Horus: A WLAN-Based Indoor Location Determination System

Transcription

1 orus: A WLAN-Based Indoor Location Determination ystem Moustafa Youssef 2003

2 Motivation biquitous computing is increasingly popular equires Context information: location, time, Connectivity: b, Bluetooth, Location-aware applications Location-sensitive billing Tourist services Asset tracking E911 ecurity

3 Location Determination Technologies GP Cellular-based ltrasonic-based: Active Bat Infrared-based: Active Badge Computer vision: Easy Living Physical proximity: mart Floor Not suitable for indoor Does not work equire specialized hardware calability

4 WLAN Location Determination Triangulate user location eference point Quantity proportional to distance WLAN Access points ignal strength= f(distance) oftware based

5 oadmap Motivation Location determination technologies Introduction Noisy wireless channel orus components Performance evaluation Conclusions and future work

6 WLAN Location Determination (Cont d) ignal strength= f(distance) Does not follow free space loss se lookup table adio map adio Map: signal strength characteristics at selected locations

7 WLAN Location Determination (Cont d) (x i, y i ) [-50, -60] (x, y) 5 [-53, -56] 13 ffline phase Build radio map adar system: average signal strength nline phase Get user location Nearest location in signal strength space (Euclidian distance) [-58, -68]

8 WLAN Location Determination Taxonomy WLAN Location Determination ystems Ad-hoc Mode Infrastructure Mode [Lundberg02] Cell of rigin ignal trength Time of Arrival Daedalus Model-based adio-map Based [Li00] Classification Wheremops Deterministic Probabilistic Example adar orus

9 orus Goals igh accuracy Wider range of applications Energy efficiency Energy constrained devices calability Number of supported users Coverage area

10 Contributions Taxonomy of WLAN location determination systems Modeling the signal strength distributions using parametric and non-parametric distributions andling correlation between successive samples from the same access point Allowing continuous space estimation Clustering of radio map locations andling small-scale variations Compare the performance of the orus system with other systems

11 oadio-map Motivation Location determination technologies Introduction Noisy wireless channel orus components Performance evaluation Conclusions and future work

12 ampling Process Active scanning end a probe request eceive a probe response 2n-1. Probe equest 2n. Probe esponse Channel n 3. Probe equest... Channel 2 4. Probe esponse 1. Probe equest 2. Probe esponse Channel 1

13 ignal trength Characteristics Temporal variations ne access point Multiple access points patial variations Large scale mall scale

14 Temporal Variations

15 Number of amples Collected Temporal Variations eceiver ensitivity Average ignal trength (dbm) 0

16 Temporal Variations: Correlation

17 ignal trength (dbm) patial Variations: Large- cale Distance (feet)

18 patial Variations: mall- cale

19 oadio-map Motivation Goals Introduction Noisy wireless channel orus components Performance evaluation Conclusions and future work

20 Testbeds A.V. William s 4 th floor, AVW 224 feet by 85.1 feet MD net (Cisco APs) 21 APs (6 on avg.) 172 locations 5 feet apart Windows XP Prof. FLA rinoco/compaq cards 3rd floor, 8400 Baltimore Ave 39 feet by 118 feet Linkys/Cisco APs 6 APs (4 on avg.) 110 locations 7 feet apart Linux (kernel 2.5.7)

21 orus Components Basic algorithm [Percom03] Correlation handler [InfoCom04] Continuous space estimator [nder] Locations clustering [Percom03] mall-scale compensator [WCNC03]

22 Basic Algorithm: Mathematical Formulation x: Position vector s: ignal strength vector ne entry for each access point s(x) is a stochastic process P[s(x), t]: probability of receiving s at x at time t s(x) is a stationary process P[s(x)] is the histogram of signal strength at x

23 Basic Algorithm: Mathematical Formulation

24 Basic Algorithm: Mathematical Formulation Argmax [P(x/s)] x sing Bayesian inversion Argmax x [P(s/x).P(x)/P(s)] Argmax x [P(s/x).P(x)] P(x): ser history

25 Basic Algorithm ffline phase adio map: signal strength histograms nline phase Bayesian based inference

26 WLAN Location Determination (Cont d) (x, y) (x i, y i ) P(-53/L1)=0.55 [-53] P(-53/L2)=

27 Basic Algorithm: ignal trength Distributions

28 Basic Algorithm: esults Accuracy of 5 feet 90% of the time light advantage of parametric over non-parametric method moothing of distribution shape

29 Correlation andler Need to average multiple samples to increase accuracy Independence assumption is wrong

30 Correlation andler: Autoregressive Model s(t+1)=.s(t)+(1- ).v(t) : correlation degree E[v(t)]=E[s(t)] Var[v(t)]= (1+ )/(1- ) Var[s(t)]

31 Correlation andler: Averaging Process s(t+1)=.s(t)+(1- ).v(t) s ~ N(0, m) v ~ N(0, r) A=1/n (s 1 +s s n ) E[A(t)]=E[s(t)]=0 Var[A(t)]= m 2 /n 2 { [(1- n )/(1- )] 2 + n+ 1-2 *(1-2(n-1) )/(1-2 ) }

32 Var(A)/Var(s) Correlation andler: Averaging a

33 Correlation andler: esults Independence assumption: performance degrades as n increases Two factors affecting accuracy Increasing n Deviation from the actual distribution

34 Continuous pace Estimator Enhance the discrete radio map space estimator Two techniques Center of mass of the top ranked locations Time averaging window

35 Center of Mass: esults N = 1 is the discrete-space estimator Accuracy enhanced by more than 13%

36 Time Averaging Window: esults N = 1 is the discrete-space estimator Accuracy enhanced by more than 24%

37 orus Components Basic algorithm Correlation handler Continuous space estimator mall-scale compensator Locations clustering

38 mall-scale Compensator Multi-path effect ard to capture by radio map (size/time)

39 mall-scale Compensator: mall-scale Variations AP1 AP2 Variations up to 10 dbm in 3 inches Variations proportional to average signal strength

40 mall-scale Compensator: Perturbation Technique Detect small-scale variations sing previous user location Perturb signal strength vector (s 1, s 2,, s n ) (s 1 d 1, s 2 d 2,, s n d n ) Typically, n=3-4 is chosen relative to the received signal strength d i

41 mall-scale Compensator: esults Perturbation technique is not sensitive to the number of APs perturbed Better by more than 25%

42 orus Components Basic algorithm Correlation handler Continuous space estimator mall-scale compensator Locations clustering

43 Number of amples Collected Locations Clustering educe computational requirements Two techniques Explicit Implicit eceiver ensitivity Average ignal trength (dbm) 0

44 Locations Clustering: Explicit Clustering se access points that cover each location se the q strongest access points =[-60, -45, -80, -86, -70] =[-45, -60, -70, -80, -86] q=3

45 Locations Clustering: esults- Explicit Clustering An order of magnitude enhancement in avg. num. of oper. /location estimate As q increases, accuracy slightly increases

46 Locations Clustering: Implicit Clustering se the access points incrementally Implicit multi-level clustering =[-60, -45, -80, -86, -70] =(-45, =[-45, -60, -70, -80, -86) -86]

47 Locations Clustering: esults- Implicit Clustering Avg. num. of oper. /location estimate better than explicit clustering Accuracy increases with Threshold

48 orus Components Continuous-pace orus ystem Components Correlation Modeler adio Map Builder adio Map and clusters Clustering Applications Location API Estimator mall-cale Compensator Discrete-pace Estimator Correlation andler Estimated Location ignal trength Acquisition API Device Driver (MAC, ignal trength)

49 oadio-map Motivation Location Determination technologies Introduction Noisy wireless channel orus components Performance evaluation Conclusions and future work

50 Avg. Num. of per. per Loc. Est. orus-adar Comparison orus adar Median Avg tdev Max orus (all components) orus (basic) adar

51 Training Time 15 seconds training time per location

52 adio map pacing Average distance error increase by as much as 100% (20 feet) 14 feet gives good accuracy

53 adar with orus Techniques Average distance error enhanced by more than 58% Worst case error decreased by more than 76%

54 oadio-map Motivation Location Determination technologies Introduction Noisy wireless channel orus components Performance evaluation Conclusions and future work

55 Conclusions The orus system achieves its goals igh accuracy Through a probabilistic location determination technique moothing signal strength distributions by Gaussian approximation sing a continuous-space estimator andling the high correlation between samples from the same access point The perturbation technique to handle small-scale variations Low computational requirements Through the use of clustering techniques

56 Conclusions (Cont d) calability in terms of the coverage area Through the use of clustering techniques calability in terms of the number of users Through the distributed implementation Training time of 15 seconds per location is enough to construct the radio-map adio map spacing of 14 feet orus vs. adar More accurate by more than 11 feet, on the average More than an order of magnitude savings in number of operations required per location estimate orus vs. Ekahau

57 Conclusions (Cont d) Modules can be applied to other WLAN location determination systems Correlation handling, continuous-space estimator, clustering, and small-scale compensator Applied to adar Average distance error enhanced by more than 58% Worst case error decreased by more than 76% Techniques presented thesis are applicable to other F-technologies a, g, iperlan, and BlueTooth,

58 Future Work sing the user history in location estimation and clustering Dynamically change the system parameters based on the environment Experimenting with other continuous distributions ptimal placement of access point to obtain the best accuracy Techniques to ensure user privacy

59 Future Work (Cont d) Different clustering techniques Automating the radio-map generation process Changing the radio map based on the environment Effect of adding/removing access points Designing and developing applications and services andling difference between different manufactures