Http://nesl.ee.ucla.edu/ 1 Dynamic Location Discovery in Ad-Hoc Networks Andreas Savvides, Athanassios Boulis and Mani B. Srivastava.

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 1

Dynamic Location Discovery in Ad-Hoc

NetworksAndreas Savvides, Athanassios Boulis

and Mani B. Srivastava(asavvide,boulis,[email protected])

Networked and Embedded Systems Lab(NESL)http://nesl.ee.ucla.edu

Electrical Engineering Department

Session 7

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 2

What is location discovery?

• Given a network of sensor nodes where a few nodes know their location how do we calculate the location of the nodes?

Known Location

Unknown Location

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 3

Why?

• Support Location Aware Applications• Navigation• Track Objects • Sensor Networks

– report event origins– evaluate network coverage– assist with routing

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 4

Basic Concepts

• Distance measuring methods– Signal Strength

• Uses RSSI readings and wireless propagation model

– Time based methods• ToA, TDoA• Used with radio, IR, acoustic, ultrasound

– Angle of Arrival (AoA)• Measured with directive antennas or

arrays

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 5

Basic Concepts II

Hyperbolic Trilateration

Triangulation

Multi-lateration– Considers all available beacons

A

B

Cab

c

c

C

b

B

a

A

sinsinsin

)cos(2

)cos(2

)cos(2

222

222

222

aBCCBC

bBCCAB

cABBAC

Sines Rule

Cosines Rule

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 6

Existing TechnologiesINFRASTRUCTURE:• Automatic Vehicle Location system (AVL)

– Base stations keep track of police cars ( uses time based and signal strength methods)

• GPS, Loran• 911 Emergency Location System (ToA, TDoA)

• BAT System(AT&T Cambridge Labs), Cricket (MIT) • RADAR – indoor, uses signal strength maps• RFID tags – IR proximityAD-HOC:• Picoradio (UC Berkeley) – indoor, based on signal strength

maps• GPS-less outdoor localization (Bulusu et. al) – proximity based

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 7

Location Discovery in Ad-Hoc Networks

• No infrastructure support• GPS may not always work

– Costly, Power Hungry, does not work everywhere

• Our Approach– Use RSSI for measuring node separation– But how should the beacons be placed?

• Multiple tradeoffs still an open problem

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 8

Long Range Beaconing

• Long Range BeaconingAdvantages:– Multi-hop Coverage– Works well even in low densities

• Disadvantages:– Low fault tolerance– Requires Dedicated Beacons– Some infrastructure is required

B

B

B

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 9

Our Approach

• Single hop beaconing• Iterative multilateration• Dynamic estimate the wireless channel

parameters• Can be done in conjunction with routing

Advantages:• Data packets are also act as beacon signals• Distributed – relies on neighborhood information• Fault tolerant• Location discovery is almost free!!

Beacon

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 10

Iterative Multilateration• Start with a small number of beacons• Number of beacons increases as more nodes estimate their positions

Initial Beacon

Step 1:

Step 2:

Step 3:

becomes beacon

becomes beacon

becomes beacon

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 11

Challenges

• Multi-path and shadowing effects– Difficult to work in indoor environments

• Beacon placement problem• Bad geometry can affect the quality of the

solution• Variable wireless channel characteristics

– signal propagation differs from place to place

(n=1.5 ... 6)

XdngP

P

t

r ))log(()log()log(

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 12

Solution

• Setup as an over-constrained optimization problem and solve for– Wireless propagation model parameters– Node Locations

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 13

Problem Setup

Wireless Channel Model

Error Distance Representation

nt

r r

gPP

20

2000 )()(),,,(

1

0

yyxxP

gPgnyxf ii

r

ti

n

i

),( 00 yx

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 14

Optimization Problem

• This is a non-linear optimization problem

• Hard to compute in one step• We solve the problem in 2 phases

over multiple iterations• Keep in mind beacon errors!

N

ii ifgnyxF

2

1

2200 )(),,,(

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 15

Two-Phase Approach• Obtain a propagation model estimate based on initial set of beacons• Certainly of node estimates used as weights for the channel estimate• Follow a rip-up and retry method until a predefined set of constraints

is met

Channel Estimator

LocationEstimator

Convergence Criteria?

ResetLocations

NO YES

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 16

Simulations

100 Nodes100 x 100 gridRange = 10Beacons = 10

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 17

Without Beacon Error

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 18

With Beacon Error = 10 %

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 19

Effect of Beacon Error

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 20

Implementation & Measurements

• Implemented Location Discovery Algorithm as part of DSDV routing protocol in SensorSim

• Obtained RSSI measurements using RSC nodes in outdoor environments

• Analyzing the results

http://nesl.ee.ucla.edu/http://nesl.ee.ucla.edu 21

Conclusions and Future Work

• Radio signal strength methods can provide a low cost scalable location discovery

• BUT does not work well indoors – experimenting with ultrasound

• Exploring Collaborative Multilateration• Beacon placement problem needs to be

explored