list of binary has the
Mirai IP address of
credentials the C&C server.
It also has the
IP addresses
hard-coded into
it
Table 5 lists the botnets analyzed in this work along with their topology, targets,
types of attacks and the infection/operation stages studied in Sect. 2; each column gives
information about how the IoT botnet operates, making possible the use of detection
mechanisms. Some mechanisms such as signatures-bases detection are referring to the
analysis of known or abnormal patterns or characteristics of threats from intruders into
a system, or by honeypots, which are used as traps to collect bot´s information and
activities, making possible to analyze them to detect botnets. These detections are
performed when the attack is executed [31].
5 Discussion
Botnets grow in size and complexity as potential nodes in the Internet of Things time to
time, making difficult to identify malicious from benign traffic even monitoring the
common ports used, like 23,2323 and 22 to gain access to the IoT device.
From a diverse set of IoT attacks described in Sect. 3 and the botnet communi-
cation in Sect. 2, different botnets were chosen in the state-of-art to identify the type of
infection and operation used. These botnets have been analysed according to their
topology, target, and type of attack, for determining the way it infects new hosts and
how it operates.
It was observed similar characteristics in most IoT botnets, such as centralized
topology due to the IoT devices resources cannot function as a server the main cons of
this topology is the single failure point by detecting the C&C server, although the pros
are management and monitoring of the botnet by the botmaster that communicates
directly with each bot, and the low latency.
As compare to decentralized and hybrid topology, there are more than one C&C
server, and because of the IoT devices resources, these can not function as a server,
these are cons. Although, the pros lies in the detection complexity of a failure point
256 P. Beltrán-García et al.
because if one C&C server is taking down, the other servers can manage and monitor
the botnet.
The application of a centralized topology still exists, by creating new botnets
following the evolution of applications on the Internet. According to the previous
research, it is provided a comprehensive state-of-art review of botnets that are evolving
time to time making important and understandable point their application to a new
target, IoT devices, highlighting how evolution and behaviour can be expanded to
perform different attacks in the IoT context.
6 Conclusion
This paper presented a comprehensive review and analysis that discusses the IoT botnet
development with distinct variation targets, attacks, and type of propagation and
operation. Being helpful to have a better knowledge of the communication between
different IoT botnets that use similar techniques or a combination of them, so that
botnet prevention can be created or even disable the botnet, i.e., the use machine
learning to extract message exchanges from network traffic for identifying possible
botnet communication.
Acknowledgment. The authors would like to thank the Instituto Politécnico Nacional (IPN), the
Centro de Investigación en Computación (CIC) and the Consejo Nacional de Ciencia y Tec-
nología (CONACYT) for the support in this research.
References
1. Escamilla-Ambrosio, P.J., Rodríguez-Mota, A., Aguirre-Anaya, E., Acosta-Bermejo, R.,
Salinas-Rosales, M.: Distributing computing in the internet of things: cloud, fog and edge
computing overview. Stud. Comput. Intell. 731, 87–115 (2018)
2. Madakam, S., Ramaswamy, R., Tripathi, S.: Internet of Things (IoT): a literature review.
J. Comput. Commun. 03(05), 164–173 (2015)
3. Nunberg, G.: The Advent of the Internet: 12th April, Courses (2012)
4. Stavrou, A., Voas, J., Fellow, I.: DDoS in the IoT Mirai and Other Botnets-2017-Computer
(2017)
5. Botnet de IoT Homepage (botnet de internet de las cosas). https://searchdatacenter.
techtarget.com/es/definicion/IoT-botnet-botnet-de-internet-de-las-cosas
6. Angrishi, K.: Turning Internet of Things (IoT) into Internet of Vulnerabilities (IoV): IoT
Botnets, pp. 1–17 (2017)
7. Tyagi, A., Aghila, G.: A wide scale survey on botnet. Int. J. Comput. Appl. 34(9), 9–22
(2011)
8. Zhaosheng, Z., Zhi, J.F., Guohan, L., Phil, R., Yan, C., Keesook, H.: Botnet research survey.
In: Proceedings of the International Computer Software and Applications Conference,
pp. 967–972 (2008)
9. Vormayr, G., Zseby, T., Fabini, J.: Botnet communication patterns. IEEE Commun. Surv.
Tutorials 19(4), 2768–2796 (2017)
10. Sundareswaran, N.: Botnet life cycle and topologies. Int. J. Pure Appl. Math. 119(17), 421–
429 (2018)
IoT Botnets 257
11. Dwivedi, S.K., Bist, A.S., Chaturvedi, P.K.: Recent trends in botnet research. Int. J. Eng.
Sci. Res. Technol. 6(7), 280–295 (2017)
12. Khattak, S., Ramay, N.R., Khan, K.R., Syed, A.A., Khayam, S.A.: A taxonomy of botnet
behavior, detection, and defense. IEEE Commun. Surv. Tutorials 16(2), 898–924 (2014)
13. De Donno, M., Dragoni, N., Giaretta, A., Spognardi, A.: DDoS-capable IoT malwares:
comparative analysis and mirai investigation. Secur. Commun. Netw. 2018, 1–30 (2018)
14. Spognardi, A., De Donno, M., Dragoni, N., Giaretta, A.: Analysis of DDoS-capable IoT
malwares. In: Proceedings of the 2017 Federated Conference on Computer Science and
Information Systems, vol. 11, pp. 807–816, September 2017
15. Meidan, Y., et al.: N-BaIoT-network-based detection of IoT botnet attacks using deep
autoencoders. IEEE Pervasive Comput. 17(3), 12–22 (2018)
16. Durfina, L., Kroustek, J., Zemek, P.: PsybOt malware: a step-by-step decompilation case
study. In: Proceedings of the Working Conference on Reverse Engineering WCRE, pp. 449–
456 (2013)
17. Hallman, R., Bryan, J., Palavicini, G., Divita, J., Romero-Mariona, J.: IoDDoS – the internet
of distributed denial of service attacks. In: IoTBDS (2017)
18. Janus, M.: Heads of the Hydra. Malware for Network Devices, 16 August 2011. https://
securelist.com/heads-of-the-hydra-malware-for-network-devices/36396/
19. Barnett, R.: New Tsunami/Kaiten Variant: Propagation Status, 11 September 2018. https://
blogs.akamai.com/sitr/2018/09/new-tsunamikaiten-variant-propagation-status.html
20. Symantec “Linux.Aidra” Writeup By: Kaoru Hayashi. https://www.symantec.com/security-
center/writeup/2013-121118-5758-99
21. Cyware “The Moon IoT botnet is proxying traffic for Youtube ad fraud scheme”, 1 February
2019. https://cyware.com/news/themoon-iot-botnet-is-proxying-traffic-for-youtube-ad-fraud-
scheme-e17d6945
22. Netlab website. https://blog.netlab.360.com/themoon-botnet-a-review-and-new-features/
23. NJ Cybersecurity and Communications Integration Cell “Linux/IRCTelnet”, 3 November
2016. https://www.cyber.nj.gov/threat-profiles/botnet-variants/linux-irctelnet
24. Silva, S.S.C., Silva, R.M.P., Pinto, R.C.G., Salles, R.M.: Botnets: a survey. Comput. Netw.
57(2), 378–403 (2013)
25. AsSadhan, B., Moura, J.M.F., Lapsley, D., Jones, C., Strayer, W.T.: Detecting botnets using
command and control traffic. In: Proceedings of the 2009 8th IEEE International Symposium
on Network Computing and Applications NCA 2009, no. 4, pp. 156–162 (2009)
26. Symantec Internet Security Threat Report: Trends for July–December 2007 (Executive
Summary), vol. XIII, April 2008
27. Makhdoom, I., Abolhasan, M., Lipman, J., Liu, R.P., Ni, W.: Anatomy of threats to the
internet of things. IEEE Commun. Surv. Tutorials 21(2), 1636–1675 (2019)
28. Ceron, J.M., Steding-Jessen, K., Hoepers, C., Granville, L.Z., Margi, C.B.: Improving IoT
botnet investigation using an adaptive network layer. Sensors 19(3), 1–16 (2019)
29. Ahmed, M.E., Kim, H.: DDoS attack mitigation in internet of things using software-defined
networking. In: Proceedings of the 3rd IEEE International Conference on Big Data
Computing Service and Applications, BigDataService 2017, pp. 271–276 (2017)
30. Gupta, B.B., Badve, O.P.: Taxonomy of DoS and DDoS attacks and desirable defense
mechanism in a cloud computing environment. Neural Comput. Appl. 28(12), 3655–3682
(2017)
31. Lange, T., Kettani, H.: On security threats of botnets to cyber systems. In: 6th International
Conference on Signal Processing and Integrated Networks, SPIN 2019, pp. 176–183 (2019)
32. Spamhaus Malware Labs: Spamhaus Botnet Threat Report 2019, pp. 1–15 (2018)
Traffic Light Control Through Intelligent
Agents, Sensors and Arduino
1 Introduction
Traffic congestion in the cities of the world is a problem which directly impacts both in
the journey time and in the number of car accidents. This kind of incidents increase
during rush hour with well-known consequences like delays, drivers and passengers
annoyance, material damages and even casualties. Traffic jam consequences often end
up in car accidents mostly because drivers lose control when they are forced to wait for
a long time in a saturated street, which in turn stresses drivers to the point of speeding
up when the street is clear. The excess of traffic may evolve into violent episodes,
disturbing other drivers and increasing time loss and traffic congestion. In the same way
other collateral damages may be produced, like excessive gas consumption while cars
wait in line for a long time in the same place, thus environmental pollution with CO2
rises considerably [1–3].
The most common system in the cities to control car flow is the traffic lights, which
basically work with green, yellow and red lights to signal go, go with caution and stop,
respectively. Their aim is to regulate and speed up car flow in the streets. However,
traffic lights nowadays do not have the ability to intelligently manage traffic congestion
on a public street at any given time; traffic lights work with timing devices which allow
cars flowing on streets according to synchronized allotted time slots. There are a few
traffic lights which consider the general state of the surrounding driveways to adjust
their performance intelligently, as to give way or stop the car flow out of their pro-
grammed schedule. This is one challenge still to be solved for IoT and smart cities
development [3].
Cars in México are the main source of pollution which contributes to the formation
of tropospheric ozone, with serious consequences on cardiopulmonary illnesses, so the
government must not keep delaying the update of regulations for emissions at national
scale, as declared by officers of the Centro Mexicano de Derecho Ambiental AC
(CEMDA, Mexican Center for Environmental Rights), the CTS Embarq México, and
International Council of Clean Transportation (ICCT); these institutions remark that
México currently has very mild standards to control emissions of ozone precursors
pollutants [2].
Particularly, the city of Mérida, capital of the state of Yucatán, has become a city
where the automobiles are reducing spaces for the more than 2 million citizens and they
are flooding the streets of the historical downtown. According to the IMPlan, the
Municipal Institute for Mérida’s Planning, six hundred thousand private motor vehicles
are estimated in the city, that is to say there is one car for every two inhabitants in
average [4]. Since Mérida is a colonial capital city, with narrow streets, it is important
to find alternatives that help clear traffic congestions which are inevitably often pro-
duced in the main streets and surrounding areas downtown Mérida [5].
This paper describes a traffic control system based on intelligent agents, sensors and
Arduino to control traffic lights to improve every day traffic and emergency vehicles.
A scale model of a two streets intersection with traffic lights and sensors is built to
perform tests.
2 Reference Framework
This project involves intelligent agents, such as deliberative, reactive and information
agents; there are also four type of sensors which are key to show the conditions on the
streets and the system’s behavior during testing.
imply that they are not able to interact with other agents to achieve specific joint
objectives. They do not save previous information therefore they are fast an efficient in
decision making but they do not adapt to change. Information agents are those that
have access to one or more sources of information and they are capable of filtrating or
manipulating data obtained from those sources in order to respond to user petitions or
to other information agents in cooperative environments. Information sources can be of
different nature from databases up to other agents. It is necessary to mention that a
particular implementation of an intelligent agent may show more than one behavior at
the same time, in which case it is classified as a hybrid agent. This allows to maximize
the strengths of some agents and minimize deficiencies of others, thus taking the best of
each type such as decision taking precision from the deliberative agents and speedy
response time from the reactive agents.
The emergency agent is essentially a reactive one and it receives readings form the
color and sound sensors to determine the presence of emergency vehicles; it disregards
information related to traffic and time of the day, and it is sensible only to red or blue
lights. Once light is detected the emergency agent compares the sound levels on each
street to select the one with the highest noise so that such a street is identified as the one
with an emergency vehicle, (police patrol, ambulance or fire truck). Once an emergency
is detected and the street is tagged as that of preferential way, the emergency agent
sends data to the control agent which adds an information component to the former
agent and it is possible to classify the emergency agent as a hybrid one. The traffic
sensor agent receives readings from the distance and motion sensors which means it is
an information agent. All three distance sensors are identified individually and they
indicate if the line of cars waiting on each street has reached a certain point (where the
sensor is placed). In other words, these sensors alert when there is a traffic jam on each
street. Also, the motion sensors placed under each of the traffic lights count the passing
vehicles and this data is saved on a file for statistical purposes. Also, there is the
intelligent control agent which collects data from the emergency and traffic agents and
262 R. A. Ojeda-Cepeda et al.
it decides the light sequence to be applied to the traffic lights in order to clear a traffic
jam or to give way to emergency vehicles. Figure 2 illustrates the described agents.
The scale model was built with the components described in detail as follows.
Arduino Mega 2560. A key component of the system is a programmable Arduino
Mega 2560 where the agents have been implemented and sensors are set to receive
information. Figure 3 shows the Arduino design with the following components.
Motion Sensor PIR HC-SR501. It detects when an object crosses directly in front of
the sensor and so it is programmed on the Arduino as a counting device. There are three
motion sensors to count vehicles passing on each street; this information shows how
many vehicles cross under the traffic lights after a light change. See number 1 on Fig. 3.
Ultrasonic Distance Sensor HC-SR04. This sensor continuously detects the distance
between its position and an object directly in front of it from 3 to 300 cm (see number 2
on Fig. 3). In this case, when consecutive readings remain constant and under 7 cm for
an allotted period of time it means that a car has remain in front of the sensor without
moving, hence it is interpreted as a jam on the street.
Sound Sensor KY038. It analogically detects the amount of generated noise in the
environment; the values range from 0 to 1023 (see number 3 on Fig. 3). The highest
value identifies the street where an emergency vehicle may be coming.
Color Sensor TCS230. This sensor is able to detect red, blue and green light around
it. To sense emergency vehicles it was programmed to react only to red and blue light,
because they are commonly used on such vehicles. The color sensor activates the sound
sensors on each street. See number 4 on Fig. 3.
3.2 Implementation
In Mérida, Yucatán, one of the busiest streets is the 50th st. which runs North to South;
at one point it is intersected by 95th street, which is a two-way road (see Fig. 4). This
intersection is represented on the scale model; it has 3 traffic lights, one is on 50th and
two more are on 95th. There are sound, distance, motion and light sensors installed on
this scenario. There are 3 ultrasonic sensors placed approximately at a distance of 3 to 5
average cars from the intersection. Traffic lights have a motion sensor installed
underneath each one. One light sensor on this scale model has been installed next to the
programmed Arduino as to have easy access to the sensor during testing. Sound sensors
are placed approximately at the same distance as the ultrasonic sensors on each street.
As it is still a prototype the whole system is sensible to malfunction of the sensors, in
which case the results will not be as expected.
4 Results
A number of tests were designed to verify the behavior of the agents on each set of
conditions (states) designed purposely to evaluate the results. Five typical states were
selected and they are explained in detail in the following paragraphs, according to the
corresponding 50th St, 95th East St and 95th West St. with a 10 s allocated time for each
traffic light on a regular basis.
1. Normal state. This state shows a normal sequence for the three traffic lights. It
starts with Traffic Light 1(TL1) allowing 8 cars to go for 10 s, followed by Traffic
Light 2 (TL2) allowing 5 cars for 10 s and finally Traffic Light 3 (TL3) allowing 3 cars
for 10 s, as detailed in Table 1. The count of cars is performed by each street´s motion
sensor. Since none of the distance, sound or color sensors are activated, the normal
sequence is followed.
2. Emergency state. When an emergency vehicle approaches on one road, the color
sensor (C) is activated by the red or blue light; the color sensors allows the sound
sensors to be evaluated in order to decide which street has an emergency vehicle
according to the highest sound sensor reading. Table 2 shows how the 95th East St is
cut short after allowing only 3 cars in 6 s on TL2 and then the sound sensors identify
50th street needing clearance for ambulance or patrol car; TL1 stays on for 20 s
allowing 6 cars to pass including the emergency vehicle. When that vehicle is far
enough and the sound is no longer perceived, the traffic lights go back to a normal
schedule, as shown in TL3 with 10 s for 4 cars to pass on 95th West St.
3. Traffic jam state. In this context, a traffic jam is identified on a street with a long
line of unmoving cars according to the ultrasonic distance sensor value which will be
constant and less than 7 cm. after an allotted time. When such a situation arises, that
sensor requests a preference on the way to clear the blocked street, in other words, the
sensor requests a green light to be activated for as long as needed and when the
blockage is no longer sensed the system will go back to normal state. Table 3 illustrates
this situation, where TL2 allows 3 cars on a normal 10 s time slot; then it is TL3 turn
but two distance sensors on streets 50th and 95th East are activated, thus cutting down
to 8 s the time of TL3 to allow the system to decide which of the two jammed streets
should be served first: it gives way to the 50th street allowing 12 cars to circulate for
15 s while 95th East street needs to wait. After 50th street is cleared, 95th street is
granted a green light and a longer time slot of 12 s of green light was needed to allow 9
cars to go. Since both jammed streets are now cleared, the traffic light schedule goes
back to the normal state with 10 s on TL 3 with only 2 cars passing through.
4. Night state. From 00:01 h to 04:00 h, the night schedule is activated; this means
that the 50th street will have only flashing yellow light to yield or slow down and cross
the road carefully; on the other hand, the drivers on both ways of the 95th street will
have flashing red lights as to indicate drivers to be aware and cross after a full stop. The
night state disregards the distance sensor since no traffic congestions are expected
during late night. See Table 4.
5. Night state with emergency. During the night, TL1 stays on flashing yellow
mode and streets 95th East and West stay on red lights. As Table 5 shows, when an
emergency vehicle is identified on 95th East street, a green light is granted for as long
as needed (in this example it was 10 s) before going back to the preventive flashing
lights.
266 R. A. Ojeda-Cepeda et al.
The intelligent agent system devised for traffic light control on a scale model has been
successful to identify a set of conditions and it has reacted adequately according to each
situation. The control agent implemented on the Arduino allows sensors to perceive the
environmental variables to test the response of the multi agent system performance in
each situation, showing good decisions to regulate traffic flow on both streets. The
results show the system is able to change the environment favorably as traffic flow
concerns. The agent system implemented on the scale model is able to identify a traffic
jam and offers a way to free the blocked roads, as showed on state 3 Table 3, which
portrays an everyday situation in that particular street crossing in Mérida. Emergency
situations are recognized and handled adequately by the control agent thanks to the
sensors reading the environment in conjunction with the programmed Arduino.
The prototype was built using the Arduino which is easy to program using a variant
of C Language. This also allows the user to modify some parameters and test the results
on the scale model.
There is also the advantage of having extra ports in which case the system can be
scaled to model more sensors o traffic lights.
There is a lot to be done as future work on this incipient project: one line of study is
for this model to be a generator of patterns which can be the input for a more complex
traffic control learning system.
The authors would like to thank the Tecnológico Nacional de México, campus
Mérida, for the support and resources devoted to this work.
References
1. New York Times. https://www.nytimes.com/2019/01/21/upshot/stuck-and-stressed-the-
health-costs-of-traffic.html?ref=nyt-es&mcid=nyt-es&subid=article. Accessed 01 Aug 2019
2. Milenio. https://www.milenio.com/estados/autos-primera-fuente-de-contaminacion-en-el-
país. Accessed 01 Aug 2019
3. INEGI. http://www3.inegi.org.mx/contenidos/app/saladeprensa/aproposito/2015/trafico31.
pdf. Accessed 01 Aug 2019
4. Diario de Yucatán. http://yucatan.com.mx/merida/urbanismo/pobre-educacion-vial. Acces-
sed 16 Apr 2019
Traffic Light Control Through Intelligent Agents, Sensors and Arduino 267
1 Introduction
Cellular communication systems will play a key role in the provision of services
in the Internet of Things (IoT) systems. This is particularly critical in scenarios
where mobile sensing objects are expected to be deployed over large areas (e.g.,
logistics, fleet management, etc.), requiring the support of cellular networks to
seamlessly transmit/receive information to/from Internet. However, the support
of IoT imposes several challenges relative to the cellular communication systems,
such as: (a) the expected number of connected devices per access point or base
station (BS), leading to a significant traffic increase per service area [1]; (b) het-
erogeneity of devices and quality of service (QoS) requirements; and (c) resource
constraints either in the device (e.g., energy, computation) or network side (e.g.,
resource availability) [2]. In order to efficiently support IoT in cellular networks,
resource management strategies are required which in turn must guarantee QoS
requirements of connected devices, as well as limiting interference and tackling
new resource bottlenecks like the backhaul [3–5].
c Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 268–283, 2019.
https://doi.org/10.1007/978-3-030-33229-7_23
Clustering for BSA Problem 269
In this context, one of the most challenging problem is to find that assignment
of BS that optimizes radio transmission to/from mobile devices. This poses an
optimization problem where the objective function is defined typically in terms
of utility and cost functions associated to a potential assignment. Such an opti-
mization problem can be subject to a set of constraints (i.e., device and system
capacity constraints) which in turn can induces non-convex solution spaces, in
which case several authors resort to heuristic approaches [CITE]. For instance,
given the heterogeneous characteristics of IoT devices, some scenarios poses con-
straints in terms of computation, sensing, and communication capabilities.
Implicitly the assignment of BS induces a partition with subsets that include
devices with “similar” features wherein the similarity can be defined in terms
of a metric (e.g. relative to channel conditions). This assumption allows us to
treat such a problem as a clustering problem. This approach has been explored
in [6], where the similarity metric is based exclusively on a proximity measure
between devices (Euclidean distance). This implies that the approach in groups
nearby devices without taking into account channel conditions (i.e., slow and fast
fading), and thus the resulting BS assignment might not be the most appropriate
from the air interface performance point of view.
In order to define a clustering-based BS assignment, we define a device rep-
resentation which is a scalar value that denotes the overall channel conditions
of each device with respect to the potential candidate set of BSs. Such represen-
tation is needed to be used as criterion to group devices according to channel
conditions instead of Euclidean distance. Furthermore, we focus on IoT cellular
systems, in which devices are expected to be connected to the Internet either via
a macro-cell BS or by means of a cluster heads acting as the serving BS or Sink.
In this work we concentrate in this latter case to determine the BS assignment
solution (i.e., the associations between IoT devices and cluster heads), where
a given mobile and the cluster head is supported via device-to-device (D2D)
communications.
1.1 Contributions
2 Related Work
The BS assignment problem has been extensively studied in the literature, where
greedy to heuristic-based algorithms have been proposed so far. On the one hand,
in greedy solutions, the devices are associated with the BS that offers the highest
signal to noise ratio (SNR). On the other hand, for heuristic approaches the
problem is formulated as an optimization problem, imposing different resource
constraints [7–10]. In [7], the authors formulated the BS assignment problem
as the Multiple-Choice Multidimensional Knapsack Problem and proposed a
heuristic approach to solve the problem. In [11] the authors studied the joint
BS assignment and resource allocation problem, and proposed solution based on
dynamical programming. Similarly, in [9] the authors modeled the problem as
a supply and demand scheme. Game theoretic-based solutions have also been
applied in [10] leading to a distributed-based assignment decision.
Recently, there is a trend to study the design of clustering-based solutions for
the BS assignment problem in cellular networks [12–14]. This is mainly boosted
by the supporting of device-to-device communications, where devices are mod-
eled to communicate directly with the minimum coordination of the BS [15–17].
Most of the approaches proposed so far are based on the extensively work pre-
sented for wireless sensor networks, where clustering solutions have been pro-
posed to perform the grouping of devices and the selection of the sink nodes
[18,19]. In particular, the attributes of devices are determined by their spatial
position inside the cell, and a clustering algorithm determines the membership
of a device to one group. The criterion to decide the clusters of devices is mainly
based on the Euclidean distance. In [20], the authors studied the support of
machine-to-machine communications under cellular networks scenarios and pro-
posed a method for grouping the devices and cluster head selection. In [13]
Clustering for BSA Problem 271
a clustering approach is studied for the joint BS assignment and load balanc-
ing problem. Authors in [6] proposed a cluster head selection and a clustering
scheme based on k-means. Similarly, in [21], the authors proposed a cluster head
selection method based on resource availability at devices. In [22], the authors
proposed a clustering solution based on physical position and social interaction
metrics between devices. The work in [23] proposed a method to group devices
that are configured to operate in D2D mode. Similarly, an energy-constrained
scenario is studied in [24], where the selection of cluster heads is determined by
the energy budget of devices.
It is worth noting that in most of the analyzed works the attributes of devices
are characterized by their position in the cell. However, this representation does
not account for the channel conditions observed between devices and their poten-
tial cluster head candidates. Moreover, devices are associated with nearest cluster
head according to the similarity function based on the Euclidean distance, which
may lead to a poor system performance since devices may not be associated with
an adequate cluster head that satisfies their minimum requirements.
3 Reference Framework
Fig. 1. Generic hierarchical single-cell scenario. Cluster heads are assumed to be con-
nected to the macro-BS, whereas the association of IoT devices to a cluster head is
determined using a clustering-based approach.
Rimin
αij = (3)
uij
where uij is the utility achieved by device i assigned to device j as a function of
the transmission rate rij achieved by device i, and αij = 1 would mean that the
Clustering for BSA Problem 273
As illustrated in Fig. 1, the BS assignment problem aims to assign each IoT device
to a given cluster head in order to optimize the system’s throughput. We assume
that IoT devices have a minimum downlink data rate requirement, whereas the
cluster heads are assumed to have a limited amount of radio resources. Over
such a basis, we define B as the assignment matrix of D × C elements, where
each element bij is determined as follows:
1, if device i is assigned to cluster head j
bij = (4)
0, otherwise
Therefore, the optimization problem is defined as follows:
C
D
max uij bij (5a)
bij
j=1 i=1
M
s.t. αij bij ≤ 1, j = 1, . . . , C (5b)
i=1
C
bij = 1, i = 1, . . . , D (5c)
j=1
Ri ≥ Rimin , i = 1, 2, . . . , D. (5d)
bij ∈ {0, 1} (5e)
Fig. 2. Characterization process of IoT devices in cellular networks: (a) Candidate set
of BSs for each user, and (b) each user characterized in terms of SNR and path loss
attributes.
Once the number of k has been found in the first phase, the process continue
to execute the k-means algorithm. Firstly, k devices are selected randomly to
be the initial centroids. Then, devices are associated with the closest cluster
head according to the maximum SNR (step 38). The centroids are updated by
calculating the mean of the devices in each cluster (step 39). This process is
repeated until no more changes are observed in the cluster centroids.
5 Simulation Experiments
In this section, we study the performance of the proposed clustering-based base
station assignment algorithm. Firstly, we developed a custom made simulator
in MATLAB to simulate the network scenario, which is composed of a single
macro-BS at the center of the cell with a radius of 500 m. Then, IoT devices
are uniformly distributed over the coverage area of the cell keeping a separation
distance of 125 m between devices and the macro-BS. The maximum transmit
power of the macro-BS is set to 47 dBm, whereas cluster heads have a transmit
power of 18 dBm. A path-loss margin of 6 dbm is selected to define the can-
didate set of each device. Moreover, all devices are assumed to have the same
downlink data rate requirement Rimin . Log normal shadowing is accounted with
standard deviation of 8 dB. The rest of simulation parameters are summarized
in Table 1. We compare the proposed clustering-based approach with the tra-
ditional k-means algorithm. For the k-means, the spatial position of the IoT
devices is considered as the attribute during the clustering process. Moreover,
an arbitrary value of k selected as the number of clusters for the k-means algo-
rithm. On the contrary, in our approach, we firstly determine the number of k
in the first stage of the algorithm, and then the k-means algorithm performs the
grouping process.
Results are divided into clustering analysis and performance evaluation of the
proposed clustering-based solution. The former consists in the analysis of the
average number of cluster heads and devices per cluster head. The latter consist
in a comparative between the proposed algorithm and the conventional k-means
algorithm.
Tables 2 and 3 show the average number of clusters and the average number
of device per clusters. The comparative was made for a variation of devices in the
system and minimum data rate requirement ranging from 15 to 100 kbps. From
the results there should be noted that as the rate requirement and the number of
devices increases, the number of formed cluster heads also increases. In addition,
there is observed that the average number of devices per cluster head remains
equal for the variation in the number of devices and the rate requirement.
Clustering for BSA Problem 277
Parameter Value
Cell radius 500 m
Operating frequency 2000 MHz
Bandwidth Macro: 10 MHz
Macro Path-Loss 128.1 + 37.6log10 (distance km)
D2C Path-Loss 148 + 40log10 (distance km)
max
Macro-BS transmit power PBS 46 dBm
Cluster head transmit power PCmax 18 dBm
Fig. 3. Comparative of the assignment solution found by the k-means and our pro-
posed clustering-based approach. The number of devices was set to 150 with a rate
requirement of 25 Kbps.
Fig. 4. Cumulative distribution function (CDF) for different data rate requirements.
The number of devices is set to M = 150, and k = 5 for the k-means algorithm.
Analysis of the Average Sum Cost: Figure 5 shows the average sum cost per
cluster head achieved by each algorithm under the variation of the number of
devices. There should be noted that in our proposed approach, we have defined
a cost constraint, which allows us to balance the load conditions between the
cluster heads. From Fig. 5, the average sum cost of our proposed approach is
below one, whereas the assignment solution found by the k-means algorithm
always exceeds the maximum capacity. For instance, in the case of 250 devices,
the k-means algorithm has an average cost of six times the maximum allowed
capacity at each cluster head. The reason behind is that the k-means algorithm
has a fixed value k which leads to overhead in clusters causing a cost increase.
Clustering for BSA Problem 281
Fig. 5. Average sum cost for each cluster head as the number of devices is varied. The
devices’ data rate requirement is also varied ranging from 15 to 30 kbps.
6 Concluding Remarks
In this paper, a novel clustering-based base station assignment algorithm has
been proposed to support the deployment of IoT devices under cellular com-
munication systems. Unlike other clustering approaches, in our solution, we
have developed a device representation that accounts for the channel conditions
between the device and its potential candidates. We have shown through exper-
iments that our proposed approach has significant gains in terms of the average
sum throughput. Moreover, in terms of the average sum cost between the cluster
heads, our approach has a better performance than the k-means algorithm. As
future work, we aim to improve the process of the cluster heads discovery by
exploring combinatorial approaches. In addition, we aim to study the behavior of
282 E. A. Esquivel-Mendiola et al.
other clustering algorithms from data mining for examining the behavior under
the massive presence of IoT devices.
References
1. Cisco Systems: Cisco visual networking index: Global mobile data traffic forecast
update, 2017–2022 white paper. Technical report, Cisco Systems, February 2019
2. Dhillon, H.S., Huang, H., Viswanathan, H.: Wide-area wireless communication
challenges for the internet of things. IEEE Commun. Mag. 55(2), 168–174 (2017)
3. Ge, X., Tu, S., Mao, G., Wang, C., Han, T.: 5G ultra-dense cellular networks.
IEEE Wirel. Commun. 23(1), 72–79 (2016)
4. Jafari, A.H., López-Pérez, D., Song, H., Claussen, H., Ho, L., Zhang, J.: Small
cell backhaul: challenges and prospective solutions. EURASIP J. Wirel. Commun.
Netw. 2015(1), 206 (2015)
5. Cheng, L., Gao, Y., Li, Y., Yang, D., Liu, X.: A cooperative resource allocation
scheme based on self-organized network in ultra-dense small cell deployment. In:
2015 IEEE 81st Vehicular Technology Conference (VTC Spring), pp. 1–6, May
2015
6. Zhao, Y., Liu, K., Xu, X., Yang, H., Huang, L.: Distributed dynamic cluster-head
selection and clustering for massive IoT access in 5G networks. Appl. Sci. 9(1),
132 (2019)
7. Galeana-Zapien, H., Ferrus, R.: Design and evaluation of a backhaul-aware base
station assignment algorithm for OFDMA-based cellular networks. IEEE Trans.
Wirel. Commun. 9(10), 3226–3237 (2010)
8. Dhillon, H.S., Andrews, J.G.: Downlink rate distribution in heterogeneous cellular
networks under generalized cell selection. IEEE Wirel. Commun. Lett. 3(1), 42–45
(2014)
9. Shen, K., Yu, W.: Downlink cell association optimization for heterogeneous net-
works via dual coordinate descent. In: 2013 IEEE International Conference on
Acoustics, Speech and Signal Processing, pp. 4779–4783, May 2013
10. Semiari, O., Saad, W., Bennis, M.: Downlink cell association and load balanc-
ing for joint millimeter wave-microwave cellular networks. In: 2016 IEEE Global
Communications Conference (GLOBECOM), pp. 1–6, December 2016
11. Lai, W.K., Liu, J.: Cell selection and resource allocation in LTE-advanced hetero-
geneous networks. IEEE Access 6, 72978–72991 (2018)
12. Hajjar, M., Aldabbagh, G., Dimitriou, N., Win, M.Z.: Hybrid clustering scheme for
relaying in multi-cell LTE high user density networks. IEEE Access 5, 4431–4438
(2017)
13. Kollias, G., Adelantado, F., Verikoukis, C.: Spectral efficient and energy aware clus-
tering in cellular networks. IEEE Trans. Veh. Technol. 66(10), 9263–9274 (2017)
14. Kazmi, S.M.A., Tran, N.H., Ho, T.M., Manzoor, A., Niyato, D., Hong, C.S.: Coor-
dinated device-to-device communication with non-orthogonal multiple access in
future wireless cellular networks. IEEE Access 6, 39860–39875 (2018)
15. Tehrani, M.N., Uysal, M., Yanikomeroglu, H.: Device-to-device communication in
5G cellular networks: challenges, solutions, and future directions. IEEE Commun.
Mag. 52(5), 86–92 (2014)
16. Vlachos, C., Friderikos, V.: Optimal device-to-device cell association and load bal-
ancing. In: 2015 IEEE International Conference on Communications (ICC), pp.
5441–5447, June 2015
Clustering for BSA Problem 283
17. Xiao, S., Zhou, X., Feng, D., Yuan-Wu, Y., Li, G.Y., Guo, W.: Energy-efficient
mobile association in heterogeneous networks with device-to-device communica-
tions. IEEE Trans. Wirel. Commun. 15(8), 5260–5271 (2016)
18. Rostami, A.S., Badkoobe, M., Mohanna, F., Keshavarz, H., Hosseinabadi, A.A.R.,
Sangaiah, A.K.: Survey on clustering in heterogeneous and homogeneous wireless
sensor networks. J. Supercomputing 74(1), 277–323 (2018)
19. Sarkar, A., Murugan, T.S.: Cluster head selection for energy efficient and delay-less
routing in wireless sensor network. Wirel. Netw. 25(1), 303–320 (2019)
20. Tu, C., Ho, C., Huang, C.: Energy-efficient algorithms and evaluations for massive
access management in cellular based machine to machine communications. In: 2011
IEEE Vehicular Technology Conference (VTC Fall), pp. 1–5, September 2011
21. Fodor, G., Parkvall, S., Sorrentino, S., Wallentin, P., Lu, Q., Brahmi, N.: Device-
to-device communications for national security and public safety. IEEE Access 2,
1510–1520 (2014)
22. Wang, L., Araniti, G., Cao, C., Wang, W., Liu, Y.: Device-to-device users clustering
based on physical and social characteristics. Int. J. Distrib. Sens. Netw. 2015(8),
1:1 (2015)
23. Koskela, T., Hakola, S., Chen, T., Lehtomaki, J.: Clustering concept using device-
to-device communication in cellular system. In: 2010 IEEE Wireless Communica-
tion and Networking Conference, pp. 1–6, April 2010
24. El-Feshawy, S.A., Saad, W., Shokair, M., Dessouky, M.I.: An efficient cluster-
ing design for cellular based machine-to-machine communications. In: 2018 35th
National Radio Science Conference (NRSC), pp. 177–186, March 2018
Fuzzy Logic Controller for Automation
of an Autonomous Irrigation System Designed
for Habanero Pepper
(Capsicum Chinense Jacq.)
1 Introduction
The origin of irrigation systems goes back thousands of years when the rivers were
used to irrigate the crops using a staircase design [1]. However, nowadays there is a
water crisis due to the misuse that is made of it, causing the environment and millions
of people to suffer the consequences [2]; according to the World Water Council, there
are six major problems related to water: (1) scarcity, (2) lack of access, (3) decrease in
quality, (4) decrease in the allocation of financial resources, (5) fragmentation in the
administration and (6) the lack of awareness of the people when making an inadequate
use of the water [3]. The Organization of the United Nations and Agriculture reports
that 70% of the world’s water is used mainly for activities such as agriculture [4].
Habanero pepper (Capsicum chinense Jacq.), after the tomato, is one of the main
products of agriculture, grown in the states of Baja California Sur, San Luis Potosi,
Sonora, Tabasco and in the states that form the Yucatan peninsula: Campeche,
Quintana Roo and Yucatan, which have the denomination of origin of habanero pepper
since 2010 [5]. The traditional cultivation techniques for this plant are obsolete, this is
because the producers are limited in technology, income, among others [6]. According
© Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 284–293, 2019.
https://doi.org/10.1007/978-3-030-33229-7_24
Fuzzy Logic Controller for Automation of an Autonomous Irrigation System 285
to [7], the cultivation of habanero peppers is generally in the open field and with
manual irrigation, which ends up affecting production by diverse environmental fac-
tors, capable of reducing the quality of the product, yield and profitability of the
crop. Nowadays, the way to improve the crop has been sought using a controlled area
where external factors are reduced [8].
Traditional farming techniques such as conventional irrigation systems have
become obsolete due to limited technology and lack of income from producers. This is
why irrigation systems have been updated to date by using automatic, autonomous and
even intelligent controllers in order to improve the crop and reduce the physical work
of farmers to a level where they only are required to perform a supervision of the
system instead of a manual control of it [9].
Fuzzy logic is a branch of artificial intelligence that refers to the way people
describe the environment that surrounds them, so it has turned out to be successful
when applied to problems related to uncertainty and vagueness [10]. There are con-
trollers based both on classic crisp sets and fuzzy sets. Crisp sets are sets with a discrete
border and are an important tool for mathematics and science, however, when used in
control applications, they do not demonstrate the thinking of the human being or the
way in which he or she interprets the surrounding environment. Crisp sets are defined
as a collection of objects or elements, which can belong to the set or not (true or false,
respectively); there are no intermediate values [11]. On the contrary, a fuzzy set does
not have a discrete border, so that belonging or not to some set occurs gradually [12].
The fuzzy set is an extension of the classical set, to which a defined membership
function is added with a range between zero and one. A fuzzy set allows then to
simulate the process of human reasoning to control systems based on their experience
and solve problems using mathematical methods based on degrees of membership or
belonging, associating them with linguistic values, defined with words or labels used
naturally by a human being. When modeling information using statements that are not
totally true or totally false, they are applied to expressions that can take a truth value
from a set of values composed of two extremes and all their possible intermediate
values. This modeling is carried out by using linguistic rules that approximate functions
of inputs and outputs of the system [11].
Incorporating fuzzy logic with control systems results in fuzzy control systems or
fuzzy inference mechanisms, which are responsible for determining what should be
done to fulfill a process in the best possible way based on knowledge and experiences
provided by a human being in charge of operating the system, since without the
knowledge base it is impossible to develop a controller that works correctly [13].
A system based on fuzzy logic necessarily requires the use of different concepts such as
fuzzy set, membership function, rules, linguistic variables and values [14]. The
membership function is responsible for representing the degree of belonging of an
element to a set, that is, represents the degree to which an element is part of such a set
[14]. There are different ways to represent membership functions, among the most
common are Gaussian, triangular, singleton, trapezoidal and sigmoidal. If-then rules are
used and form a fundamental component in an inference system that can represent
human experience in a specific process of a controller [14].
A fuzzy inference system (FIS) is a computational model based on fuzzy sets, rules
and reasoning. To develop a FIS three conceptual components are required: (1) a set of
286 J. Molina-Puc et al.
rules with a selection of fuzzy rules, (2) a database in which the membership functions
to be used in the fuzzy rules are defined, and (3) a reasoning mechanism. There are
basic inference systems in which you can have fuzzy and discrete inputs, but the output
must always be fuzzy, however, sometimes it is necessary to have a discrete output, as
in the case of controllers. In these cases, it is necessary to have a defuzzification method
to obtain a discrete value that best represents the fuzzy set. The different types of fuzzy
inference systems are Mamdani, Sugeno and Tsukamoto; the type of system that should
be used depends primarily on the nature of the output [11].
Currently, there are works that implement fuzzy logic in control modules, which
logically determine the solutions to achieve the control objectives from the knowledge
that is provided by the human being [11]. In [15] the operation of a fuzzy controller
applied to an irrigation system is shown; the controller uses triangular and trapezoidal
membership functions for the fuzzification of the variables and a rule base of the
Mamdani type. In [16] a fuzzy logic controller is used, also based on the Mamdani
model for the automation of a greenhouse irrigation system, taking into account as the
input value the difference between the desired soil moisture and the actual soil mois-
ture, and as output value a valve that allows to open and close the water passage; this is
done to effectively estimate the amount of water avoiding that too much water is
applied to the plants. In [17] an irrigation system based on the information provided by
humidity, temperature and soil moisture sensors is proposed; considering the inputs, the
output of the system determines the irrigation time for the crop while optimizing water
use. In [18] and [19] climatic factors are used to determine the ignition of the irrigation
pump of a greenhouse and then send the information obtained by the sensors by SMS
and perform operations with a smartphone. In [20] an irrigation system is proposed that
allows the optimization of water, using humidity, temperature sensors and a flow sensor
in the field as inputs for a fuzzy Mamdani controller, while the outputs are the ignition
of a lamp and an irrigation pump. Finally, in [21] an intelligent irrigation system is
presented that uses an Android application to optimize water consumption in medium-
scale gardens and fields using sensors that capture plant data and weather factors in real
time. Fuzzy logic is used for decision making and activating the irrigation pump.
This paper presents the design and development of a controller for an automated
irrigation system using fuzzy logic. The controller developed in this project uses
sensors to measure soil moisture, ambient humidity, luminosity and ambient temper-
ature to determine the activation of an irrigation pump. The soil moisture is periodically
monitored to determine if the humidity needed to measure the other sensors is met and,
if so, the data obtained from the sensors feeds a Sugeno FIS based on specific rules for
habanero pepper plants; the system output activates or deactivates a pump for drip
irrigation in a greenhouse.
2 System Design
The development of the controller consists of two subsystems, (1) detection and control
by means of the sensors, and (2) the internal development of the controller. The
development of these subsystems is described in more detail in the following
subsections.
Fuzzy Logic Controller for Automation of an Autonomous Irrigation System 287
A conventional solenoid valve is used to supply the water, since the irrigation
system works only in two states: on and off. The solenoid valve is connected to a hose
that works in conjunction with sprinklers and is activated by the fuzzy controller when
complying with the rules presented in Table 1. These rules were defined by interviews
with experts in the care of habanero pepper when presented with different scenarios
with respect to the input variables. Soil moisture sensors are the ones that mainly
determine if the activation of the system is necessary, because when the soil is com-
pletely humid, irrigation is not carried out regardless of the values of the other sensors,
however, if the soil moisture sensors are in a “not wet” range (as described below), then
the values of the other sensors are considered to determine the activation of the sole-
noid valve with the calculation of the rules.
Table 1. Rules to determine whether the irrigation pump should be turned on or off, according
to experts in the care of habanero pepper.
Soil moisture Ambient temperature Ambient humidity Luminosity Irrigation pump state
High – – – Off
Medium Low High – Off
Medium Low Medium Not low Off
Medium Low Medium Low On
Medium Low Low Not low Off
Medium Low Low Low On
Medium Medium – Not low Off
Medium Medium – Low On
Medium High – Not low Off
Medium High – Low On
Low Low High – On
Low Low Medium – On
Low Low Low Not high On
Low Low Low High Off
Low Medium – Not high On
Low Medium – High Off
Low High – Not high On
Low High – High Off
functions, whose parameters were defined empirically and, in the same way as the
fuzzy rules, by means of interviews with experts in the care of habanero peppers. These
values represent the input of the FIS, whose output determines the switching off or on
of the irrigation pump. All the membership functions of the output variables in a
Sugeno FIS are linear, and in our case, constants of the form z = c, where z represents
the output variable and c a scalar value, in our case 0 (off) or 1 (on).
Fig. 3. The membership functions used in the FIS, (a) those for soil moisture, (b) luminosity,
(c) ambient humidity and (d) ambient temperature.
A value of Z close to 1 turns on the irrigation system (or keeps it on, according to
the situation), while a value close to 0 turns it off (or keeps it off).
3 Results
According to the input values of each sensor, the changes in the readings of the values
obtained can be observed, therefore, compliance with the rules is evaluated to deter-
mine whether the solenoid valve should be turned on or off. Figure 4 graphically shows
the operation of the system. Each row represents each of the 18 rules that regulate the
behavior of the controller. The first four columns represent the different variables
obtained from the sensor data: soil moisture, luminosity, ambient temperature and
ambient humidity. The last column represents the activation (or not) of the irrigation
pump based on each of the rules. The graph in the lower right corner represents the
aggregation of the different outputs for each rule; it is this result that determines
whether the irrigation pump is to be turned on or off.
Fuzzy Logic Controller for Automation of an Autonomous Irrigation System 291
Fig. 4. Result of the operation of the FIS. Each row represents each of the 18 rules that regulate
the behavior of the controller. The first four columns represent the different variables obtained
from the sensor data. The last column represents the activation (or not) of the irrigation pump
based on each of the rules. The graph in the lower right corner represents the aggregation of the
different outputs for each rule. The graph shows a case of 35.3% moisture in the soil, 72.2%
luminosity, 26.7 °C ambient temperature and 78.4% ambient humidity; the irrigation pump
would be turned off in this case due to an aggregated value close to 0.
The operation of the fuzzy controller turns out to be as expected, since the irrigation
system stays off when the soil is totally humid and there are high temperatures. In
addition, the controller correctly complies with readings of the values of the sensors
every 10 min when the pump is off and readings every 2 min when it is turned on.
4 Conclusions
In this work we describe the design of a fuzzy controller based on the Sugeno model for
the control of an automatic irrigation system specifically designed for habanero pepper.
Based on interviews with experts in the care of habanero pepper, 18 if-then fuzzy rules
were established and the different membership functions corresponding to linguistic
values of the four linguistic variables obtained from different sensors were designed:
soil moisture, luminosity, ambient temperature and ambient humidity. These mem-
bership functions were designed with Gaussian and sigmoidal functions.
The implementation of fuzzy logic in a controller for an irrigation system implies a
great impact and benefit for agriculture and the people who dedicate themselves to this
activity, due to the automation of tasks that allows people to reduce time and processes
that would normally need to be carried out by more than one person. It allows pre-
serving the care of the water by having an automatic on/off control based on weather
factors, in order to care for crops and thus help the farmer to increase productivity. The
design of an irrigation system like the one proposed in this paper can also be extended
292 J. Molina-Puc et al.
to the specific conditions of plants other than the habanero pepper plant by redesigning
the fuzzy rules and associated linguistic values.
References
1. Traxco: Historia ancestral del riego agrícola - El agua de riego. https://www.traxco.es/blog/
noticias-agricolas/historia-ancestral-del-riego
2. World Water Council: Visión Mundial del Agua-Resumen Ejecutivo. http://www.
worldwatercouncil.org/fileadmin/wwc/Library/Publications_and_reports/Visions/Spanish
ExSum.pdf
3. Ruiz, F., Esquivel, K., Rodríguez, D., Rodríguez, M., Duarte, R.: Internet of things (IoT), an
alternative for the care of water, vol. 40, pp. 2318–2330 (2018)
4. Calzada, J., Narváez, J., Aguilar, R., Velasco, H.: Agenda Técnica Agrícola Yucatán,
México (2015)
5. Narváez, M.: Chile habanero de Yucatán, único en el mundo. http://www.cienciamx.com/
index.php/ciencia/mundo-vivo/22848-chile-habanero-yucatan
6. Andrade Rodríguez, M., Villegas Torres, O.G., Juárez López, P., López Gómez, J.D.,
Martínez Fernández, E., Sotelo Nava, H.: Rendimiento y calidad del chile habanero
(Capsicum chinense Jacq.) por efecto del régimen nutrimental. Rev. Mex. Ciencias
Agrícolas 8, 1717 (2018). https://doi.org/10.29312/remexca.v8i8.699
7. Tun, J.: Chile habanero características y tecnología de producción. Secr. Agric. Ganad.
Desarro. Rural. Pesca y Aliment. 79 (2001)
8. Ruiz, N., Medina, F., Martínez, M.: El chile habanero: su origen y usos (2011)
9. Scott, C., Banister, J.: The dilemma of water management “regionalization” in México under
centralized resource allocation. Int. J. Water (2008). https://doi.org/10.1080/0790062070172
3083
10. Baum, B.A., Titlow, J., Vasanth, T., Welch, R.M.: Automated cloud classification of global
AVHRR data using a fuzzy logic approach (1997)
11. Jang, J.-S.R., Sun, C.-T., Mizutani, E.: Neuro-Fuzzy and Soft Computing Jang: A
Computational Approach to Learning and Machine Intelligence. Prentice-Hall Inc., Upper
Saddle River (1997)
12. Zadeh, L.: Fuzzy sets, Berkeley, California (2013). https://doi.org/10.1007/978-1-4614-
1800-9_73
13. Ying, H.: Fuzzy control and modeling : analytical foundations and applications. IEEE Press
Series in Biomedical Engineering (2000)
14. Ying, H.: The simplest fuzzy controllers using different inference methods are different
nonlinear proportional-integral controllers with variable gains. Automatica 29, 1579–1589
(1993). https://doi.org/10.1016/0005-1098(93)90025-O
15. Touati, F., Al-Hitmi, M., Benhmed, K., Tabish, R.: A fuzzy logic based irrigation system
enhanced with wireless data logging applied to the state of Qatar. Comput. Electron. Agric.
98, 233–241 (2013). https://doi.org/10.1016/j.compag.2013.08.018
16. Hussain, M.H., Min, T.W., Siraj, S.F., Rahim, S.R.A., Hashim, N., Sulaiman, M.H.: Fuzzy
logic controller for automation of greenhouse irrigation system (2011)
Fuzzy Logic Controller for Automation of an Autonomous Irrigation System 293
17. Suruthi, N., Saranya, R., Subashini, S., Shanthi, P., Umamakeswari, A.: Managing irrigation
in Indian agriculture using fuzzy logic – a decision support system. Int. J. Eng. Technol. 7,
321 (2018). https://doi.org/10.14419/ijet.v7i2.24.12075
18. Hilali, A., Alami, H., Rahali, A.: Control based on the temperature and moisture, using the
fuzzy logic. Int. J. Eng. Res. Appl. 07, 60–64 (2017). https://doi.org/10.9790/9622-
0705036064
19. Parmar, B., Chokhalia, J., Desarda, S.: Terrace garden monitoring system using wireless
sensor networks. Eng. Technol. Manag. 02, 51–54 (2019)
20. Izzuddin, T., Johari, M., Rashid, M., Jali, M.: Smart irrigation using fuzzy logic method.
ARPN J. Eng. Appl. Sci. 13, 517–522 (2018)
21. Munir, M., Bajwa, I., Cheema, S.: An intelligent and secure smart watering system using
fuzzy logic and blockchain. Comput. Electr. Eng. 77, 109–119 (2019). https://doi.org/10.
1016/J.COMPELECENG.2019.05.006
22. Boman, B., Smith, S., Tullos, B.: Control and Automation in Citrus Microirrigation Systems,
Florida (2008)
Internet of Things for Irrigation System
J. M. Guzmán-Toloza, D. F. Villafaña-Gamboa(&),
L. J. Peniche-Ruiz, R. A. Gómez-Buenfil,
J. K. Molina-Puc, and M. J. Rodríguez-Morayta
1 Introduction
In the last 100 years, the world’s population has tripled, while water withdrawals have
increased six fold, increasing the pressure on natural resources and specifically on
water resources worldwide [1].
Agriculture is one of the main pillars of the Mexican economy, 76% of all available
water is destined for agricultural use, while for public supply only 14.4%. In Mexico,
most irrigation systems operate manually, which in practice ends up being of little use,
promotes excessive use and even empirical use of water by not having a means to
verify whether the plants have had optimal watering. Among the main difficulties for
the technification of irrigation stand out: the high costs of conditioning the irrigation
infrastructure, low efficiency of conduction and distribution due to filtration problems,
excessive parceling of agricultural land, low efficiency of water application, due to
ignorance of efficient irrigation methods and the requirement of water per crop; the
weak organization of farmers and their low participation, so that innovation in irrigation
2 System Architecture
The first two cases verify irrigation history and check sensor status; they provide a
broad overview of the functionality of the system: the first one shows the user the exact
date and time when irrigation started while the second shows the status of each sensor;
this is useful to identify problems in the data collection or any malfunction. The last
298 J. M. Guzmán-Toloza et al.
function, check sensor values, allows real time visualization of the values obtained by
the installed sensors.
The V methodology was applied to develop the system [15], where tests for each
module such as the data server and the IoT server, are first carried out. Next, hardware
and software are integrated to verify the correct operation of both systems as a whole.
Finally, the functionality tests of the system in general are carried out.
Figure 3 shows the pins through which each sensor sends the recorded readings;
each pin is connected in series with a resistance of 220 X to limit the current and thus
avoid burning the Raspberry input terminal. The Raspberry Pi 3 module provides each
sensor with a +5 Vdc supply to operate. The GPIO Raspberry terminals of the irrigation
system electronic circuit are illustrated in more detail in the interconnection diagram in
Fig. 4.
Internet of Things for Irrigation System 299
Once the information server has collected current data from all sensors, these are
displayed in Raspberry terminal, so it is easy to visually verify that both the server and
the sensors are working correctly, thus there is communication between them without
interruptions or loss of information. Table 1 illustrates an example of the detected
values.
The information server sends the current data to the IoT server, which has the
necessary configurations to make the connection through the internet with the database
in real time.
300 J. M. Guzmán-Toloza et al.
In this project, the design and development of a controller for an automated irri-
gation system using fuzzy logic is carried out. The developed controller uses sensors
for soil moisture, humidity in the environment, brightness and temperature in the
environment to manage the activation of an irrigation pump; the moisture in the soil is
monitored every 10 min to determine if the humidity necessary to read the other
sensors is met and if each of the necessary factors is met, the irrigation system is
activated.
A conventional solenoid valve is used to supply the water, since the irrigation
system works only in two states: on and off. The solenoid valve is connected to a hose
that works in conjunction with sprinklers and it is activated by the fuzzy controller
when complying with the rules presented in Table 2. These rules were defined based on
interviews with experts in the care of habanero pepper when presented with different
scenarios with respect to the input variables. Soil moisture sensors are the ones that
mainly determine if the activation of the system is necessary, because when the soil is
completely humid, irrigation is not carried out regardless of the values of the other
sensors.
The mobile application allows representing the values obtained from each of the
sensors through graphs due to the integration of a Microsoft Office Excel® plugin,
which facilitates the analysis and understanding of the behavior of the data. Two graphs
are shown in Figs. 6 and 7, corresponding to average values obtained from the lumi-
nescence and humidity sensors in the soil.
302 J. M. Guzmán-Toloza et al.
The graphs allow seeing the behavior of the sensors and facilitate the decision
making to perform the irrigation.
3 Conclusions
There are already autonomous and automated irrigation systems in the market, but in
most cases their costs are too high for the common farmer, making them inaccessible.
This is why the proposed system considerably reduces the cost to make it available to
rural people, who in most cases lack economic resources. Its operation is as simple as
knowing the basics to manage applications and a smartphone or tablet with internet
access. Furthermore, if we consider that the situation of water scarcity will worsen this
kind of projects for the agricultural sector may contribute to solve this global problem
and keep the vital liquid in optimal usage.
The architecture that makes up this project allows monitoring of parameters in the
irrigation system, using a data server and an IoT server implemented in an integrated
Internet of Things for Irrigation System 303
system based on a distribution of Linux for Raspberry known as Raspbian. The data
server receives the information from the sensors every 2 s to avoid data overlapping
and information loss. Therefore, the proposed architecture can perform real-time
monitoring at low cost on the irrigation system. The IoT server sends the information of
the sensors to a database in real time, which in turn provides the updated information to
a mobile application implemented on a smartphone or tablet.
Implementing fuzzy logic in a controller for an irrigation system implies a great
impact and benefit for agriculture and people who are engaged in such activity, due to
the automation of tasks that allow people to reduce time and processes, based on
weather factors, using water in an efficient way avoiding waste and providing the plants
with sufficient quantity for optimum growth.
On a first stage, the irrigation system was developed and tested at laboratory level;
it will subsequently be installed in a greenhouse to grow habanero pepper plants. In the
future, more sections would be implemented within the mobile application, to have a
greater amount of information for the irrigation system, with the advantage to be able to
monitor real time data for better crops administration. It is also intended to integrate a
digital image processing system with a neural network, for the prediction of crop pests
and registration of plant growth, thus increasing the benefits of the intelligent irrigation
system.
References
1. Comisión Nacional del Agua: Estadísticas del Agua en México, pp. 45–50 (2018)
2. Fereres, E., García-Vila, M.: Irrigation management for efficient crop production. In: Savin,
R., Slafer, G. (eds.) Crop Science. ESSTS, pp. 345–360. Springer, New York (2019). https://
doi.org/10.1007/978-1-4939-8621-7_162. https://link.springer.com/referenceworkentry/10.
1007%2F978-1-4939-2493-6_162-3
3. IICA: Innovation and Water Management for Sustainable Development in Agriculture,
pp. 20–29 (2015). http://repositorio.iica.int/bitstream/11324/3035/2/BVE17068948i.pdf
4. Organización de las Naciones Unidas para la Alimentación y la Agricultura: El estado
mundial de la agricultura y la alimentación. Roma (2018)
5. Pastor, E.J.: Introducción al Internet de las Cosas, pp. 80–120 (2015)
6. Clúster ICT-Audiovisual de Madrid: Internet de las cosas: Objetos interconectados y
dispositivos inteligentes, pp. 68–75 (2016). https://actualidad.madridnetwork.org/
imgArticulos/Documentos/635294387380363206.pdf
7. Negrete, J.: Internet of things in Mexican agriculture; a technology to increase agricultural
productivity and reduce rural poverty, pp. 46–47 (2018)
8. Alper Akkas, M.: An IoT-based greenhouse monitoring system with Micaz motes, pp. 604–
608 (2017). https://www.sciencedirect.com/science/article/pii/S187705091731709X
9. World Water Vision Report: Visión Mundial del Agua, México (2017)
10. Karim, F., Karim, F.: Monitoring system using web of things in precision agriculture. In:
The 12th International Conference on Future Networks and Communications (2017). https://
reader.elsevier.com/reader/sd/pii/S1877050917312590
304 J. M. Guzmán-Toloza et al.
11. Shekhar, Y., Dagur, E., Mishra, S., Tom, R.J., Veeramanikandan, M., Sankaranarayanan, S.:
Intelligent IoT based automated irrigation system. Int. J. Appl. Eng. Res. 12(18), 7306–7320
(2017)
12. OBSoil-01: Octopus Soil Moisture Sensor Brick, p. 5 (2016)
13. D-Robotics: DHT11 Humidity & Temperature Sensor, Mouser Electronics. https://www.
mouser.com/ds/2/758/DHT11-Technical-DataSheet-Translated-Version-1143054.pdf.
Accessed 27 May 2019
14. Raspberry Pi Foundation: Raspberry Pi 3 Model B, Raspberry Pi Website, p. 137 (2016)
15. Perez, A., et al.: Una Metodología Para El Desarrollo De Hardware Y Software Embebidos
En Sistemas Críticos De Seguridad. Sist. Cibernética e Informática 3(2), 70–75 (2014)
Architecture Proposal for Low-Cost
Hydroponic IoT Based Systems
1 Introduction
Agriculture is the third economic activity in the country, about 22 million hectares are
dedicated to the agriculture, of which, 75.67% implement some level of mechanization,
and 64.92% of this is seasonal crop [1].
Of the agronomic population, approximately 35% of the farmers are in the micro, or
small business sector with ties in the domestic market and 50% produce for self-
consumption. Despite the existence of diverse imported solutions for the agro-food
field, these are focused on intensive producers of the industrial scale. They require high
initial investment and qualified personnel in the technology for its installation, oper-
ation, and maintenance. For this reason, its implementation has not been generalized
for producers in the micro and small enterprise sectors as well as self-consumption
orchards, who are more vulnerable to climate variability and change [2], generating a
technological and economic gap between small farmers versus intensive producers.
Also, with the growing urban expansion, areas dedicated to agriculture have been
reduced or displaced to forests and jungles, irremediably damaging the ecosystem by
decreasing the participation of the population in agriculture.
According to this, the design of low-cost solutions based on alternative systems that
potentiate the results of the agricultural campaign in small spaces, such as hydroponics,
along with precision agriculture and the Internet of Things (IoT), allow the modern-
ization of agriculture in the mentioned sectors without huge investments, obtaining
better quality products, this with the aim of generating economic growth in the region.
Fig. 1. Three-layer architecture for IoT proposed by the IEEE Working Group P2413.
The IoT solutions have many challenges to be facing [10–12]: Scalability, inter-
operability, mobility and adaptable solutions in order to provide evolved services. The
edge computing and Fog architectures helps to stablish the IoT architectural definition
and support its applications [13, 14] trough the implementation of different security
mechanism stablished for Internet protocols and smart objects standards.
The first stage for the IoT solutions is about obtain all the behavior information of
the study case, then analyze all the data collected in order to propose the algorithms for
the decision making and finally the final solution needs to be tested and corrected to
improve the results of the solution trough some advance control algorithms.
This article describes an architectural proposal for the development of a hydroponic
management system using IoT technologies. In the following section, the analysis and
design of the system are addressed, focusing on the development of two different
networks, the intra-greenhouse network and the extra-greenhouse network which, in
conjunction with sensors and actuators, monitors and control the greenhouse operation
variables.
The third section shows the development of communication networks as well as the
development of specific frames within the network, which allow optimizing the
implemented communication. Subsequently, the implementation and the obtained
results of the development of the system are briefly explained, and finally, we conclude
with possible future improvements of the system.
2 Design
Between crop campaigns, the distribution and extension of crops is variable, for this
reason, it´s important design a network topology of the solution with the follow
characteristics: modular in order to cover the needs of crop management at different
308 G. L. Evelyn et al.
scales; flexible and dynamic to allow variations in the geographical arrangement and
configuration of the nodes; low power to offer the autonomy of the nodes; and tolerant
to failures to ensure the operation of the network. According to this, the solution was
divided into two action zones:
Intra-Greenhouse. A wireless network is proposed to facilitate the installation and
communication of the modules without tools or extra material, the network is based on
a mesh topology to meet the requirements of a robust and dynamic communication
with enough coverage between the nodes arranged inside the greenhouse. Within this
network can define different types of nodes which focus on different roles within the
greenhouse, at least 3 types of nodes are proposed: Irrigation Node, Sensor Node and
Master Node. In the following sections will detail each of them.
Extra-Greenhouse. It’s required the transmission of data collected inside the green-
house to the server for further processing. The solution considers the communication of
the server to one or more greenhouses, so the network topology is defined as a star
topology in this communication stage. A wireless network is proposed to facilitate the
installation and communication of the modules without tools or extra material, the use
of a mesh topology could satisfy the requirements of a robust and dynamic commu-
nication with enough coverage among the nodes arranged inside the greenhouse.
Subsequently, the mobile terminal requires obtaining the data stored on the server.
To do so, it is necessary to consider the following cases:
Local network connection: The server will serve as a wireless access point to
download the information collected in the mobile terminal.
Anywhere connection through the Internet: To achieve this, it is necessary to
contract a service in the cloud for the exchange of data from the server to the mobile
terminal through the Internet and the necessary infrastructure exists.
Finally, we can define a hybrid network architecture integrated by one or more
“intra-greenhouse” networks which operate with a general “extra-greenhouse” network.
These networks can be of different protocols if there is a node that works as middleware
between both networks.
Fig. 2. The general architecture of the proposed network for the intelligent management system
for hydroponic crops, minimum case.
where we have multiple greenhouses, each one with its Intra-Greenhouse network
connected to the same Extra-Greenhouse network and with it to the same server.
This modularity can be applied in the same way at the greenhouse level, so that we
can have multiple greenhouses with different needs, in this case, it’s the Intra-
Greenhouse network that varies in the number of nodes, adjusting to the characteristics
of each place, this can be visualized in the Fig. 4.
310 G. L. Evelyn et al.
Fig. 4. Intra-Greenhouse network diagram: (a) minimum case, (b) expansion of the network
after adding sensor nodes.
domestic power line (110V in México) to energize said actuators and control them
through the change of state of optocoupled relays. The use of batteries for the irrigation
node is added, to provide redundancy in power for the perception phase in case of
disconnection in the electric supply. The processing unit is carried out (as with the
sensor and master node) with the AtMega328p microcontroller and the wireless
transceiver as the nRF24L01+ (see the Fig. 6).
2.3 Server
The server application is developed on the single-board computer Raspberry Pi 3
model B+ development card with Raspbian operating system (based on Debian -
GNU/Linux) due to its processor, memory and radio transceiver characteristics based
on the IEEE 802.11 standard already included in the card. The design of the server is
divided into two parts:
Communication. Two forms of communication with the server were established to
guarantee the link, through the local area network using the Ethernet port, taking
Architecture Proposal for Low-Cost Hydroponic IoT Based Systems 313
Fig. 8. Communication lines of the master nodes and the mobile application with the server.
Functionality. For the development of the web server, we use Django which is a high-
level application programming environment based on Python, Django allows us to
create web servers in a friendly way and create the REST services necessary for the
solution. The web application is constituted by Models, Uniform Resource Locators
(URLs), and Views. The information is stored and managed by the database
management system MariaDB, which offers the services of storage and control of the
tables. The relational model is selected due to the need to link the data of different
tables to establish the information of the crops, their belonging, and configuration (see
Fig. 9).
Fig. 10. State machine of the views (activities) of the mobile application.
Fig. 11. Payload of the configuration message for irrigation node and sensor.
Table 1. Parameters included within the payload field of the nRF24L01+ frame.
Node type Parameter VMIN VMAX Precision Steps Data size
Sensor Temp 2 °C 35 °C 0.2 °C 166 1 byte
Hum 50% 85% 0.25% 141 1 byte
Ilum 0 60 Klx 10 lx 6001 2 bytes
Irrigation TempSol 10 °C 30 °C 0.1 °C 201 1 byte
pH 5 7.5 0.1 26 1 byte
make the memory space per parameter more efficient, the Table 1 shows data size of
the parameters monitored for hydroponic crops: Temperature (Temp), Humidity
(Hum), Light intensity (Ilum), temperature of the Nutrient solution (TempSol) and pH
level (Table 1).
For the communication of the message configuration by the master node to the
sensor and irrigation nodes, it is necessary to add to the payload (Fig. 11) the maximum
and minimum values of the parameters, their status (if enabled), the period in which the
sensing sample will be taken. Also, the data about start and end time in which the
configuration is valid, and in the case of the irrigation node, information referring to the
irrigation activity (irrigation periods and duration), so 21 bytes of data was defined for
data configuration message to be transmitted in this stage of design (see Fig. 12).
3 Development
This section briefly explains how the communication between the intra-greenhouse
network and the extra-greenhouse network is developed, as well as the structure of the
JSON (JavaScript Object Notation) messages that are used for the communication of
the master node with the server.
316 G. L. Evelyn et al.
Fig. 12. Data structure of the JSONs sent from the master node to the server
After initializing the nRF24L01+ transceiver and the sensors installed in the nodes,
the default range values are read from the internal MCU EEPROM; those values act as
extreme values for the monitoring of the environment (Temperature, Humidity, Illu-
mination, pH, Temperature of the nutrient solution). Likewise, the identification
information of the node (node type and serial number) and the configuration of the
control parameters of the system (sampling periods, times in which the configuration is
valid and node status) are read.
The configuration and management of the mesh network begin as the master node
dynamically assigns an identifier to the slave nodes (sensor node and irrigation node),
promoting the dynamic registration (commissioning) of them and the joining and
leaving of nodes to the network (see Fig. 13), thus allowing relocation of the nodes at
any time. The connection between nodes is updated periodically to maintain the link
between the nodes.
Finally, the operation of the slave nodes complies with the following steps: (see
Fig. 14).
Fig. 13. Flowchart of the initialization of the nodes for the creation of the BLE network.
318 G. L. Evelyn et al.
Fig. 14. Flowchart of functionality of the Sensors, Irrigation and master nodes
The development and implementation of the system were carried out between the
facilities of the ESIME Zacatenco and a micro greenhouse in the town of Santa Isabel
Tola, the network of nodes was developed with the proposed sensors, a master node, a
server, and a prototype of the mobile application were developed (Fig. 15).
Fig. 15. Basic electronic diagram of Sensor Node and Irrigation Node
The user utilizes the mobile application to enter the system to monitoring and
configure the state of crop actuators in a friendly way (Fig. 17).
320 G. L. Evelyn et al.
Then select a crop to visualize, and the application displays the information of the
crop, its location, its assigned greenhouse, and the last recorded reading of the sensors
in the location (see Fig. 18).
Fig. 18. List of registered crops and display of the details of a crop
Architecture Proposal for Low-Cost Hydroponic IoT Based Systems 321
This system is developed focusing on low-cost technologies that allow its implemen-
tation by independent farmers, self-consumption and micro and small businesses,
which in most cases cannot acquire commercial systems due to its high cost. However,
we have observed that the flexibility and modularity of the system allows it to be
installed in many more places, for example, in urban and roof gardens or in indoor
hydroponic systems, which can have a significant impact on the reduction of land used
for traditional culture.
The above allows us to conclude that with minimal modifications this architecture
could be implemented in massive hydroponic farms, improving the monitoring systems
of these and with it the quality of the production. Finally, these types of modular IoT
architectures can revolutionize the industry, lowering the cost of infrastructure and,
soon, increasing the quality of life of the users.
The first stage of design was completed for the Intelligent system management for
hydroponic crops, the intends of the solution pretend to be an easy tool for farmers and
urban agricultures to improve their crops through the IoT technologies without high
investments.
Currently, we are working on the implementation in a real hydroponic greenhouse
to verify the correct functioning of the system. Based on the results of this imple-
mentation, the aim is to develop a machine-learning based algorithm that allows
adjusting the monitoring ranges automatically, to optimize crop quality and yield.
It is also proposed to implement an LTE-M modem that allows monitoring
remotely, as well as the sending of alerts by SMS, as an additional way of informing
the user of possible ongoing problems.
Finally, we are working on a second version of the system focused on medium-
scale crops, based on WSN, with greater coverage and flexibility, including a backup
server in the cloud, so that this new system can be managed anywhere with an Internet
connection.
References
1. Servicio de Información Agroalimentaria y Pesquera (SIAP). Uso de tecnología y servicios
en el campo Cuadros tabulares (2016)
2. World Bank, CIAT, CATIE. Climate-smart agriculture in Mexico. CSA country profiles for
Latin America Series. World Bank Gr (2014)
3. Oasis, S.: Manual de hidroponia, p. 32 (2002)
4. Patokar, A., Gohokar, V.: Precision agriculture system design using wireless sensor network,
pp. 169–177 (2017). https://doi.org/10.1007/978-981-10-5508-9_16
5. Verdouw, C.: Internet of Things in agriculture. CAB Rev. Perspect. Agric. Vet. Sci. Nutr.
Nat. Resour. 11 (2016). https://doi.org/10.1079/pavsnnr201611035
6. Azcón-Bieto, J., Talón, M.: Fundamentos de fisiología vegetal (2003)
7. Araceli, M.: Manual de hidroponia, Primera Ed. Instituto de Biología, Distrito Federal
(2014)
322 G. L. Evelyn et al.
8. Kiani, F., Seyyedabbasi, A.: Wireless sensor network and internet of things in precision
agriculture. Int. J. Adv. Comput. Sci. Appl. 9, 99–103 (2018). https://doi.org/10.14569/
ijacsa.2018.090614
9. Minerva, R., Biru, A., Rotondi, D.: Towards a definition of the Internet of Things (IoT).
IEEE Internet Things 86 (2015). https://doi.org/10.5120/19787-1571
10. Lin, J., Yu, W., Zhang, N., Yang, X., Zhang, H., Zhao, W.: A survey on Internet of Things:
architecture. Enabling Technol. Secur. Priv. Appl. 4, 1125–1142 (2017)
11. Chiang, M., Zhang, T.: Fog and IoT: an overview of research opportunities. IEEE Internet
Things J. 3, 854–864 (2016). https://doi.org/10.1109/JIOT.2016.2584538
12. Yang, Y., Wu, L., Yin, G., Li, L., Zhao, H.: A survey on security and privacy issues in
Internet-of-Things. IEEE Internet of Things J. 4, 1250–1258 (2017)
13. Shi, W., Cao, J., Member, S., Zhang, Q., Member, S.: Edge computing: vision and
challenges. IEEE Internet Things J. 3, 637–646 (2016). https://doi.org/10.1109/JIOT.2016.
2579198
14. Razzaque, M.A., Milojevic-Jevric, M., Palade, A., Cla, S.: Middleware for internet of things:
a survey. IEEE Internet Things J. 3, 70–95 (2016). https://doi.org/10.1109/JIOT.2015.
2498900
Informatics Security
Learning and Automatization of Searching
in Digital Media, Based on Concurrency,
Weighing and Decay
Abstract. The search for information of real interest for a particular topic has
always been a challenge, especially when the source of this information is on the
Internet, where there are many digital media, platforms, databases and from
which you can explore data. Large companies such as Google or Yahoo help us
find these sources of information through their search algorithms, but they are
not very effective when filtering a particular information, so it is necessary to
formulate a model that responds to this need, in the present The document
proposes a model based on 3 components, using information filtering tools,
word valorization techniques based on their concurrences, weights, anti-words
and a proposal to apply a word decay log within the information, in order to
maintain a “score” and avoid false weights (triggered weights) of the
information.
1 Introduction
In this way, it is possible to determine the variables that define the obtainment
process. These are found on all digital media which fulfils the condition of delivering
information in the Feed RSS format (Really Simple Syndication) [4].
• Obtainment Variables: RSS Feed Structure: From this RSS structure taken from the
FeedParser library, we will base the information searching on collecting the data
now shown. These will change their content as the information source feeds update,
but the structure always remains unchangeable.
– feed.tittle
– feed.url
– feed.content
– feed.tags
• Information Digital Media: The Information Digital Media (news websites, blogs,
etc) will deliver the information requested by the model. These could be classified
by topics, and the collecting condition will be given by the information delivery
format (Feed RSS).
Learning and Automatization of Searching in Digital Media 327
• Feed Update Time: The different information digital media update their feeds when
modifying the information. For example, every time a digital newspaper publishes a
news, the feed changes and generates new information captured by the model.
• Collecting Cycle Time: This refers to the time the model will last to collect all the
digital information media before starting again.
• Total Collecting Time: The total time employed to collect the information
(Table 1).
Concurrency: The concurrent value will be given by the time a key word shows on
the information (Table 2).
3 Component Metrics
The sample taking is conducted to register the metrics of each component, in which the
values of the identified variables are obtained. This is conducted by a determined
hardware and software, which are mentioned as following:
Hardware:
• Intel I5 8th Generation
• 8 GB Ram
• 256 GB SSD
The software. Libraries employed:
• Linux Kubuntu 16
• Python 3.6
• Framework Django 2.1
• Virtualenv 17.0
• PostgreSQL 11.2
• NLTK library, Text Blob
• FeedParser 5.2.0
Each collected fee changes from the news media (the news title, the body and the
related tags). The update average time of these 100 media is of 15 min. This is variable
because it changes depending on the news’ contingency. At the moment of this study,
the contingency developed normally, without existing a high number of interesting
330 D. Cordero Vidal and C. Barria Huidobro
news in the news media. The model lasts approximately 11 min to track all the reg-
istered websites. From these metrics, 17.000 feeds were collected and these will sustain
the information classification component.
Table 5. Filter of key words over 17.000 news from a 100 information media
Number of key words Number of candidate news Percentage of total news
01 4.320 25,41%
02 338 1,98%
03 44 0,25%
4 Model Proposal
From the components described previously, we propose the following model based on
three components (Fig. 2), and a flow chart which explains the model’s addressing
(Fig. 3).
5 Continuous Learning
Once the system has been sustained with the information of interest, it is possible to
automatically determine new key words, example: words not considered by the
expert’s judgement at the beginning of the procedure. With the library NLTK (Natural
Language Toolkit), all corpuses available can be vectorized, stopwords erased and
significant words are selected. Likewise, it can be determined which words are more
concurrent so to propose them to the dictionary indexation.
Consideration: It is the value or importance a key word has within a dictionary, and it
relates to the topic of interest. In this way, the words are prioritized according to their
importance in the topic. The same logic is employed by the procedure based on
representation of Bag-of-Words’ text [10], and used by the NLTK library to assign the
relevance meaningless or too repetitive words could have (Fig. 4).
The total weighing dictionary has to be equal to 100% in the amount of words
registered to define the correct prioritization. The value of each word depends on topic,
as the same word in different topics could have different weighing values (importance).
The weighing value could be assigned by the expert’s judgement.
Decay Logarithm: If there are repeated words within the body information, a
logarithm of decay will be applied [11]. The value for this logarithm will be obtained
from the number of manifestations shown in the vector. This procedure is conducted
because a single key word can alter the acceptance criteria, as many concurrencies can
show a high weighing followed by a value decay within the total sum of words found.
If the vector of the frequency (repetitions) of the dictionary words f^(t) in a given
text t, then to calculate the score of the text it would be enough to calculate the next
sum:
where f_i^(t) is the number of times the word number i in the dictionary appears in the
text t and d_i are functions of the form:
An example will be Twitter [12], which provides an API with which we can look
for Tweets related to a topic. Another will be the tools of Google, particularly Google
alerts, which provide information available on the website through sides that have not
been registered in the system (digital information media).
The integration of a system that connects to any API and sends real-time notifi-
cations at the moment of finding a piece of information is crucial. Libraries such as
Telebot (Telegram) allows to manage these notifications directly from the source code.
It is necessary to create only a BOT from Telegram [13], to include the application
token key directly to the source code in order to send the messages with the results.
References
1. Dua, S., Du, X.: Data Mining and Machine Learning in Cybersecurity. Auerbach
Publications, New York (2016). 9780429063756
2. Fawaz, S., AbuZeina, D.: Toward an enhanced Arabic text classification using cosine
similarity and Latent Semantic Indexing. J. King Saud Univ. – Comput. Inf. Sci. 29(2), 189–
195 (2017)
3. Feedparser. https://pythonhosted.org/feedparser/. Accessed 06 May 2019
4. Bharathi, S., Geetha, A., Sathiynarayanan, R.: Sentiment analysis of twitter and RSS news
feeds and its impact on stock market prediction. Int. J. Intell. Eng. Syst. (2017). https://doi.
org/10.22266/ijies2017.1231.08
5. Saif, H., Fernández, M., Yulan, H., Harith, A.: On stopwords, filtering and data sparsity for
sentiment analysis of Twitter. In: Proceedings of the Ninth International Conference on
Language Resources and Evaluation, LREC 2014, pp. 810–817 (2014)
6. Prat, M.: Posicionamiento web: estrategias de SEO: Google y otros buscadores. s.l.: Eni
(2016). 2409004903
7. Cabero, J., Llorente, M.: The expert’s judgment application as a technic evaluate information
and communication technology (ict), Revista de Tecnología de Información y Comunicación
en Educación, vol. 7, no. 2 (2013)
8. Díaz, I., Suárez, S., Sidorov, G.: Creation and evaluation of a dictionary tagged with
emotions and weighted for Spanish, Onomázein: Revista de lingüística, filología y
traducción de la Pontificia Universidad Católica de Chile, no. 29, pp. 31–46 (2014). ISSN-e
0717-1285
9. Perkins, J.: Python 3 Text Processing with Nltk 3 Cookbook (2014). ISBN-10: 1782167854
10. Hasan, K.S., Ng, V.: Stance classification of ideological debates: data, models, features, and
constraints. In: Proceedings of the Sixth International Joint Conference on Natural Language
Processing, pp. 1693–1701 (2015)
11. Kumar, S., Morstatter, F., Liu, H.: Twitter Data Analytics. Springer, Heidelberg (2013).
https://doi.org/10.1007/978-1-4614-9372-3. ISBN 1461493722
12. Setiaji, H., Paputungan, I.: Design of telegram bots for campus information sharing. In: IOP
Conference Series: Materials Science and Engineering, vol. 325, p. 012005 (2018). https://
doi.org/10.1088/1757-899x/325/1/012005
13. Salas Alvarez, D.: Revolución 4.0, RII, vol. 4, no. 2 (2016)
Security Hierarchy for the NDN Network:
Description and Experimentation
1 Introduction
The Internet is a worldwide data network whose main objective is to allow pairs
of computers to exchange information. In order to identify each computer, it
becomes necessary to employ a unique number, the IP address. However, most
users do not need to know their own devices’ address, nor the procedures required
to exchange information in this network. These procedures are relegated to dif-
ferent protocols, like TCP and UDP. Since what it is important to the Internet’s
users is to obtain information and services, there are currently proposals to
implement new network architectures that are oriented to obtain, transmit and
deliver information, instead of establishing a connection between two computers.
This shift of the main architectural objective of the Internet has been named
Information-Centric Networking or ICN [1,2].
The main objective of ICNs is to implement a network infrastructure that
is better suited on how users currently employ the Internet, which are more
interested in accessing information, rather than its physical location. Special
attention is given to content distribution, security and mobility support. There
have been different proposals to implement ICNs. Examples of these are Data-
Oriented Network Architecture (DONA) [3], Publish-Subscribe Internet Tech-
nology (PURSUIT) [4], Scalable and Adaptive Internet Solutions (SAIL) [5] and
Named Data Networking (NDN) [6].
Named Data Networking or NDN is a type of ICN that is currently being
developed and implemented by the NDN Project. This project is a team effort
c Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 335–344, 2019.
https://doi.org/10.1007/978-3-030-33229-7_28
336 P. D. Arjona-Villicaña and A. G. Silva-Trujillo
Fig. 1. The main functional elements of the current Internet and NDN.
by many universities and research laboratories around the world, including the
Universidad Autónoma de San Luis Potosı́ (UASLP) [7]. Currently, participants
to this project are trying to implement the main functionalities of an NDN node,
for example: hierarchical name structure, security methods and procedures, rout-
ing, forwarding and in-network storage [6]. Others are working on applications
that need to employ some of the functionalities described before. The main idea
behind these developments is to prove NDN functionality by testing it in a setting
that is as close to a real network as possible. Therefore, the NDN project is also
implementing a global network to test the protocols, services and applications
under development [8]. This network includes a node at UASLP.
Another of the main objectives of NDN is to maintain a structure that is as
simple as the one currently implemented for the Internet. This has been called the
thin waist model (Fig. 1). For today’s Internet, the simplest (thinnest) functional
element is the IP packet; for the case of NDN, is the Content or the name of the
information, which is sent through two different types of packets, Interest and
Data.
Addresses in an NDN network are the name of the information or service
that the user requests [6]. These addresses follow a hierarchy which enables
names that are more specific to have priority over more generic ones. For exam-
ple, a name like /uaslp/ingenieria/computacion has precedence over /uaslp or
/uaslp/ingenieria. This is similar to how current routing protocols work, where
a network using a /24 mask (a set of specific IP addresses) has priority over
another using a /20 mask (a less specific set of addresses). NDN employs only
two types of packets: Interest and Data. Interest packets are used to request
information and Data packets are used to answer each Interest packet. Neither
of these packets is associated with the computer that originally requested the
information or the one that generated the Data packet.
Security Hierarchy for the NDN Network 337
2 NDN Security
This section first describes the NDN security hierarchy (Subsect. 2.1). Followed
by a description of the software needed for implementing the experiments (Sub-
sect. 2.2). Finally, two experiments are described in Subsects. 2.3 and 2.4.
NDN follows a hierarchical security structure (see Fig. 2). A root is at the top of
this hierarchy and is the entity in charge of certifying each site in the network.
Then, each site certifies one or more operators which, in turn, certify each router
in the network. Each one of these different entities must be defined and certified
338 P. D. Arjona-Villicaña and A. G. Silva-Trujillo
in an NDN network. For networks that have few nodes, it is possible that one
router simultaneously assumes the function of more than one of these entities.
It is important to consider that this security hierarchy needs to be defined and
configured before any other protocol or application can exchange messages. This
includes even routing protocols, like NLSR.
Figure 2 shows an example of a security structure for and NDN network with
two sites. Site 1 employs two operators: operator 1 is responsible for router 1
and router 2, while operator 2 is responsible for router 3 and router 4. Site 2 has
a simpler configuration since it only has one operator (operator 3) which only
handles router 5. This hierarchy is independent of how the routers are connected
between themselves and, as it was said before, more than one entity could be
defined in a single router or node.
In order to implement this hierarchy, it is necessary to exchange keys and
certificates using the following procedure: The dependent entity (the one that
is below in the hierarchy) generates a key which is then copied to the parent
entity (the one that is above). The parent entity employs the key as an input to
generate a certificate, which is then copied back to the dependent entity. Finally,
the dependent entity installs the certificate. This procedure is shown in Fig. 3.
The only exception to this procedure is the root, which is the only entity allowed
to self-certify its key.
Before performing the experiments described in Sects. 2.3 and 2.4, it is necessary
to install ndn-cxx and NFD in each node in the test network. The ndn-cxx library
is a set of programs developed in C++ that implement different NDN support
applications. The NDN website [13] includes information and instructions on
how to install this application’s library. NFD is an application that implements
an NDN forwarder node that supports the NDN protocol and is able to receive
and send Interest and Data packets. The NDN site [14] includes instructions on
Security Hierarchy for the NDN Network 339
2.3 Experiment 1
This experiment shows how to configure the security hierarchy for the sim-
plest NDN network: two routers communicating between themselves (see Fig. 4).
This network has two routers, R1 and R2, which connect to each other using a
crossover Ethernet cable. R1 is configured to be the root, site, operator 1 and
R1, while the other computer is only configured to be R2. The commands needed
to configure R1 are included below.
Since R1 is already configured as the root, site and operator 1, R2 only needs
to be configured as a router whose parent entity is the same operator (operator
1). The commands included below need to be executed in R2 and R1 to perform
this configuration.
3. The certificate is copied back to R2 and then it is installed using the following
command:
ndnsec-cert-install -f R2.cert
Security Hierarchy for the NDN Network 341
Fig. 5. Extract of NLSR log showing successful key registration and message exchange.
2.4 Experiment 2
To configure the security hierarchy for R1, it is only necessary to repeat the
same steps that were shown at Sect. 2.3 for R1.
The commands included below may be used to configure security for R2.
1. Operator 2 generates its own key:
ndnsec-key-gen /ndn/edu/uaslp/%C1.Operator/op2 > op2.key
2. The key is copied to the site (R1) and it is then certified using the following
command:
ndnsec-cert-gen -s /ndn/edu/uaslp op2.key > op2.cert
3. The certificate is copied back to R2 and it is then installed using the following
command:
ndnsec-cert-install -f op2.cert
4. The following commands configure R2 as a router:
ndnsec-key-gen /ndn/edu/uaslp/%C1.Router/R2 > R2.key
ndnsec-cert-gen -s /ndn/edu/uaslp/%C1.Operator/op2 R2.key
> R2.cert
ndnsec-cert-install -f R2.cert
To configure R3, R4 and R5 it is necessary to repeat the same steps that
were used at Sect. 2.3 for R2. However, these instructions need to be adjusted
to consider that R3’s parent is operator 1, while the parent for R4 and R5 is
operator 2.
After performing the security configuration for the five nodes in this network,
NLSR was executed again and it was verified that all the computers exchanged
messages and were able to generate a routing table.
Security Hierarchy for the NDN Network 343
3 Conclusions
ICNs are a proposal to redefine the Internet from a network that allows to estab-
lish communication between two computers, to an infrastructure whose main
objective is to obtain information from the network. This, in turn provides the
opportunity to examine new ideas and to reflect on the fundamental principles
that allowed to construct today’s network.
For the specific case of NDN networks, security mechanisms are required to
guarantee that Interest and Data packets are not modified as they travel through
the network. This paper has shown how to properly configure this security using
two different test networks. This correct configuration allows the implementation
of an Internet which should be safer by definition.
Besides the topic of security, there is a large number of protocols and appli-
cations under development and experimentation for NDN. These developments
represent an opportunity to contribute to a project which may have a signifi-
cant impact in the design of new architectures for the Internet. However, engi-
neers interested in contributing to the NDN project, first need to understand
its security hierarchy. The experiments included in this article are a first step to
understand this hierarchy and its main concepts.
The authors of this article are trying to propose new routing strategies for
NDN and this article shows the path that may be followed in the next stages of
research.
Acknowledgments. The authors would like to thank Lan Wang and Ashlesh
Gawande form the University of Memphis which helped to understand and clarify
many of the concepts exposed in this article.
References
1. Ahlgren, B., Dannewitzh, C., Imbrenda, C., Kutscher, D., Ohlman, B.: A survey
of information-centric networking. IEEE Commun. Mag. 50(7), 26–36 (2012)
2. Xylomenos, G., et al.: A survey of information-centric networking research. IEEE
Commun. Surv. Tutor. 16(2) (2014)
3. Koponen, T., et al.: A data-oriented (and beyond) network architecture. In: ACM
SIGCOMM, pp. 181–192 (2007)
4. FP7 PURSUIT project. http://www.fp7-pursuit.eu/
5. FP7 SAIL project. http://www.sail-project.eu/
6. Zhang, L., et al.: Named data networking. ACM SIGCOMM Comput. Commun.
Rev. 44(3), 66–73 (2014)
7. NDN Testbed Status. http://ndndemo.arl.wustl.edu/
8. NDN Traffic Map. http://ndnmap.arl.wustl.edu/
9. Bellovin, S., Schiller, J., Kaufman, C.: Security Mechanisms for the Internet, RFC
3631 (2003)
10. Zhang, Z., et al.: An overview of security support in named data networking, NDN,
Technical report NDN-0057 (2018)
11. Dijkstra, E.: A note on two problems in connection with graphs. Numerische Math-
ematik 1(1), 269–271 (1959)
344 P. D. Arjona-Villicaña and A. G. Silva-Trujillo
12. Lehman, V., Mahmudul Hoque, A., Yu, Y., Wang, L., Zhang, B., Zhang, L.: A
secure link state routing protocol for NDN, NDN, Technical report NDN-0037
(2016)
13. ndn-cxx: NDN C++ library with eXperimental eXtensions. http://named-data.
net/doc/ndn-cxx/current/
14. NFD: NDN Forwarding Daemon. http://named-data.net/doc/NFD/current/
15. NDN Project Team, NFD Developer’s Guide, Technical report NDN-0021 (2016)
16. NLSR, Named Data Link State Routing. http://named-data.net/doc/NLSR/
current/
17. Arjona-Villicaña, P., Renterı́a-Vidales, O., Gawande, A.: A beginner’s guide
to installing and testing NLSR in Fedora. http://named-data.net/doc/NLSR/
current/beginners-guide.html
Methodology for Malware Scripting Analysis
in Controlled Environments Based on Open
Source Tools
1 Introduction
cause of concern for cybersecurity managers who must generate computer security
awareness instances or culture.
Many of the attacks that occur today are based on malicious software and the
different variations that may arise from a single threat, it is for this reason that the need
for malware analysis is born, which in definition could be considered as the art of
dissecting malicious software, to understand how it works, how to identify it and thus
be able to defeat it.
Malicious codes or scripts will continue to be one of the main threats to the modern
world, becoming an endless battle between those who write new scripting malware and
those who seek to detect their execution in time and avoid all types of fraud through
cyberspace. The exploitation of systems through the infection of scripting malware also
propagated by phishing, are the perfect mix to continue deceiving the end user and to
penetrate even the most sensitive information we store.
Currently there are different methods to perform different activities framed within the
field of cybersecurity and cyber defense, one of these disciplines is malware analysis.
As a way to help researchers and analysts on this matter, the present analysis of
malicious software is focused on malware scripting, that is very simple code sequences
that automate malicious tasks in real time and that do not need to be compiled as is the
case with files executables (.EXE). These can infect, download, redirect or slow down a
computer, in addition to not being discovered developers obfuscate the code to make
them undetectable and thus not be discovered by antivirus engines.
Some languages to perform this type of malware can be used different program-
ming languages, such as Visual Basic Script (VBS), Java Script (JS) or Python (PY).To
carry out this type of methodology, the following 5 stages are proposed:
This type of analysis happens when it is necessary to calculate the different hashes
that the sample can have, such as its MD5 hash, SHA1, among others, it is also possible
to rely on integrated development environments (IDE), or it can even be studied
quickly with any text editor, since they do not need to be compiled for execution.
348 D. Muñoz et al.
After the background or data obtained, it is important to look for information in open
sources, in order to know if this threat has already been studied by other researchers.
3 Applied Example
Once the MD5 hash is obtained, we proceed to extract the metadata contained in
the sample, this may vary with each case (Fig. 3):
When extracting the metadata and realizing that the type of file is unknown, we use
the “File” utility that comes by default in Kali Linux to obtain more data, as shown
below (Fig. 4):
350 D. Muñoz et al.
This gives us an indication that the sample has ASCII characters and that it is also
composed of large lines of code.
If we go to the end of the result we can observe the following (Fig. 7):
We already have some data regarding how the obfuscation of this script works,
which by the way uses WScript, an environment developed by Windows for the
execution of scripts in its operating system.
Now, with the background information obtained, it is possible to perform searches
in open sources, using tools such as “total virus” to analyze, for example, some of the
hashes of the sample, as it is possible to observe in the following image:
A good activity now is to review each of the processes generated to identify their
properties and where they are stored.
3.4.3 Comparison
To compare what changes exist in a machine once infected, a tool called “Regshot” can
be used which makes a comparison of the before and after a sample is executed, as can
be seen in Fig. 10, a “shot” to a clean machine, then the sample is executed, then a
second shot” is taken and finally the results must be compared.
This will give us a text file with all the changes that arose in the equipment once the
sample was executed, being able to show how many values were deleted, modified, as
well as folders or keys within the system.
4 Model Proposal
Based on the above, the following malware scripting analysis methodology is proposed
(Fig. 11):
This methodology aims to facilitate the analysis of various types of scripts which may
contain some type of malicious use, considering that it is a quick study and must be
executed before conducting an in-depth analysis, in order to save time in making
decisions regarding an infection that could affect the normal development of activities
both in an organization and in the personal life of an analyst.
In addition, it is important to mention that this methodology is the basis for a deeper
research, leaving as future work to adapt new tools and lessons learned from other
disciplines [10], such as computer forensics, in terms of custody and recognition of
evidence. Thus, this paper is intended as a first approach to the issue at hand, in order to
provide the foundations for a thorough methodology proposal, tested against real-life
use cases.
References
1. Gutmann, P.: The commercial malware industry. In: DEFCON Conference (2007)
2. Baltimore, B.: Baltimore’s out-of-date and underfunded IT system was ripe for ransomware
attack (2019)
3. Baltimorebrew classification homepage. https://www.baltimorebrew.com/2019/05/21/
baltimores-out-of-date-and-underfunded-it-system-was-ripe-for-a-ransomware-attack/.
Accessed 21 May 2019
4. Gandotra, E., et al.: Malware analysis and classification: a survey. J. Inf. Secur. 5, 56–64
(2014). https://doi.org/10.4236/jis.2014.52006
5. Mieres, J.: Ataques informáticos. Debilidades de seguridad comúnmente explotadas.
Recuperado (2009). http://proton.ucting.udg.mx/tutorial/hackers/hacking.pdf
6. Mitnick, K.: The Art of Invisibility: The World’s Most Famous Hacker Teaches You How to
be Safe in the Age of Big Brother and Big Data. Hachette, UK (2017)
7. Gómez, J., Villar, E.: Introducción a la virtualización (2018)
8. Fedorenko, E., Ivanova, A., Dhamala, R., Bers, M.U. The Language of Programming: A
Cognitive Perspective. Trends in cognitive sciences (2019)
9. Sikorski, M., Honig, A.: Practical Malware Analysis: The Hands-On Guide to Dissecting
Malicious Software (2012)
10. Pandey, S., Mehtre, B.: Performance of malware detection tools: a comparison. In: 2014
IEEE International Conference on Advanced Communications, Control and Computing
Technologies, pp. 1811–1817. IEEE (2012)
11. Uppal, D., Mehra, V., Verma, V.: Basic survey on malware analysis, tools and techniques.
Int. J. Comput. Sci. Appl. (IJCSA) 4(1), 103 (2014)
Validation of ICS Vulnerability Related
to TCP/IP Protocol Implementation
in Allen-Bradley Compact Logix PLC
Controller
Abstract. Industrial Control Systems (ICS) research and testing process was
implemented to validate the existence of a well known security vulnerability in a
Rockwell Automation Allen-Bradley Compact Logix PLC controller. The study
was conducted considering a public advisory from the manufacturer, which
includes a large list of families of affected products by the vulnerability. The
established hypothesis of the study considered the existence of the vulnerability
in a specific available PLC model, included by Rockwell Automation manu-
facturer in the list of affected products. An exploit was developed and multiple
testing was performed to trigger the vulnerability.
Testing methodology and results indicates there is sufficient evidence to
establish that Rockwell Automation Allen-Bradley Compact Logix 5370 L2
controllers, are not affected by the same type of Improper Input Validation
vulnerability, than the Compact Logix 5370 L3 controllers, as it was stated by
the manufacturer in a public advisory.
1 Introduction
Industrial Control Systems (ICS) play a very important role in world economy and
current way of living. ICS became highly automated and network connected, intro-
ducing benefits but also new security vulnerabilities. This is particularly important in
Critical Infrastructure ICS such as energy, water, transportation, etc. [12].
Automation components of ICS include sensors, actuators, PLCs and SCADA
systems, among others components [8]. New ICS vulnerabilities are discovered every
© Springer Nature Switzerland AG 2019
M. F. Mata-Rivera et al. (Eds.): WITCOM 2019, CCIS 1053, pp. 355–364, 2019.
https://doi.org/10.1007/978-3-030-33229-7_30
356 J. Pavesi et al.
year. In several countries, regulation forces manufacturers to recognize and inform ICS
vulnerabilities in public advisories, which usually include mitigation procedures and
recommendations to their customers, such as software or firmware upgrades, ICS
network isolation, traffic block, renew equipment, etc.
Such mitigations actions may have an important cost for the companies and are also
time consuming. Most of the companies rely on the information provided by ICS world
class manufacturers, such as Rockwell Automation, Siemens, etc.
When an ICS vulnerability become public, including a list of ICS affected products,
companies currently using those products feels they are at risk of being attacked. This
is particularly important in the case of critical infrastructure companies, whose oper-
ation depends on ICS systems that are supposed to be vulnerable [9].
Those companies don’t hesitate to follow the risk mitigation recommendations of
their provider/manufacturer, assuming the involved costs. But, what if the information
provided by manufacturers, related to ICS vulnerabilities, is not enough accurate?
What if ICS manufacturers decide to include families of products supposedly
affected by a vulnerability, in their public advisories, to prevent a possible damage,
without having complete demonstrated confirmation of the vulnerability, in some of the
products?
Looking for an answer to all these questions, this study focuses on technical val-
idation of PLC vulnerability in Rockwell Automation Allen-Bradley Compact
Logix PLC affected products, declared in NCICC Advisory ICSA-18-172-02 [1].
That specific PLC was selected for this research, because the research team has full
access to a Lab with 4 of those PLCs and also has access to the PLC configuration and
communication licensed software, which are needed for the purpose of the study.
In June 2018, researcher Alexey Perepechko from Applied Risk company based in the
Netherlands, discovered a PLC vulnerability and sent an Advisory Report [2] to
Rockwell Automation manufacturer. The report include the following main
information:
“A vulnerability in the Rockwell Automation Allen-Bradley CompactLogix 5370
Controller 1769-L30ERMS could allow an unauthenticated, remote threat actor to
reboot the device and switch the device to the “Major Non-Recoverable Fault” mode,
resulting in a Denial of Service (“DoS”) condition” [2].
“Vulnerability Details. The vulnerability is due to incorrect processing of TCP ACK
packet additional options by the listener at Ethernet/IP TCP port (default 44818). An
incorrect order on the NOP option leads to a immediate device reboot and enters a
“Major Fault” mode which must be resolved manually. To trigger the vulnerability, the
NOP option must be put first and the number of options must be more than one” [2].
The vulnerability discovered and reported, was recognized by the PLC manufac-
turer. Rockwell Automation determined there were additional products affected by the
vulnerability and reported this information to NCCIC [3], which result in NCICC
Advisory ICSA-18-172-02 [1]. List of affected products by the vulnerability, according
to Rockwell Automation [1, 13]:
Validation of ICS Vulnerability Related to TCP/IP Protocol Implementation 357
3 Methodology Design
The methodology to prove the stated hypothesis was based on the “vulnerability
details” contained in Sect. 2 of this paper, plus the analysis and definition of all the
elements that had to be considered. Several components needed to be defined:
• Confirmation of available controller with affected firmware version
• Confirmation of required conditions to trigger the vulnerability
• Define platform to perform testing
• TCP/IP protocol details study
• TCP packet Option Field details study
• TCP/IP network protocol analyzer
• TCP Parameters and Options for specific PLC model available
• Exploit method
• Restore procedure of the PLC system, in case of successful exploitation.
Fig. 1. Nmap scan confirms TCP port 44818 open for Ethernet-IP service
In this study, the Kali Linux environment is the Client and the PLC is the Server or
Host (Fig. 3).
The above figure allow to understand the 3WHS sequence needed to establish a
connection. It’s also important to notice that proper handle of sequence and
acknowledge numbering in TCP protocol, is required.
Fig. 4. Valid TCP packet format with multiple TCP options field
3.7 TCP Parameters and Options for Specific PLC Model Available
Each Rockwell Automation PLC model has specific communication parameters over
TCP/IP network protocol, with the communication and configuration software Logix
Designer. Those same parameters were used in the exploit code, to establish
Fig. 5. Wireshark capture showing valid communication parameters with the PLC
362 J. Pavesi et al.
communication with the PLC. Wireshark was used to capture valid 3WHS traffic
between the PLC and Logix Designer (Fig. 5).
Therefore, for Compact Logix 5370 L24ER-QB1B specific available controller
model, the valid parameters and options are:
• Window = 2300
• Maximum Segment Size (MSS) option = 1460
• Window Scale (WS) option = 2
• Selective ACK ok (SAok) option = Yes
• Timestamp option = none.
The testing consisted on running multiple exploit versions, while monitoring and
capturing network traffic with Wireshark. Valid parameters and options previously
shown were used, for Compact Logix 5370 L24ER-QB1B PLC model.
A total of 90 tests were done, changing the python exploit code to achieve the
following combinations:
• SYN crafted packet with 1 NOP option first and then MSS, WS and SAok options
• SYN crafted packet with 2 NOP option first and then MSS, WS and SAok options
Validation of ICS Vulnerability Related to TCP/IP Protocol Implementation 363
• SYN crafted packet with 1 NOP option first and 1 NOP between MSS, WS, SAok
• Same previous one, but changing the location of the 2nd NOP in the Options
• ACK crafted packet with 1 NOP option first and then MSS, WS and SAok options
• ACK crafted packet with 2 NOP option first and then MSS, WS and SAok options
• ACK crafted packet with 1 NOP option first and 1 NOP between MSS, WS, SAok
• Same previous one, but changing the location of the 2nd NOP in the Options
• SYN & ACK crafted packets with 1 NOP option first and then MSS, WS and SAok
options
• SYN & ACK crafted packets with 2 NOP option first and then MSS, WS and SAok
options
• SYN & ACK crafted packets with 1 NOP option first and 1 NOP between other
options
• Same previous one, but changing the location of the 2nd NOP in the Options
• Additional tests, suppressing a RESET (RST) packet that was automatically sent by
Linux Kernel to the PLC, after receiving a SYN + ACK packet from the PLC
• Tests ended using the exact same python exploit code, with valid parameters, that
was capable of exploiting the vulnerability in the Compact Logix 5370
L30ERMS PLC model [10, 11].
In all these tests, communication with the Compact Logix 5370 L24ER-QB1B PLC
was established and confirmed, always having the 3 packets of the 3 Way Handshake,
always getting the SYN + ACK reply from the PLC. Multiple Wireshark network
traffic captures were saved for documentation purpose.
However, the Improper Input Validation vulnerability was never triggered for this
PLC model. There was no major fault of the controller and no denial of service
condition.
5 Conclusions
It´s important to notice that Alexey Perepechko, the same researcher that discovered
and exploited the vulnerability in the Rockwell Automation Allen-Bradley Compact
Logix 5370 L30ERMS PLC model, participated in this research to trigger/exploit the
same vulnerability in the Compact Logix 5370 L24ER-QB1B PLC model. Both PLC
models are supposed to be affected products by the tested vulnerability, according to
Rockwell Automation manufacturer.
Multiple exploit tests were done, including the exact same python exploit code,
with valid parameters, that was used in successful exploitation when this vulnerability
was discovered. Considering the experimental testing and results shown in Sect. 4, the
hypothesis of this research was false.
However, there is sufficient test evidence to establish that Allen-Bradley Compact
Logix 5370 L2 controllers, are not affected by the same type of Improper Input Val-
idation vulnerability, than the Compact Logix 5370 L3 controllers. This evidence is
opposed to the statement of the manufacturer, which informed NCICC that both
controllers are affected products by the same CWE-20 Improper Input Validation
vulnerability.
364 J. Pavesi et al.
References
1. ICS CERT US Homepage. https://ics-cert.us-cert.gov/advisories/ICSA-18-172-02. Accessed
15 June 2019
2. APPLIED RISK Homepage. https://www.applied-risk.com/resources/ar-2018-002. Acces-
sed 15 June 2019
3. NCICC Homepage. https://www.us-cert.gov/. Accessed 15 June 2019
4. NIST Homepage. https://nvd.nist.gov/vuln/detail/CVE-2017-9312. Accessed 15 June 2019
5. MITRE Homepage. http://cwe.mitre.org/data/definitions/20.html. Accessed 15 June 2019
6. RFC Homepage. https://www.rfc-editor.org/info/rfc793. Accessed 15 June 2019
7. RFC Homepage. https://www.rfc-editor.org/info/rfc791. Accessed 15 June 2019
8. Serhane, A., Raad, M. Raad, R., Susilo, W.: PLC Code-level vulnerabilities. In: 3rd
International Conference on Computer and Applications (ICCA), Beirut, Lebanon, pp 348–
352. IEEE (2018)
9. Lee, R., Assante, M., Conway, T.: Analysis of the cyber attack on the Ukrainian power grid.
Technical report, E-ISAC, Washington, DC (2016)
10. Echeverri, D.: Hacking with Python. Zeroxword Computing, Madrid (2015)
11. Echeverri, D.: Python for Pentesters. Zeroxword Computing, Madrid (2014)
12. Krotofil, M., Gollmann, D.: Industrial control systems security - what is happening? In: 11th
International Conference on Industrial Informatics (INDIN), Bochum, Germany, pp 664–
669. IEEE (2013)
13. Milinkovic, S., Lazic, Ljubomir.: Industrial PLC security issues. In: 20th Telecommunica-
tions Forum (TELFOR), Belgrade, Serbia, pp. 1536–1539. IEEE (2012)
14. CVE Details Homepage. https://www.cvedetails.com/cve/CVE-2016-9343. Accessed 15
June 2019
15. NIST Homepage. https://nvd.nist.gov/vuln/detail/CVE-2016-9343. Accessed 15 June 2019
Identifying Components Belonging to Wireless
Connectivity Security
Abstract. Information systems security has been a need for people and
enterprises. This paper presents a review about the different variables studied in
the literature affecting normal operations in Wi-Fi, these variables can change or
affect the normal scenario, conditions, even vulnerabilities, hindering or facili-
tating attack by a third party. According to variables number identified, they are
grouped and related into four categories components. All of this in support of to
achieve a higher level of physical and information security.
1 Introduction
Currently, data security has an important role for enterprises and people. Different
institutions have been trying to unify standards for risks evaluation linked to infor-
mation handling, creating criteria to guarantee data security, integrity and confiden-
tiality. This paper delivers base analysis to establish security level for WiFi networks,
focused on the way the different instructions are performed by the different programs
algorithms and their interaction and, the process management required to coordinate the
different components needed to build the communication up.
The study objective is identifying components belonging to wireless connectivity
security considering the variables that directly connect the physic and data link layers
from OSI model or the access layer from the TCP/IP model (see Fig. 1).
Fig. 1. A figure OSI model, physical and data link layers (Source: prepared)
2 Development
2.1 Security
This component groups the variables in the protection level (see Fig. 3).
The “User Knowledge Level”, refers to the knowledge level the user has to detect a
security fault or to configure the system to avoid the security issue. Users with
knowledge are able to apply security and use support tools like software and apps.
There are three types of “Password”, according to user level: device, technical
administrator and web administrator.
The variable “Router Security Feature”, as there are different types of devices, some
of them with defense incorporated for some kind of attacks and are considered a
security part and the first filter for any suspicious activity.
“Encryption Protocol”, are a first security barrier as cypher the data transmission
through WEP, WPA, WPA2 and soon WPA3.
Because of device variety and different kind of attacks available, the “Firewall” is a
containment barrier considered a variable. In case of an attack, can mitigate attacker
chances and increase router security.
The “Radius Server” is accountable of network access security. It controls if the
user and password are valid, gives access to servers to the user that is attempting to
access the wireless network.
Effective “bandwidth” is the frequency band where is concentrated the most part
of the signal, which is limited by the transmission speed. Therefore, there is a direct
relation between transmission speed and bandwidth: the higher the transmission speed
of a signal, the greater the effective bandwidth [4].
The “attenuation” described as factors, like distance and atmospheric, are con-
ditions affect the transmission energy level and consequently the receiver errors level.
“Noise Signals” are present in the transmission media, non-wanted signals affect
between the transmitter and receiver, distorting the information signal. The noise can
be classified in four categories: Thermal noise, Intermodulation noise, Crosstalk and
Impulsive noise.
“Signal Loss” is consequence of different effects [5], like atmospheric absorption,
and absorption through the interaction with the matter like walls, glass, concrete,
marble, wood, stone, etc. that can be defined as signal absorbers.
The “Physics Effects”, such as diffraction, is energy redistribution in a wave front
when goes at the border of an object; allows wave propagation through the corners [6].
Reflection, when the electromagnetic wave pass through the border between two
mediums, a part of the wave is reflected in the opposite direction [6]. Refraction,
changes the signal direction by passing in an oblique direction from a medium to
another, with different propagation velocity [6].
The “Distance”, defined as the range of space or time that mediates between two
things or events [7]. In wireless transmissions, the signal is dispersed by distance,
therefore the greater the distance from the transmitter, the transmission power range
will be considerably reduced [8].
This variable considers the differences that wireless device manufacturer, in this
case described like “Vendor”, have according to their policies, production schemes,
support and updates. hat the first paragraph of a section or subsection is not indented.
According to the vendor, it is the degree of security at the administration level and
protection barriers that can be estimated as preliminary data to consider, at the time of
performing a pentesting.
Updates are a characteristic associated with the manufacturer, it defines whether the
device is vulnerable or not. It is also possible to mention, the location installed device,
normally they are not installed according to what the installation manual says, It may
not be installed as recommended by the manufacturer’s manual, for ignorance or
changes made without considering the initial implementation, This is an advantage for
any attacker, since the power varies and change the emitted signal direction by the
transmitter. Facilitating detection attacker activity, where the attacker can detect a
signal wireless networks, from any vulnerable place for any of penetration attack.
To pentesting level process, the amount “Connected wireless clients” to the devi-
ces, it is considered a variable. Mainly in the process attack, there are different types,
depend of traffic users generated when navigate on the network, for be used by a
software exploitation execution.
The Operative System of the endpoint, wireless router, and any tools of pentesting
software, are considered as an important variable for vulnerability level. The config-
uration systems and applications installed by operative systems, influence bad practices
adopted by users and administrators, for simple carelessness.
Identifying Components Belonging to Wireless Connectivity Security 371
In process of the activities to be carried out, the “Equipment for Penetration Test”,
it is considered like a variable, considering different stages or phases of process, such as
recognition, attack, among others. The activities depend on different devices such as a
cell phone, notebook or tablet, among others. An interesting example is a notebook or
mobile used from a car in motion to detect wireless networks (wardriving) [13].
3 Conclusion
Identified components have associated the variables that affect the wireless transmis-
sion medium. Some of these variables are not considered in other studies or have been
considered on an isolated way. As result of this paper, these variables have been
grouped to find a relationship between them and define a real scenario affecting the
pentesting procedures. Penetration tests fail to achieve their objectives as only use ideal
environment parameters leaving aside other key parameters.
Therefore, each component variables compilation offers a new transmission med-
ium vision, providing decision making support to feasible attacks and make wireless
network security improvements.
This investigation defines future works on a guide design to evaluate wireless
network security level according to identified components.
Identifying Components Belonging to Wireless Connectivity Security 373
References
1. International Organization for Standardization (ISO homepage) (2019). https://www.iso.org/
standard/20269.html
2. Pacio, G.: Data Center hoy. Buenos Aires - Argentina, año, Editorial Alfaomega (2014)
3. Stallings, W.: Comunicaciones Redes de Computadores. 7ma Edición (2004)
4. Hi, Y., Boile, K.: Antennas from Theory to Practice. Wiley, United Kingdom (2008)
5. Morán, J.: Efectos de interferencia sobre redes basadas en el estándar IEEE 802.11 en el
espectro de 2.4 GHz (2012)
6. Eleftheriades, G.V., Balmain, K.G.: Negative Refraction Metamaterials: Fundamental
Principles and Applications. Wiley-IEEE Press, Hoboken (2005)
7. https://dle.rae.es/?id=Dy4u4UW
8. Barclay, L.W.: Propagation of Radio Wave. IEE (2003)
9. Beard, C.: Wireless Communication Networks and Systems, William Stallings. Pearson,
London (2016)
10. Joshua, W., Cache, J.: Hacking Exposed Wireless: Wireless Security Secrets & Solutions.
McGraw-Hill Osborne Media, New York (2014)
11. Aircrack-ng Homepage, June 2019. https://www.aircrack-ng.org/doku.php?id=
compatibility_drivers
12. Ron, S., et al.: Analytic Methods in Systems and Software Testing. Institute, Technion, Israel
Fabrizio Ruggeri CNR-IMATI, Italy Frederick W. Faltin The Faltin Group and Virginia
Tech, USA (2018)
13. Hurley, C., et al.: WarDriving & Wireless penetration Testing. Syngress Publishing, Inc.
Rockland, Canada (2007)
Author Index