차세대 무선 네트워크 및 보안 2008 Fall CS710 Class in KAIST m ulti m edia c omputing laboratory MAL(Mobile-Assisted Localization) in Wireless Sensor Networks Choi

차세대 무선 네트워크 및 보안2008 Fall CS710 Class in KAIST

multimediacomputing laboratory

MAL(Mobile-Assisted Localization) in Wireless Sensor Networks

Choi Chang-hee MMC lab.

Proceedings of IEEE INFOCOM, March 2005. Nissanka B. Priyantha(MIT)


multimediacomputing laboratory

Index

1. Introduction2. Rigidity Theory3. MAL – Distance Measurement4. MAL – Movement Strategy5. Performance Evaluation6. Conclusion

2


multimediacomputing laboratory3

Introduction



IntroductionWhat is the Localization Problem?

• Determine an assignment of coordinates

Node 1

Node 2 Node 3

Input

Node 1

Node 2 Node 3

(0,0) (4,0)

(0,3)

Output



IntroductionSteps of Localization

DistanceMeasurement Localization

Using MAL

Using MAL & AFL( Another paper )

In this paper,



Introduction

• Manually ( ex : Ruler, laser, etc… )• Ultrasonic on sensor node

Previous Methods – Distance Measurement



Introduction

• Physical obstacles ( in especially indoors )• Non-omni-directional hardware• Few distance information

Problem with Previous Methods in Practice

Response curves of the sensorSU-D2000-M30N-C1-POS

Very many obstacles in my life Few data



Introduction

• Use mobility to estimate location!!– Roving human, robot, etc…

Proposed Method

Node 2 Node 3Node 1

Node 4

Ob sta

cle



Rigidity Theory



Rigidity Theory

• Suppose C is a collection of mathematical objects , C is rigid if every c Є C is uniquely determined by less information c about than one would expect.

• Not locally rigid : local graph is not rigid• Locally rigid : locally rigid, but local graphs is not rigid• Globally rigid : global graph is rigid

Definition



Rigidity Theory

• A graph is globally rigid if it is formed by starting from a clique of four non-coplanar nodes and repeatedly adding a node con-nected to at least four non-coplanar existing nodes

Thorem1 – In 3D

Insufficient Sufficient



MAL - Distance Measurement




• In simultaneous equations– Necessary Condition : unknowns – equations ≤ 0– The more we add mn, the more we have (unknowns-equations)

Calculating Distance between Two Nodes - Proposition 2

1 2

n1(a1,b1,c1) Noden2(a2,b2,c2)

1

m1(x1,y1,z1)

Mobile Node

2

m2(x2,y2,z2)

3

m3(x3,y3,z3)

Unknowns : 3 X 5 = 15Equations : 2 X 3 = 615 – 6 = 9 ≥ 0How can we solve this problem!!

obstacle



obstacle


• We need restriction! – Fixed Height ( c1 = c2 (known) , z1 = z2 = z3 = 0 )

– Parallel Line ( b1 = b2 = y1 = y2 = y3 = 0)

Calculating Distance between Two Nodes – Proposition 2

1 2


1

m1(x1,y1,z1)

Mobile Node

2

m2(x2,y2,z2)

3

m3(x3,y3,z3)

Unknowns : 3 X 5 = 15 – 10 = 5Equations : 2 X 3 = 65 – 6 = -1 ≤ 0We can solve this problem!!




• We need restriction!– Fixed Height ( c1=c2=c3(known), z1=z2=z3=z4=z5=z6=0 )

Calculating Distance between Three Nodes – Proposition 3

1 2


1

m1(x1,y1,z1)

Mobile Node

2

m2(x2,y2,z2)3

m3(x3,y3,z3)

Unknowns : 3 X 9 = 27 – 9 = 18Equations : 3 X 6 = 1818 – 18 = 0 ≤ 0We can solve this problem!!

3

n3(a3,b3,c3)

4

m4(x4,y4,z4)

5

m5(x5,y5,z5)

6

m6(x6,y6,z6)




• There is no restriction– j nodes, k mobile positions– Unknowns : 3j-5

• 3D ( 3 X j ), 3 degrees of translational motion, 2 degrees of rotational motion

– Equations : k(j-3)– Required mobile positions : k =┌(3j-5)/(j-3)┐

– J = 4 then k = 7

Calculating Distance between Four Nodes – Proposition 4

1

3Node

1

Mobile Node3

2

45

67

2

4

Unknowns : 3 X 11-5 = 28Equations : 4 X 7 = 2828 – 28 = 0 ≤ 0We can solve this problem!!



MAL – Movement Strategy




• A) Find 4 stationary nodes that can be mea-sured from mobile

• B) Move the mobile to at least 7 spots and measure distances

• C) Compute pair-wise distances between the four stationary nodes

• D) Localize the resulting tetrahedron according to Theorem 1

Initialize




• A) Pick a stationary node that has been localized but has not yet been examined by this loop

• B) Move the mobile around the stationary node, and search not-yet-localized nodes (1~3)

• C) If not-yet-localized nodes are – One, then measure distance with Proposition 2– Two, then measure distance with Proposition 3– Three, then measure distance with Proposition 4

• D) Localize it according to Theorem 1(globally rigid)

Loop



Performance Evaluation




• Localization : AFL(Anchor Free Localization)• Simulation environment: Cricket• No of nodes : 24

Environment

• Real distance : Manual



Performance EvaluationGraph

Graph obtained by MAL Graph after applying AFL



Performance EvaluationPerformance – Error CDF

• CDF of % error between original location and estimated location




• Estimated Location after applying AFL– AFL : avoid folding problem

Performance – Estimated Locations

This spots can be lo-calized by

AFL



Conclusion



Conclusion

• Strong point– Very practical in indoor environment– Very accurate localization conjunction with AFL

• Weak point– Need for ultrasonic device– Need for human resource

Critique



Q&A

Q&A

Documents

차세대 무선 네트워크 및 보안 2008 Fall CS710 Class in KAIST m ulti m edia c omputing laboratory MAL(Mobile-Assisted Localization) in Wireless Sensor Networks Choi