CH03 - 1algorithms Used in Positioning Systems

Algorithms Used in Positioning Systems

Chapter Contents
1. Understanding basic types of wireless localization
2. Learning propagation-based positioning techniques
3. Learning location-fingerprinting-based positioning techniques
4. Introducing the fundamental algorithms for indoor positioning
techniques
5. Evaluating the performance of indoor positioning techniques
6. Comparing different indoor positioning techniques
Classification of Indoor Positioning Techniques

Indoor positioning techniques can be categorized into three general types: scene
analysis, triangulation and proximity. Figure1 provides classification of the three different
types of indoor positioning techniques.
Figure 3.1 Three types of indoor positioning techniques.
Scene Analysis
Scene analysis positioning techniques collects and extracts the features

from observed scene.
The observed features are usually specific and unique. The scene could be
radio frequency waves, acoustic sound, visual images or any other
measurable physical phenomena which usually exist near to the object.
The typical technique of RF-based scene analysis is location fingerprinting

(LF). The observed features are usually specific and unique and are used
to estimate the location of the observer or of observed objects in the scene.
We can estimate the distance by matching the similarity of features. Static
scene analysis searches for observed features in a predefined dataset that
maps them
to object locations. Differential scene analysis estimates location by
tracking the difference between successive scenes.
Triangulation
Triangulation uses the geometric properties of triangles to estimate the

target location.
It can be divided into the subcategories of lateration, using distance
measurements, and angulations using primarily angle measurements.
Lateration estimates the position of an object by measuring its distances
from multiple reference points. So, it is also called range measurement
techniques.
Instead of measuring the distance directly using received signal strengths
(RSS), time of arrival (TOA) or time difference of arrival (TDOA) is usually
measured, and the distance is derived by computing the attenuation of the
emitted signal strength or by multiplying the radio signal velocity and the
travel time.
Roundtrip time of flight (RTOF) or received signal phase method is also used
for range estimation in some systems. Angulation locates an object by
computing angles relative to multiple reference points.
Proximity
The proximity technique usually provides symbolic relative location

information.
This technique identifies an object with a tag and a dense grid of antennas
in a known location detects the tag. If the tag is detected by a single
antenna, we can say that the object is ‘near’ to that antenna.
When more than one antenna detects the mobile target, it is considered to
be collocated with the one that receives the strongest signal.
This method is relatively simple to implement. It can be implemented over
different types of physical media. The presence of the object is sensed using
a mechanism with a limited range, for example, infrared radiation (IR),

radio frequency identification (RFID), pressure and touch sensors
Monitoring is possible when a mobile device or tag is in the range of one or
more antennas. Another example is the cell identification (Cell-ID) or cell
of origin (COO) technique.
This method relies on the fact that mobile cellular networks can identify the
approximate position of a mobile handset by knowing which cell site the
device is using at a given time. The main benefit of Cell-ID is that it is
already in use today and can be supported by all mobile handsets.
After we understand the basic categories of indoor positioning techniques,

we further look at the most typical algorithms, propagation based and
location fingerprinting (LF) based.
3.2 Propagation-based Algorithms
Propagation based algorithms estimates the position by measuring the

received signal strength with path loss.
Signal propagation loss algorithm calculates the received signal strength
(RSS) with path loss as follows:
𝑅 = 𝑟 − 10𝛼𝑙𝑜𝑔10 (𝑑) − 𝑤𝑎𝑙𝑙𝐿𝑜𝑠𝑠
where r is initial RSS, d is a distance from access points (APs) to a location,

𝛼 is the path loss exponent and wallLoss is the sum of the losses introduced
by each wall on the line segment drawn at Euclidean distance d.
It is not easy to model the radio propagation in the indoor environment

because of severe multi-path, low probability for availability of LOS path,
and specific site parameters such as floor layout, moving objects and
numerous reflecting surfaces. There is no good model for indoor radio
multi-path characteristic so far.
Propagation-based techniques usually apply mathematical models to a set
of triangulation algorithms to determine the position of the device.
The triangulation approach uses the geometric properties of triangles to
estimate the target location. It has two derivations: angulation and
lateration.
• Angulation estimates the position of an object by computing angles

relative to multiple reference points.
• Lateration estimates the position of an object by measuring its
distances from multiple reference points. So, it is also called range
measurement techniques.
Figure 3.2 Classification of propagation-based algorithms.
3.3 Location-fingerprinting-based Algorithms
As a fingerprint of a human is almost unique to an individual, a given fingerprint

can be used to recognize a corresponding human.
Similarly, the fundamental principle of the fingerprinting technique for localization

is to use a pattern of a particular characteristic about the surrounding to recognize
a corresponding location.
The particular characteristic about the surrounding at each location can be, for
instance, a thermal image or a set of sensed APs, which is so-called a “fingerprint”
of the location.
The fingerprinting technique normally involves two phases: a phase of generating

fingerprints for different locations and a phase of estimating the current location
based on the fingerprints.
The former phase of generating fingerprints, according to various publications, is

known as the offline phase, learning phase, or training phase.
The latter phase of estimating locations is known as the on-line phase, positioning
phase, or recognition phase.
To keep it consistent, we use the expression of the learning phase and the
positioning phase. Figure 3.3 demonstrates the fundamental principle of
fingerprinting.
Figure 3.3: the fundamental principle of fingerprinting (the learning phase and
the positioning phase).
Fingerprinting uses the surrounding Wi-Fi information sensed at a location as a

fingerprint of the location is known as Wi-Fi fingerprinting. A database, which
contains the generated fingerprints, is referred to as a fingerprint database. It is
also known as a radio map in some Wi-Fi fingerprinting literature.
Lets take a conventional approach of Wi-Fi fingerprinting as an example, as

indicated in Figure 3.4.
Figure 3.4: the collection phase, learning phase and positioning phase of Wi-Fi
fingerprinting
Assume a Wi-Fi enabled device always receives the same signal strength in
the same location (i.e., the RSSs and coordinates), such RSSs and
coordinates would be served as a unique ‘fingerprint’ of this location.
We can collect a ‘fingerprint’ in each location and store them in a database.
Every time we come to a new location, we detect the Wi-Fi signal and
estimate the location by measuring the similarity between current and
stored fingerprints.
Figure 3.5: RF-fingerprinting approach
RF–fingerprinting consists of two phases; the offline phase and the online
phase.
In the offline phase (Training phase) the area of interest is divided into
grids, in each grid, many RSS are collected from surrounding APs and
averaged to remove the fast fading effect, averaged RSS with
corresponding location (also called reference points RP) are stored in a
database known as Radio map.
In Online phase (Real-time phase) RSS measurements are collected from

unknown locations called test points (TP), these measurements are then
compared with the database built in the offline phase.
One possible method is to estimate the smallest Euclidean distance
between the test point measurements and the radio map subspace. The
RP with corresponding smallest Euclidean distance represents the closest
location to TP as shown in the Equation below.
𝑛
2
𝑑 = √∑(𝑅𝑆𝑆𝐼𝑜𝑓𝑙𝑖𝑛𝑒𝑖 − 𝑅𝑆𝑆𝐼𝑜𝑛𝑙𝑖𝑛𝑒𝑖 )
𝑖=1
where:
𝑅𝑆𝑆𝐼𝑜𝑓𝑙𝑖𝑛𝑒𝑖 : RSSI value from 𝑖 𝑡ℎ reference nodes during offline phase;
𝑅𝑆𝑆𝐼𝑜𝑛𝑙𝑖𝑛𝑒𝑖 : RSSI value from i𝑖 𝑡ℎ reference nodes during online phase;
𝑛: Number of received reference nodes;
𝑖: The identity of the reference node
The level of achieved accuracy depends heavily on how many APs and RPs
used. Adding more APs will reduce the possibility of having ambiguous
results and tend to enhance the localization process.
Adding more RPs will enhance resolution; however, this will cost more
labour work. Another disadvantage of this approach is the need for regular
updates for the radio map as the building layout or the number of operating
APs may be changed.
This basic Euclidean Distance algorithm can be used for location estimation
assuming indoor environment is static and small. For dynamically, changing
environments the basic Euclidean Distance algorithm has to be modified by
normalization. This normalization takes into account number of reference
nodes received at a specific point within the calibration phase and
positioning phase. The resulting modified equation is
𝑚
1 2
𝑑 = √ ∑(𝑅𝑆𝑆𝐼𝑜𝑓𝑙𝑖𝑛𝑒𝑖 − 𝑅𝑆𝑆𝐼𝑜𝑛𝑙𝑖𝑛𝑒𝑖 )
𝑚
𝑖=1
where m is a number of reference nodes received

There are at least five LF-based positioning methods, the K-Nearest

Neighbour (K-NN), the probabilistic method, neural networks, support
vector machine (SVM) and the smallest M-vertex polygon (SMP).
Figure 3.5 Five types of location fingerprinting method.
3.3.1 K-nearest neighbours (KNN)

K-nearest neighbours (KNN) is a type of supervised learning algorithm used
for both regression and classification.
KNN tries to predict the correct class for the test data by calculating the
distance between the test data and all the training points.
Then select the K number of points which is closet to the test data. The KNN
algorithm calculates the probability of the test data belonging to the classes
of ‘K’ training data and class holds the highest probability will be selected.
In the case of regression, the value is the mean of the ‘K’ selected training
points.
The K-Nearest Neighbor algorithm requires the online RSS to search for K
closest matches of known locations in signal space from the previously-built
database according to root mean square errors principle.
By averaging these K location candidates with or without adopting the
distances in signal space as weights, an estimated location is obtained via
weighted K-NN or unweighted K-NN.
K-NN classifies a new data point based on the majority of its K-nearest
neighbors. For different applications, different distance functions are
defined to quantify the ‘similarity’ between the training and testing points
In the simplest case (K = 1), the algorithm finds the single closest match
and uses that fingerprint’s location as prediction. In this algorithm, K is the
parameters adapted for better performance
More specifically, the first set of data is the offline samples of RSS from N
APs in the area. Each element in a vector is an independent RSS (in dBm)
collected from APs in the location.
𝑆 = {𝑠1 … 𝑆𝑛 ∈ ℛ 𝑛 }
Where 𝑆 is a set of online sampling LF vectors in database.

The second set of data contains online RSSs,
𝑅 = {𝑟1 … 𝑟𝑛 ∈ ℛ 𝑛 }
from n APs at a particular location.
The K-NN algorithm requires the collection of data {(𝑠𝑖 , 𝑣𝑞 , 𝑖, 𝑞 ∈ 𝑁)} for n
locations in the site, where 𝑣𝑞 is the known location of the 𝑞 𝑡ℎ measurement
and the vector 𝑠𝑖 is the ‘fingerprint’ of the location 𝑣𝑖 .
When a receiver in unknown location A becomes aware of a new fingerprint
r, it searches for the fingerprint 𝑠𝑖 that is closest to r and then estimates
the location.
The unknown location for r is decided by a majority vote from the K shortest
distance fingerprints.
We can estimate the location 𝑣𝑞 by clustering the distance between online

received LF vector r and offline sampling LF vector 𝑠𝑖 as
1
𝑛 𝑞
𝑣𝑞 (𝑟, 𝑓) = (∑|𝑟 − 𝑠𝑖 |𝑞 )
𝑖=1
𝑞 is called Manhattan distance if q = 1 and Euclidean distance if q = 2; the

accuracy does not necessarily become higher as q increases

CH03 - 1algorithms Used in Positioning Systems

Uploaded by

Copyright:

Available Formats

CH03 - 1algorithms Used in Positioning Systems

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CH03 - 1algorithms Used in Positioning Systems

Uploaded by

Copyright:

Available Formats

Algorithms Used in Positioning Systems

Algorithms Used in Positioning Systems

Classification of Indoor Positioning Techniques

Figure 3.1 Three types of indoor positioning techniques.

Scene analysis positioning techniques collects and extracts the features

The typical technique of RF-based scene analysis is location fingerprinting

Triangulation uses the geometric properties of triangles to estimate the

The proximity technique usually provides symbolic relative location

a mechanism with a limited range, for example, infrared radiation (IR),

After we understand the basic categories of indoor positioning techniques,

3.2 Propagation-based Algorithms

Propagation based algorithms estimates the position by measuring the

where r is initial RSS, d is a distance from access points (APs) to a location,

It is not easy to model the radio propagation in the indoor environment

• Angulation estimates the position of an object by computing angles

Figure 3.2 Classification of propagation-based algorithms.

3.3 Location-fingerprinting-based Algorithms

As a fingerprint of a human is almost unique to an individual, a given fingerprint

Similarly, the fundamental principle of the fingerprinting technique for localization

The fingerprinting technique normally involves two phases: a phase of generating

The former phase of generating fingerprints, according to various publications, is

Fingerprinting uses the surrounding Wi-Fi information sensed at a location as a

Lets take a conventional approach of Wi-Fi fingerprinting as an example, as

Figure 3.5: RF-fingerprinting approach

In Online phase (Real-time phase) RSS measurements are collected from

where m is a number of reference nodes received

There are at least five LF-based positioning methods, the K-Nearest

Figure 3.5 Five types of location fingerprinting method.

3.3.1 K-nearest neighbours (KNN)

Where 𝑆 is a set of online sampling LF vectors in database.

We can estimate the location 𝑣𝑞 by clustering the distance between online

𝑞 is called Manhattan distance if q = 1 and Euclidean distance if q = 2; the

You might also like