2009 International Conference on Computer Technology and Development

The Architecture and Characteristics of Wireless Sensor network

LIU Yong-Min 1, 2, WU Shu-Ci 1, NIAN Xiao-Hong 2*

1. College of Computer and Information Engineering 2.* School of Information Science and Engineering
Central South University of Forestry and Technology Central South University
ChangSha 410004, China ChangSha 410075, China
lym37212004@126.com xhnian@mail.csu.edu.cn

Abstract—As a novel information acquiring and processing include in aircraft sow, rocket ejection and manual
technology, WSN (wireless sensor network) is made by the embedded, the common WSN system architecture is shown
convergence of sensor, micro-electro-mechanism system and in figure 1.
networks technologies. WSN consist of small nodes with
sensing, computation and wireless communication capabilities.
Various architectures have been developed for WSN,
depending upon the requirement of application. WSN are used
in different applications e.g. environmental monitoring, habitat
monitoring, home automation, military application etc. In this
paper we present the architecture and characteristics of WSN.
As viewed from the existing work, the hot spots recent
representative routing protocols are analyzed, and their
characteristics and application areas are compared. Finally,
the future research issues in this area are pointed out.
Figure 1. The architecture of Wireless sensor network.
Keywords- architecture; wireless sensor network; routing
technology; characteristics Architecture may include the topology, organization of
sink node, other nodes, and a global view of the whole
I. INTRODUCTION network [3]. The important things that have to be considered
are the autonomy and adaptability. Autonomy specifies that
Wireless sensor networks are composed of a large
sensors can be deployed in an unattended area or physically
number of sensor nodes; communicate with each other
unreachable area and hence they are required to operate with
through wireless transmission. Many feasible applications
the minimum efforts from the base stations or human
are proposed such as industrial sensor networks [1], volcano
administrators. Sensor networks should support adaptability
monitoring networks, habitat monitoring [2], health
in the way to operate through adapting to the environmental
monitoring, and home automation etc. The organization of
changes that sensors monitor, e.g. sensors may decrease their
internal software and hardware should be in a manner that
duty cycles in order to reduce the power consumption in the
will allow them to work properly and be able to adapt
sensors when there is no significant change in sensor
dynamically to new environments, requirements and
readings [4]. Some applications of sensor networks involve
applications. Similarly, it makes sure to be general enough to
mobility. Communication in sensor networks is totally
be suited for as many applications as possible.
depending upon the application and the way of collecting
A sensor node typically consists in five components:
data, mainly it consist of the tasks of data sent to nodes and
sensing, memory, processor, transceiver (transmitter and
the data recorded by nodes about the environment [5].
receiver) and battery. Nowadays, nodes are intended to be
Also, Node deployment is the basic problem in wireless
small and cheap. Consequently, their resources are limited
sensor network. According to the applications, the node
(typically, limited battery, reduced memory and processing
deployment strategy is decided. Most of these work [6], [7],
capabilities). Because of the restrained transmission power,
[8] assume that the environment is sufficiently known and
wireless sensor nodes can only communicate locally, with a
under control. However, when the environment is unknown
certain number of local neighbors. So, nodes have to
or hostile such as remote harsh fields, disaster are as and
collaborate in order to accomplish their tasks: sensing, signal
toxic urban regions, sensor deployment can’t be performed
processing, computing, routing, localization, security, etc. So,
manually. To scatter sensors by aircraft is one possible
WSN are, by its nature, collaborative networks [3].
solution. However, using this technique, the actual landing
Generally, only a small number of sink node take charge
position cannot be controlled due to the existence of wind
of sending instructions, collecting data and communicating
and obstacles such as trees and buildings. Consequently, the
with Internet. The typical WSN deployment approaches
coverage may be inferior to the application requirements no

978-0-7695-3892-1/09 $26.00 © 2009 IEEE 561

DOI 10.1109/ICCTD.2009.44
matter how many sensors are dropped. In building toxic- gateways and the routers to achieve. Not only the nodes of
leaks detection [9], [10], chemical sensors must be placed WSN act as data collector and sender, but also do it acts as
inside a building from the entrance of the building. In such the information router.
cases, it is necessary to make use of mobile sensors, which Application relevance: WSN is centralized data
can move to the correct places to provide the required gathering, multi-hop communication and many-to-one traffic
coverage. pattern. WSN are different from traditional networks and
highly dependent on applications, its ultimate work is
II. SENSOR NETWORK CHARACTERISTICS acquiring the environment data. Different sensor networks
Currently, the wireless network includes mobile application process different the physical signals, a sensor
communications network, wireless LAN, Bluetooth network, networks routing protocols can’t be applied efficiently to the
Ad hoc network and so on. Similar to Ad hoc network, other one. Application relevance is one of the most important
sensor network has many characteristics such as mobility, issues to the WSN.
switching character and the limit capability of the battery
power. Comparing to these wireless networks, WSN also has III. SENSOR NETWORK ARCHITECTURES
some distinctive properties. The characteristics of WSN is The environment is said to be dynamic where, there may
shown as following. be various types of sensor nodes in a single sensor network
Computing capabilities: Because of the limit of cost, i.e. the environment is heterogeneous in terms of software as
size and battery power consumption, program space and well as hardware. BiSNET which is based on the concept
memory space of the sensor is very limited. that bees act autonomously, influenced by local conditions
Battery energy: sensor nodes are often become invalid and local interactions with other bees. A bee colony adapts to
and abandoned because the power is exhaust. Therefore, the dynamic environmental conditions. The BiSNET runtime
protocols and algorithms should to be considered for battery operates a top of TinyOS in each sensor node (figure 2).
energy conservation in advance. In addition, the energy
consumption of the nodes transmit data information is more
than the energy of the nodes implement computing consume.
The power consumption of sensor node transmit a bit of data
can be used to execute 3000 line calculation instructions.
Therefore, how to save power in process of network run so
as to maximize wireless sensor network’s lifetime is a
problem we are facing.
Communication capabilities: Senor network’s
communications bandwidth is narrow and changeful, and its
communications distance is only tens to several hundred
meters. Because the senor will be easily influenced by the
impact of natural environment such as mountains, buildings (a)
and storms, rains and lighting, the terrain obstacles and the
weather. Therefore, it is difficult to maintain WSN running
smoothly. This is to say, the software and hardware of WSN
must be robust and fault-tolerant, security, which is an
important research direction in the future.
Dynamic: the sensor node will exit form the networks
because of the battery exhaustion and other failures. It is
possible that some new sensor nodes be move or added into
network because of tasks requirement. These will bring
about changes in the topology of network, so the WSN
topology must have the function of the reconfiguration, (b)
dynamic and self-adjustment. It is a difficult research spot. Figure 2. The architecture of BiSNET.
No centre, self-organization: The deployment of
wireless sensor nodes does not need preinstall any network Agents are designed to follow several biological
infrastructure. Sensor node, which it can collaboratively principles such as decentralization, autonomy, food
adjust itself perform and distribute algorithm, can rapidly gathering/storage and natural selection. Each agent reads
and automatically form an independent network after nodes sensor data, and discards or reports it to a base station using
switch on. The WSN is an equal network. biological behaviors such as replication, death and migration.
Multi-hop communications: sensor node only can A platform runs on TinyOS and hosts agents. Each platform
communicate with direct neighbors in the WSN. If one node controls the state of sensor node i.e. sleep, listen and
need communicate with the nodes, which is beyond the broadcast and provides a set of runtime services that agents
coverage of the node’s radio frequency. It must be via multi- use to read sensor data and perform behaviors.
hop route transmit data through the intermediate nodes. The A brief description of the layers included can be: the
traditional wired network multi-hop route is used the physical layer addresses mainly the hardware details of the

562
wireless communication mechanism: the modulation type, information of sensing area is not known in advance or is
the transmission and receiving techniques, etc. The data link varied with time, that is the position for sensor deployment
layer is concerned with the Media Access Control (MAC) cannot be determined [9, 12].
protocol that manages communication over the noisy shared Potential field based approach for node deployment is
channel. The network layer manages routing the data deeply discussed in [11], in which nodes are treated as
between the nodes, while the transport layer helps to virtual particles, subject to virtual forces. These forces repel
maintain the data flow. Finally, the application layer contains the nodes from each other and from obstacles, and ensure
(very often) only one single user application. that an initial, compact configuration of nodes will quickly
In addition to the five network layers, the three spread out to maximize the coverage area of the network. In
management planes have the following functionality: the addition to these repulsive forces, nodes are also subject to a
power management plane coordinates the energy viscous friction force. This force is used to ensure that the
consumption inside the sensor node. The architecture based network will eventually reach the state of static equilibrium,
on the five OSI layers together with three management i.e. all nodes will ultimately come to a complete stop. The
planes that go through out the communication protocols viscous force does not, however, prevent the network from
model of Wireless sensor networks [11] (see Figure 3). reacting to changes in the environment; if something is
moved, the network will automatically reconfigure itself for
the modified environment before return once again to a static
equilibrium. Thus, nodes move only when it is necessary to
do so, saving a great deal of energy.
A hybrid approach based on clustering in [13] is used for
load balancing, where the 2D mesh is partitioned into 1D
arrays by row and by column. Two scans are used in
sequence: one for all rows, followed by the other for all
columns. Within each row and columns, the scan operation is
used to calculate the average load and then to determine the
amount of overload and under load in clusters. Load is
shifted from overloaded clusters to under load clusters in an
optimal way to achieve a balanced state. Each cluster covers
a small square area and is controlled by cluster head, knows
the information about cluster’s position in the 2D mesh and
the number of sensors in the cluster.
Figure 3. The communication protols model of WSN.
Limited motilities based approach is discussed in [14],
where sensor can flip (or hop) only once to a new location
In [6], the network space is divided into rectangular equal and the flip distance is bounded. In this framework, the
sized regions (RRs), where the size of the region is set based problem is to determine the optimal way for based sensors to
on the radio range and how many hops the region has to maximize the coverage in the network. After detecting the
cover. Each region (e.g., RRj) is responsible for a set of coverage holes, the sensors move to new position to prevent
resources. The resource key space is divided among these coverage hole. Such movement can be realized in practice by
regions, such that each resource key (Ki) is mapped to a propellers that are powered by fuel, coiled springs that
region. The key way is set to RR mapping (KSeti↔RRj) is unwinds for flipping. In this model, sensors can flip only
known by all nodes. once to a new location.
In fact, the Rendezvous Regions scheme can be built on Sensor node deployment method based on a centralized
top of any routing protocol that can route packets toward virtual force [15], which combines the idea of potential field
geographic regions. The only requirement of the routing and disk packing .In this a powerful cluster head, which
protocol is to maintain approximate geographic information, communicate with all the other sensors, collect sensor
such that given an insertion or lookup to a certain region; it position information, calculate forces and desired position
should be able to obtain enough information to route the for each sensor. The distance between two adjacent nodes
packet toward that region. Given that the packet is able to when all nodes are evenly distributed is defined as a
reach the region, there are several design options inside the threshold to distinguish attractive or repulsive force between
region itself. These design options affect the operation of two nodes. The force between two nodes is zero if their
insertions, lookups, server election, and replication inside the distance is equal to the threshold, attractive if less than and
regions. They also affect the consistency operations for repulsive if greater than. The total force on a node is the sum
mobility and failures. of all the forces given by other sensors together with
obstacles and preferential coverage in the area.
IV. SENSOR NODE DEPLOYMENT STRATEGIES In [16], three protocols are evaluated for sensor network
For static environment, deterministic deployment is used to maximize the sensor coverage with less time, movement
since the location of each sensor can be predetermined distance and message complexity. These protocols first
properly. The stochastic deployment is used when the discover the existence of coverage holes in the target area
based on the sensing service required by the application.

563
After discovering a coverage hole, the protocols calculate the because this design concept has many advantages: reducing
target positions of these sensors, where they should move. the complexity of the core network, facilitate the upgrading
These three protocols are VEC (VECTOR-based), VOR of the network, increasing the versatility and flexibility of
(VORONOI-based) and Mini-max based on the principle of network. So, if it is reasonable choice to use end-to-end
moving sensors from densely deployed are as to sparsely argument in WSN. It has not answer up to now.
deployed areas.
In [18], a centralized deterministic sensor deployment Low energy adaptive clustering hierarchy (LEACH) is a
method, DT-score is the basis. Given a fixed number of cluster-based protocol that minimizes energy dissipation in
deployable sensors, DT-score aims to maximize the area sensor networks by randomly selecting sensor nodes as
coverage of sensing area with obstacles. In the first phase of cluster heads. Power Efficient Gathering in Sensor
DT-score, a contour based deployment is used to eliminate Information System (PEGASIS) is a near optimal chain
the coverage holes near the boundary of sensing area and based protocol. The basic idea of the protocol is to extend
obstacles. In the second phase, a deployment method based network lifetime by permitting nodes to communicate
on the Delaunay Triangulation is applied for uncovered exclusively with their closest neighbors, employing a turn
regions. Before deploying a sensor, each candidate position taking strategy to communicate with the Base Station (BS).
generated from the current sensor configuration is scored by
Minimum Energy Communication Network (MECN)
a probabilistic sensor detection model.
establishes and maintains a minimum energy network for
V. ROUTING PROTOCOLS wireless networks by utilizing low power geographic
WSN is oriented to such business as events monitoring positioning system (GPS). The main idea of MECN is to
and characteristic measuring, mostly data flow is not end-to- find the sub-network with the smallest number of nodes that
end, and multimedia flow is less. In the early time, the study requires the least transmission power between any two
of WSN was once considered that mature Internet particular nodes [18]. Some other representative flat-based
technology combined with Ad-hoc routing protocols and protocol, which are flat sleep scheduling algorithm such as
algorithms are full enough to design the sensor network, but CCP[15]、SPAN[16] and grade sleep scheduling algorithm
the in-depth study shows that: there exists significant such as HEED[17].
differences between wireless sensor networks and traditional
network [8]. We use mobile ad hoc networks (MANET) as Hsiao and others [18] have constructed the balancing tree
an example, which is most similar to WSN to compare. for the wireless network which can achieve the load
1. WSN and MANET most similar performance balancing at the highest level among the wireless nodes, but
characteristics: both are wireless, self-organization, and the flow characteristics of WSN and the kinds of wireless
multi-hop network. networks is different. Dai and others [18] have set
2. WSN nodes is immovable or basically stationary, the centralized algorithm. It looks the grid structure of central-
number of WSN nodes reaches up to 1,000; while the node WSN as a static load-balancing tree, but in reality this
mobility of MANET nodes is minutely, the number of type of network structure are rare.
MANET nodes generally is only a few dozen. In order to solve the dynamic character of the sensor
3. WSN is a data-centric network, whose each node
network tree structure in the reference [12, 13] the
achieves the monitoring through a collaborative approach;
MANET is a network for the purpose of data transmission, researchers have presented dynamic query-tree energy
primarily provides network interconnecting and computing balancing (DQEB) series of protocol to adjust the tree's
power for distributed application. structure and balance the network's energy consumption.
4. WSN’s data packets are smaller than the MANET However, it is at the huge power consumption cost to
network’s, while the relativity of WSN’s data is larger, so its change the structure of tree. In order to eliminate the
data transmission will cost more energy. At the same time, question of data access in WSN, which is caused by the
the data flow direction of WSN has the characteristics: traditional method of data storage.
many-to-one and one-to-many. The load-balancing network (LBN), which is proposed in
5. WSN nodes consume more energy in communication literature, [13] belongs to distribution algorithm. LBN’s
process than in the calculation process. This consumption
difference has certain significance for WSN. However, this basic idea is: WSN is a “supply and demand” data network,
consumption difference is unimportant for MANT. In base station is the only ultimate data demander, and the
addition, WSN node resources (including storage, computing sensor node is the data producer and transporter, data
capacity, communications bandwidth and distance, etc.) are demander “buy” the data from the data producer.
extremely limited. So the WSN node is vulnerable to energy SPEED [7] is stateless non-deterministic geographic
depletion and failure. forwarding protocol (SNGF), which belongs to distribution
6. “End-to-End argument” emphasized that it was put all routing strategy, with the provision of soft real-time end-to-
the dealing process to the end of network. The intermediate end routing capabilities. SPEED has cyclical trigger and
nodes are only responsible for sending data Packet. threshold trigger two kinds of work mode, which can access
Therefore, so it brings huge success of the Internet today, geographic information and load conditions through

564
exchange information between neighbor node, and prefer REFERENCES
chooses a lighter load and better performance neighbor [1] Akyildiz, I., Su, W., Sankarasubramaniam, Y., Cayirci, E., “A Survey
nodes as the next hop, effectively reducing the network load on Sensor Networks”, IEEE Communications Magazine, Aug. 2002,
pp. 102- 109.
and realizing congestion control. SPEED has certain
[2] NingXu, “A Survey of Sensor Network Applications,” University of
scalability. Southern California, http://enl.usc.edu/~ningxu/papers/survey.pdf,
Berkeley-MAC (B-MAC) [8] protocol uses channel 2003.
assessment to ensure the reliability of data transmission in [3] Karl H, Willing A. “Protocols and Architectures for Wireless Sensor
the data link layer, and use low power interception Networks”. New York: Wiley, 2005. 314−340.
technology to reduce idle listening to achieve low-power [4] Pruet Boonma, Paskon Champrasert, Junichi Suzuk. “BiSNET: A
communications. Channel appraisement is adopted weighted Biologically Inspired Architecture for Wireless Sensor Networks”,
sliding average index algorithm to derive the average IEEE 2006.
channel noise through received Receives Signal Strength [5] Wan PJ, Yi CW. “On the longest edge of gabriel graphs in wireless
Ad Hoc networks”. IEEE Trans. on Parallel and Distributed Systems,
Index (RSSI), and then the derived the average channel noise 2007,18(1):111−125.
is compared with the minimum RSSI value in a short time to [6] Chen B, Jamieson K, Balakrishnan H, Morris R. “Span: An energy-
determine the channel state. B-MAC protocol provide a efficient coordination algorithm for topology maintenance in ad hoc
series of bi-directional interface for upper layer, for example, wireless networks”. ACM Wireless Networks Journal,
set Preamble Length and so on. By setting these interfaces, 2002,8(5):481−494.
MAC protocol can be applied to many different types of [7] He T, Stankovic JA, Lu C, Abdelzaher T. “SPEED: A stateless
network traffic. In order to solve the problem of network protocol for real-time communication in sensor networks”. In: Proc.
of 23rd Int’l Conf. on Distributed Computing Systems. Rhode Island:
congestion, B-MAC protocol by using the back-off IEEE Computer Society, 2003. 46−55..
algorithm to allocate channels, which is include initial
[8] Ai J, Abouzeid A. “Coverage by directional sensors in randomly
avoidance and the congestion avoidance two stages, these deployed wireless sensor networks”. Journal of Combinatorial
function can be set up through application program. This Optimization, 2006,11(1):21−41..
kind of avoidance algorithm is similar to the principle of [9] Gupta H, Zhou Z, Das SR, Gu Q. “Connected sensor cover: Self-
AIMD congestion control algorithm.. Organization of sensor networks for efficient query execution”.
IEEE/ACM Trans. on Networking, 2006,14(1):55−67.
VI. CONCLUSION [10] A. Howard, M. J. Mataric and G. S. Sukhatme, “An Incremental
Self Deployment Algorithm for Mobile Sensor Networks,”
According to the above analysis, we can conclude that Autonomous Robots, Special Issue on Intelligent Embedded Systems,
WSN have distinctive features in network structure and September, 2002.
application requirements, compared with the traditional data [11] Goel S, Imielinski T. “Prediction-Based monitoring in sensor
communications network. This paper only discusses a networks: Taking lessons from MPEG”. ACM SIGCOMM Computer
general framework of Routing Protocols, state-of-the-art of Communication Review, 2001,31(5):82−98.
architectures and existing node deployment strategies in [12] Zhang H, Hou JC. “Maintaining sensing coverage and connectivity in
wireless sensor network. Mandatory requirement haven't large sensor networks”. Ad-hoc and Sensor Wireless Networks,
2005,1(1):89−124.
been given to specific protocol and technologies on the every
layer. How to realize the each of the technical details needs [13] G. Wang. G. Caoand T. La Porta, “Movement assist sensor
deployment”, in Proceedings of IEEE Conference on Computer
to be continuously refined and further in-depth study. Communication (INFOCOM), Miami, March 2005.
Certainly, there are still many challenges that need to be [14] Capkun S, Hubaux JP. “Secure positioning of wireless devices with
solved in sensor networks. Related work will be discussed in application to sensor networks”. In: Znati T, Knightly E, Makki K,
the following article. eds. Proc. of the 24th Annual Joint Conf. of the IEEE Computer and
Communications Societies. New York: IEEE Computer Society
ACKNOWLEDGMENT Press, 2005. 1917−1928.
[15] Kawadia V, Kumar PR. “A cautionary perspective on cross-layer
Here, we express the thanks for the profession who have design”. IEEE Wireless Communications, 2005,12(1):3−11..
given the supports and suggestions to this paper, especially [16] Yun quanDong and Qing yiQuan, “Topology Based and Power
the class of students and teachers were led by Professor Adaptive Routing Algorithm for Wireless Sensor Network”, IEEE,
JIANG Xin-Hua and Professor LU Wu-Yi. Finally, we 2007.
would also like to thank the associate professor YANG [17] Younis O, fahmy S. “HEED: A hybrid, energy-efficient, distributed
Sheng-Yue. All of them are teacher in School of Information clustering approach for ad hoc sensor networks”. IEEE Trans. on
Science and Engineering, Central South University. Mobile Computing, 2004,3(4):660−669.
This work supported by National Natural Science [18] Deng J, Han YS, Heinzelman WB, Varshney PK. “Scheduling
Foundation of P. R. China under Grant (60474029, sleeping nodes in high density cluster-based sensor networks”.
ACM/Kluwer Mobile Networks and Applications (MONET),
60774023, 60774045), China Postdoctoral Science 2005,10(6):825−835.
Foundation under Grant (2005038558).

565