sensor localization presentation1&2
DESCRIPTION
what is sensor localization? distance estimation methods? localization approaches? localization accuracy enhancement.TRANSCRIPT
Prepared by:Gamal Sallam
Prepared for:Dr. Othman Baroudi
What?
◦ To determine the physical coordinates of a group of sensor nodes in a wireless sensor network (WSN)◦ Due to application context, use of GPS is unrealistic
• GPS can work only outdoors.• GPS receivers are too expensive for wide-range deployment. • It cannot work in the presence of obstructions.
Why?◦ To report data that is geographically meaningful i.e., object tracking◦ Services such as routing rely on location information; geographic
routing protocols; context-based routing protocols, location-aware services◦ coverage area management◦ Self deployment
Accuracy: Different applications have different requirements
Energy constraints: All operations involved in localization and tracking must be energy efficient
Signal interference: collisions between packets transmitted by different nodes at the same time
Physical Layer Measurements:- Signal strength, time of arrival, angle of arrival- Prone to physical layer impairments (multipath
propagation, fading, shadowing, noise, etc.)
Computational Constraints:- Sophisticated algorithms cannot be efficiently performed on
wireless sensor nodes because of processing or memory constraints
Triangulation
Finger print
Centroid localization
Next: the common follow of the triangulation approach.
Start
Exist an Unknown Node which has at Exist an Unknown Node which has at least three reference node in its
coverage area
Select an Unknown Node
Reference NodeEstimate the Distance to the
Reference Node
Select Reference Node
Any Selected Reference Node Without Estimated Distance
Any Selected Reference Node Without Estimated Distance
Selected Unknown Node Calculate the Position of the
Selected Unknown Node
Unknown Nod Selection
Distance Estimation
Position Computation
End
The method used for distance calculation:
i. RSSI
ii. LQI
iii. TOA
iv. TDOA
Received signal strength indicator.
- The idea:
- transmission power at the transmitting device (���) directly affects the receiving power at the receiving device (���).
- Using Friis’s free space transmission equation:
(1)
(2)
An ideal distribution of ��� is not applicable
in practice
In practice, the actual attenuation depends on multipath propagation effects, reflections, noise, etc.
These attenuation degrades the quality of the RSSI significantly.
Realistic models replace �� with �� (n=3..5)
Link quality indicator
it indicates how strong the communications link is.
based on the received signal strength as well as the number of errors received.
It is only made available by IEEE 802.15.4 compliant devices.
Distance between sender and receiver of a signal can be determined using the measured signal propagation time and known signal velocity
Sound waves: 343m/s, i.e., approx. 30ms to travel 10m
Radio signals: 300km/s, i.e., approx. 30ns to travel 10m
One-way ToA
one-way propagation of signal
dist��=(t�-t�)*v
Two-way ToA
round-trip time of signal is measured at
sender device
requires highly accurate synchronization of
sender and receiver clocks
two radio signals travelling at different speeds such as radio frequency (RF) and ultrasound.
example: radio signal (sent at �� and received at ���),
followed by acoustic signal (sent at �� and received at ���)
+ve: no clock synchronization
required
+ve: distance measurements can be very accurate
-ve: need for additional hardware
Range-based uses absolute point to-point distance estimates for calculating the location.
more expensive Better accuracy
Range-free doesn’t need such assumption. It assume that hop count proportional to the
their distance (less realistic) cost-effective Less accuracy
In centralized algorithms,
• nodes send data to a central location where computation is performed and the location of each node is determined and sent back to the nodes.
In distributed algorithms,
• each node determines its location by communication with its neighboring nodes
• robust and energy efficient
Centralized: expensive because the power supply for each
node is limited. latency, as well as consuming network
bandwidth.
Decentralized reduce the power-consumption Can be more complex to implement At times may not be possible due to the limited
computational capabilities of sensor nodes
Triangulation
Fingerprint
Centroid
determine the location of a target point by measuring distances to it from three different known points.
Step 1: distribute the beacon
nodes in the area of interest;
Step 2: determine the distance
between each beacon node and
the target node d1,d2, and d3
based on the RSSI, LQI, ToA, or
TDoA values;
Step 3: calculate the
intersection point (the target
node) between the three beacon
nodes with radiuses d1, d2, d3.
We have the following three equations:
Solve the above equations to get x, y.
Problem: d1,d2, and d3 will never be sufficiently accurate.
Divide the area of interest in grids.
determining how the signals will be received at every grid point.
Two phases: offline phase& online phase.
Offline phase:
Step 1: distribute the beacon nodes ��, ��, �� in the area of tracking;
Step 2: divide the area of tracking into several small grids and use the grid points as reference points (x, y)� , (x, y)�,. (x, y)�, …in the tracking area;
Step 3: get the RSS values at each reference point from beacon nodes and store them in the DB with the corresponding locations coordinates.
Online phase:
Step 1: the mobile target enters the tracking area, and then collects the RSS values from each beacon node;
Step 2: compares the collected RSS values with the stored values in the DB;
Step 3: retrieve the position from the DB with the closest RSS values.
Pros:
Better accuracy
Less computation overhead on sensor
Cons:
Collecting RSS values and send them to the server requires long period of time especially if the area is large.
the searching procedure through the stored samples is time consuming.
relies on a high density of beacons. every target sensor node can hear from several
beacons. each target node estimates its location by
measuring the centre of the location of all nodes it hears.
all beacons send their
position �� �, � (� = 1 … , �)to all target sensor nodes within their transmission range.
Then all target sensor nodes calculate their own position ��(x, y) by averaging the coordinates of all n positions of the beacons in range.
Introduces weight functions ��� to improve the accuracy of localization.
��� depends on the distance and the characteristics of the target node receivers.
g depends on the application scenario.
each node maintains a table {��, ��, ℎ�} (location of anchor node i and distance in hops between this node and anchor node i).
when an anchor obtains distances to other anchors, it determines the average hop length (“correction factor” ��), which is then propagated throughout the network.
given the correction factor and the anchor locations, a node
can perform trilateration by multiply ℎ�*c.
Calculate c: C(a1)=100+40/(6+2)=17.5 C(a2)=(40+75)/(2+5)=16.42. C(a3)=(75+100)/(5+6)=16.42. Each anchor send its c value.
Node n will receive first from A2, and will consider it the avgDistance per hop.
so the distance from anchors to node n is calculated by multiplying the minimum hop number and received c.
n−>a1=3∗16.42=49.26, n−>a2=2∗16.42=32.84, n−>a3=3∗16.42=49.26. Then use triangulation to compute node n position If nodes are randomly distributed DV-HOP results in a large
localization error.
The uncertainty of the distance determinations due to the changed application circumstance and the nature of radio signal propagation.
Environment Factor
Eliminating the Outliers of Radio Signals
Evolutionary Optimization
The tracking environment in which a target is located is, in most cases, dynamic, i.e., people waking in an indoor environment, or weather changes in an outdoor environment.
computes the environmental factors between beacon nodes with known positions, based on finding out the relationship between distances and RSS values.
The environment factor ���� can be measured between each
beacon node pair �� and ��
The average environmental factor � can be introduced as the main characteristics for the tracking environment.
Where n is the total number of beacon node covering the mobile target MT.
Each mobile target receives at least three different factors from beacon nodes, in addition to the RSS values for each beacon node.
It compute the average environment factor �.
Compute the distance using this equation:
Then use triangulation to
Calculate the position.
RSSI and LQI are affected by many environment factors such as reflections, obstacle, and other electro-magnetic fields.
Eliminating noise elements will assist in improving the accuracy of the localization.
The Dixon method is used here to eliminate the outlier of RSSI values.
The standard deviation of all the RSSI values received each time is recorded as ���.
The standard deviation threshold is defined as ���.
The RSSI value, noted as ����, obtained from the RSSI measurement is as follows:
m is the number of the RSSI values which are less than or equal to the mean of q RSSI values, alpha is calculated according to the following equation:
In the absence of noise in a system, the intersection of the
circles determines the one and only one target position.
But it yields ambiguous solutions in the presence of noise in the system, since the circles may intersect at multiple points
due to erroneous distance determination.
Consequently, the localization problem
becomes a searching problem.
the location of the target node
is calculated as follows.
A popular statistical localization algorithm
is the nonlinear least squares (NLS) techniques
PSO is a new heuristic method inspired by the social behavior of bird flocking.
particles fly through the problem hyperspace with given velocities.
At each iteration, the velocities of the individual particles are stochastically adjusted according to the historical best position for the particle itself (pBest) and the overall swarm best position (gBest). Both pBest and gBest are derived according to a user defined fitness function.
The fitness function can be defined as follows:
where
the searching space of the blind node can be defined as follow:
Where (��, ��) is the coordinates of the ith reference node;
�� is the measured distance between the blind node and the ithreference node;
���� is the maximum range error of TOF ranging engine in the tunnel environment;
N≥3 is the number of the selected reference nodes.
Then the rectangle defined by (����, ����),(����,, ����) is the searching space of the blind node.
The particles of PSO are randomly initialized in the searching
space at the beginning:
Where (��, ��) is the position of the ��� particle, rand(1) generates a random number with a range of [0,1] and M is the number of the particles.
Each particle updates its position based on its own best exploration, the best swarm overall experience and its previous velocity according to the following model:
Where (��� � , ���(�)) is the current velocity vector of particle j;
while (��� � + 1 , ���(� + 1)) is the velocity vector of particle j for the next iteration;
(�� � , ��(�)) is the current position of particle j;
(�� � + 1 , ��(� + 1)) is the position of particle j of the next iteration;
(pBest�� � , pBest��(�)) is the best position particle j achieved based on its own experience during previous k iterations;
(gBest�� � , gBest��(�)) is the best particle position based on over swarm’s experience during previous k iteration; w is the inertia weight;��, �� are two positive constants; rand(1) is a randomly generated number with a range of [0, 1]; and k is the iteration index.
Challenges:
The space shape is long and narrow: WSN deployed there is of the line or chain type and has low density, and data transmission is energy expensive because of the multiple hops;
The air is wet and dirty due to water and dust, which significantly affects the valid wireless communication distance.
The surface is usually rough and the multi-path effect on radio propagation is severe.
Population 10,
Max iteration 200
c1and c2 1.494,
w 0.729
Satisfied fitness value 1
linear least square estimation (LLSE).
seven potential estimation (SPE)
particle swarm optimization estimation (PSOE)
how to enable enough beacons in the neighborhood and if there are not enough beacons, there how to use some of the mobile target nodes whose locations have been determined as additional beacons.
Mobile target node 1 (Class A) contains three beacons in its range and can get high accuracy and can be used as a reference node.
Mobile target node 2 is covered only by 2 beacon nodes with known position, and one mobile target node with previously determined position, less accuracy.
Class C offers the worst tracking accuracy as the mobile target nodes is covered by only a single beacon nodes and the rest of the available reference nodes are the mobile target nodes with previously determined positions.
The error will be accumulated in Classes B and C.
Service Industry: robots that perform tasks such as basic patient care in nursing
homes, maintenance and security in office buildings.
Requires a mechanism for position estimation.
Skilligent uses a visual localization system based on pattern matching.
Pollution Monitoring Sensor nodes that measure specific pollutants in the air are
mounted on vehicles.
As the vehicles move along the roadways, the sensors sample the air, and record the concentration of various pollutants along with location and time.
When the sensors are in the proximity of access points, the data are uploaded to a server and published on the web.
Shooter Detection / Weapon Classification: a soldier-wearable sensor system is developed that not only
identifies the location of an enemy sniper, but also identifies the weapon being fired.
Each sensor consists of an array of microphones mounted on the helmet of a soldier.
The sensor observes both the shock wave of the projectile, as well as the muzzle blast from the weapon, and based on TDOA, as well as properties of the acoustic signal, is able to triangulate the enemy position and classify the weapon type.
Pothole Detection: a system is developed to detect potholes on city streets.
Deployed on taxi cabs, the sensor nodes contain an accelerometer, and can communicate using either opportunistic Wi-Fi or cellular networks.
1. Shuang-Hua Yang,” Wireless Sensor Networks Principles, Design and Applications”, chapter 10, Springer, 2014.
2. Tareq Alhmiedat, “Tracking Mobile Targets through Wireless Sensor Networks”, A Doctoral Thesis, Oct, 2009.
3. Qin Y., Wang F., Zhou, C., Yang, S.H.: A particle swarm optimization based distributed localization scheme in tunnel environment. Wireless Sensor Systems—IET Conference, June,London (2012) .
4. M. Keshtgary, M. Fasihy, and Z. Ronaghi,” Performance Evaluation of Hop-Based Range-Free Localization MethodsinWirelessSensorNetworks” International Scholarly Research NetworkISRN Communications and Networking, 2011.
5. Isaac Amundson and Xenofon D. Koutsoukos, “A Survey on Localization for Mobile Wireless Sensor Networks”, Mobile Entity Localization and Tracking in GPS-less Environnments, Volume 5801, 2009, pp 235-254.
6. Grossmann, Ralf, et al. "Localization in Zigbee-based sensor networks."Proceedings of 1st European ZigBee Developers Conference, EuZDC. 2007.