1. GPS-less Low-Cost Outdoor Localization for Very Small Devices Nirupama Bulusu, John Heidemann, and Deborah Estrin

2. Design Goals RF-based Receiver-based Ad hoc Responsive Low Energy Adaptive Fidelity

3. In this paper … Related Work Algorithm for Coarse-grained Localization Implementation Results

4. Related Work Fine-Grained Localization Coarse-Grained Localization

5. Fine-Grained Localization Range Finding Timing Signal Strength Signal Pattern Matching Directionality Based Electrical Phasing Small aperture Direction Finding

6. Timing Time of flight of communication signal Signal Pattern Global Positioning System Local Positioning System Pinpoint’s 3D-iD Different modalities of communication Active Bat

7. Signal Strength Attenuation of radio signal increases with increasing distance RADAR Wall Attenuation Factor based Signal Propagation Model RF mapping

8. Signal Pattern Matching Multi-path phenomenon Signature unique to given location Data from single point sufficient Robust Substantial effort needed for generating signature database

9. Fine-Grained Localization Range Finding Timing Signal Strength Signal Pattern Matching Directionality Based Electrical Phasing Small aperture Direction Finding

10. Small Aperture Direction Finding Used in cellular networks Requires complex antenna array Disadvantages Costly Not a receiver based approach

11. Coarse-Grained Localization Infrared Active Badge – fixed sensors Fixed transmitters Disadvantages Scales poorly Incurs significant installation, configuration and maintenance costs

12. Localization Algorithm Multiple nodes serve as Reference points Reference points transmit periodic beacon signals containing their positions Receiver node finds reference points in its range and localizes to the intersection of connectivity regions of these points

13. An Idealized Radio Model Perfect spherical radio propagation Identical transmission range for all radios

14. Terms d : Distance b/w adjacent ref. points R : Transmission range of reference point T : Time interval between two successive beacons t : Receiver sampling time N sent (i,t) : No. of beacons sent by R i in time t N recv (i,t) : No. of beacons sent by R i received in t

15. contd… CM i : Connectivity metric for R i S : Sample size for connectivity metric CM thresh : Threshold for CM (X est, Y est ) : Estimated location of receiver (X a, Y a ) : Actual location of receiver

16. contd… CM i = (N recv (i,t) / N sent (i,t)) * 100 t = (S + 1 + ε) * T, 0 < ε « 1 k = No. of reference points within connectivity range (X est, Y est ) = (avg(X i1 +…+X ik ), avg(Y i1 +…+Y ik )) LE = Sqrt( (X est – X a ) 2 + (Y est – Y a ) 2 )

17. Model

18. Validation of Model 78 points measured 68 correct matches Mismatches were all at the edge Error <= 2m CM thresh = 90 R = 8.94m

19. Results T = 2s S = 20 t = 41.9s d = 10m

20. contd… Average error 1.83m Standard deviation 1.07m Max. error 4.12m

21. contd…

22. Simulation to check the effect of increasing the overlap of ref. points Calculated for 10,201 points NO MONOTONIC INCREASE

23. Discussion and Future Work Collision Avoidance Tuning for Energy Conservation Non-uniform reference point placement Reference Point Configuration Robustness Adaptation to Noisy Environment

24. Questions ???