1. 1 Mobile-Assisted Localization in Wireless Sensor Networks Nissanka B.Priyantha, Hari Balakrishnan, Eric D. Demaine, Seth Teller IEEE INFOCOM 2005 March

2. 2 Outline  Introduction  MAL: Mobile-Assisted Localization Distance Measurement Movement strategy  AFL: Anchor-free Localization Proposed in 2003 April  Simulation  Conclusion

3. 3 Introduction  Knowledge of sensor location enable 1. Sensed data useful to application 2. Efficient routing protocol  Node location information is also useful in indoor environment Reference node (known location) Mobile device (like PDA)

4. 4 Introduction cont.  How reference node know location Manually configuring  Cumbersome and error-prone Automatically localize 1. Each node obtain distance to neighbor 2. Computes a coordinate assignment for all node  This paper shows that use mobile device to measure pairwise node distance Then fed the distance information into AFL

5. 5 Indoor Localization Problems OObsturction occlude line-of-sight EX. hard to obtain distance between nodes placed inside and outside a room TThe lack of omni-directional ranging hardware Directional antenna usually pointed toward to user SSparse node deployment Hard to obtain a rigid structure (necessary to obtain a unique solution) GGeometric dilution of precision (GDOP) Incur large error in its position estimate 12 43 2,3 1,4 1 42,3 Globally rigid  A graph is globally rigid if it has a unique embedding

6. 6 GDOP node The intersection point is not precise

7. 7 GDOP node The intersection point is not precise

8. 8 MAL: three distance measurement approach 1. Measure distance between two node at a time 2. Measure distance between three nodes at a time 3. Measure distance between four or more nodes at a time

9. 9 Measure distances among two nodes Assumption: At least need 3 mobile position n o,n 1,m 0,m 1,m 2 are coplanar m 0,m 1,m 2 are collinear Use a larger number of points would reduce GDOP

10. 10 The solution geometry can be computed in polynomial time using standard techniques in real constraint optimization for numerically solving systems of polynomial equations

11. 11 Measure distances among three nodes Assumption: At least need 6 mobile position n o,n 1,n 2 are non-collinear m 0 ~m 5 are coplanar no three of which are collinear

12. 12 Measure distances among four or more nodes  Require at least [(3j-5)/(j-3)] mobile position  Ex. J=4 then [(3j-5)/(j-3)] = 7 At least need 7 mobile position  Assumption n1,n2,n3,n4,m1,m2,m3,m4,m5,m6,m7 No four of which are coplanar

13. 13 Movement Strategy: use 3 nodes at a time Location 1

14. 14 Movement Strategy Location 2

15. 15 Movement Strategy Location 3

16. 16 Movement Strategy Location 4

17. 17 Movement Strategy Location 5

18. 18 Movement Strategy Location 6

19. 19 Movement Strategy Moving around the (perimeter of )visibility region of the node

20. 20 Movement Strategy Find node L

21. 21 Movement Strategy Location 1

22. 22 Movement Strategy Location 2

23. 23 Movement Strategy Location 3

24. 24 Movement Strategy Location 4

25. 25 Movement Strategy Location 5

26. 26 Movement Strategy Location 6

27. 27 Movement Strategy

28. 28 Movement Strategy Moving around the visibility region of the node

29. 29 Distance information graph

30. 30 AFL: Anchor-Free Localization 2003 April Coordinate of node i

31. 31 AFL: Anchor-Free Localization Select n1 to maximize h0,1

32. 32 AFL: Anchor-Free Localization Select n2 that has the maximize hop count from n1

33. 33 AFL: Anchor-Free Localization Find n3 that has the equal hop count to n1 and n2 n3 maximize h1,3 + h2,3 3 hop

34. 34 AFL: Anchor-Free Localization Find n4 that has the equal hop count to n1 and n2 and have max hop count to n3

35. 35 AFL: Anchor-Free Localization Select n5 to minimize |h1,5- h2,5| Pick the node that minimizes |h3,5 - h4,5|

36. 36 AFL: Anchor-Free Localization  Coordinate of node i

37. 37 AFL: Anchor-Free Localization True position AFL Phase1 a(X1,Y1) b(X2,Y2) E ab = || d m (a,b) – (X1-X1) 2 +(Y2-Y2) 2 || 2

38. 38 AFL: Phase 2 Measure distance “Compute” distance Node i Node i’s neighbor  AFL use a optimization algorithm to minimize the Graph Error

39. 39 Didn ’ t know the one hop range and obstruction effect

40. 40 Performance Evaluation Inter-node distance estimate error % CDF of the % edge error Cause by physical obstacle such as furniture

41. 41 Coordinates obtained after running AFL

42. 42 MAL Performance Radius larger reduce more GDOP error

43. 43 MAL Performance Edge length error We can obtain a large number of mobile distance estimates to reduce the distance error

44. 44 Conclusion  We evaluate the algorithm using real- world experiment  The average distance error is less than 1.5%  With sufficient distance samples, AFL can produces coordinate assignment

45. 45

46. 46

47. 47 Mobile-assisted  A mobile collecting distance information between nodes and itself  The challenge is to design movement strategies To collect sufficient distance  The pairwise node distances then be fed into a localization algorithm Anchor-Free Localization (AFL) Algorithm