Dynamic Object Tracking in Wireless Sensor Networks Tzung-Shi Chen 1, Wen-Hwa Liao 2, Ming-De Huang 3, and Hua-Wen Tsai 4 1 National University of Tainan,

Dynamic Object Tracking in Wireless Sensor Networks

Tzung-Shi Chen1, Wen-Hwa Liao2, Ming-De Huang3, and Hua-Wen Tsai4

1 National University of Tainan, Dept. Information and Learning Technology2 Tatung University, Dept. Information Management

3 Chang Jung Christian University, Dept. Information Management4 National Cheng-Kung University, Dept. CSIE

Wang, Sheng-ShihWang, Sheng-ShihDec. 26, 2005Dec. 26, 2005

IEEE International Conference on Networks (ICON 2005)

Outline

• Introduction• Movement Object Tracking• Simulation• Conclusion

Introduction

• Motivations Target tracking is an important issue in WSNs

• Enemy vehicle tracking, habitat monitoring, etc Cooperation of multiple sensors

• Objectives Mobile object tracking

• Accurate• Quick• Energy efficiency

Movement Object Tracking --- Overview

A

F1

F3

F6

F7

F10

F2 F11

F12

F13

source

mobile target

ingress node

sensor

Movement Object Tracking --- Operation

• Target discovery• Mobile target detection• Target tracking• Track improvement

Face track adjustment Loop face track adjustment

Target Discovery

Target enters the network

Target Discovery (cont’d)

Three sensor nodes detect the target

Target Discovery (cont’d)

Where ?

Three sensor nodes build their corresponding faces

Target Discovery (cont’d)

querypacket

Source sends a query packet to discover the target

Target Discovery (cont’d)

replypacket

• state active• active time infinite

Ingress nodeThe sensor closest to the target replies the packet to the source

Target Discovery (cont’d)

Mobile Target Detection

wak

eup

Wakeup packet• ingress id (node A)• face hop count (1)

A

B

Node A is the 1st ingress node

wakeup

wakeup

wakeup

Mobile Target Detection (cont’d)

wak

eup

Wakeup packet• ingress id (node A)• face hop count (1)

A

B

Node A is the 1st ingress node

wakeup

wak

eup

wakeup

Mobile Target Detection (cont’d)

A

B

Target may enter face F0, F2, or F5

F2

F5

F0

F1

Mobile Target Detection (cont’d)

A

B

Node B sends the wakeup packets to all of its neighboring face nodes (F1, F2, F3, F4, and F5) Node A knows that node B is the next ingress node

F2

F3

F4

F1

F0

F5

Node B is the 2nd ingress node

Mobile Target Detection (cont’d)

A

BF2

F3

F4

F1

active time expires go to sleep

F0

F5

Target Tracking

A

B

C

D

• Node A knows the next ingress node (node B)• Node B knows the next ingress node (node C)• Node C knows the next ingress node (node D)

• Source knows the 1st ingress node (node A)

Face Track Adjustment

A

Face Track Adjustment (cont’d)

AG

B

C

D

E

F

• Track : <A, B, C, D, E, F> (F1, F2, F3, F6, F7, F8, F9)• Face hop count = 7

Not optimaltrack

F1

F2

F3

F4

F5

F6

F7

F8

F9

Face Track Adjustment (cont’d)

AG

B

C

Optimal track : <A, G, F>

F1

F3

F4

F5

F6

F7

F8

F9F

D

EF2

Face Track Adjustment (cont’d)

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9

if (face hop count = k)then send infoadj packet to the last checkface (node A)

F

D

EF2

Face Track Adjustment (cont’d)

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9F

D

EF2

Face Track Adjustment (cont’d)

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9F

D

E

The ingress nodes, B, C, D, and E, are canceled

F2

Face Track Adjustment (cont’d)

AG

B

C

Node A finds the next ingress node closest to the kth ingress node (F),and then sends an adjustment packet to the next ingress node

F1

F3

F4

F5

F6

F7

F8

F9F

D

E

adjustment adjustment

F2

Loop Face Track Adjustment

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9

D

EF2

F

H

Loop Face Track Adjustment (cont’d)

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9

D

EF2

• Track : <A, B, C, D, E, F, H> (F1, F2, F3, F6, F7, F8, F9, F4, F3)

F

H

loop

Loop Face Track Adjustment (cont’d)

A

H

B

C

F1

F3

F4

F5

F6

F7

F8

F9

F

D

EF2

• Node C receives the wakeup packet from node H and detects the loop, node C sends a deletion packet to node H

deletionG

Loop Face Track Adjustment (cont’d)

AG

B

C

F1

F3

F4

F5

F6

F7

F8

F9

D

EF2

F

H

Simulation Model

• ns-2 simulator• Sensing field: 500 m 500 m• 1000 sensor nodes (random deployment)• Communication range = sensing range = 25

m• Power

Tx = Rx = 175 mW Initial: 5 joule

• Moving speed Target: 20 m/s Source: 10 m/s, 20 m/s, or 30 m/s

• Simulation time: 2000 secs

Simulation --- Comparison

• Threshold Flooding (TF) When the source reaches the location of the

target, target discovery process is executed again• Schedule Flooding (SF)

The source executes the target discovery process every 2 secs

• Schedule Updating (SU) The source queries once The sensor which detects the target sends the

update message to the source every 2 secs• Proposed Object Tracking scheme (OT)

Simulation Result --- First Catching Time

infrequent query flooding and face track adjustment

Simulation Result --- Energy Consumption

• before the first catch • after the first catch

flooding-based

less query flooding

Simulation Result --- Lifetime

query packet is required although the source is near the target

Conclusion

• Target tracking protocol Dynamic Energy efficiency Shorter tracking path Loop avoidance