A Practical Approach to Landmark Deployment for Indoor Localization - PowerPoint PPT Presentation

A Practical Approach to Landmark Deployment for Indoor Localization Yingying Chen, John Chen, John- -Austen Francisco, Austen Francisco, Yingying Wade Trappe, and Richard P. Martin Wade Trappe, and Richard P. Martin Dept. of Computer Science Dept. of Computer Science Wireless Information Network Laboratory Wireless Information Network Laboratory Rutgers University Rutgers University th , 2006 May 15 th , 2006 May 15 WINLAB IAB, May 2006

Background RSS Reading Transmit Packet at unknown unknown Transmit Packet at (x 1 ,y 1 ) location location time t Landmarks Rx Rx Landmarks [-35,-68,-56] Modality Modality (x?,y?) Received Signal Strength (RSS) Received Signal Strength (RSS) [(x,y),s1,s2,s3] Time- Time -Of Of- -Arrival (TOA) Arrival (TOA) Angle- -Of Of- -Arrival (AOA) Arrival (AOA) Angle [(x,y),s1,s2,s3] Principle to compute position to compute position Principle Lateration/Angulation Lateration/Angulation angle θ (x 3 ,y 3 ) Scene matching Scene matching Training data/radio map Training data/radio map (x 2 ,y 2 ) Localization results results Localization WINLAB IAB, May 2006 WINLAB IAB, May 2006

Motivation Localizing sensor nodes is a critical input for high- - Localizing sensor nodes is a critical input for high level networking applications: level networking applications: Tracking, monitoring, and geometric- -based routing based routing Tracking, monitoring, and geometric Location- -based services become more prevalent based services become more prevalent Location Recent active research efforts have resulted in a Recent active research efforts have resulted in a plethora of localization methods. plethora of localization methods. Study to improve the deployment of landmarks deployment of landmarks and and Study to improve the thus help a wide variety of algorithms a wide variety of algorithms. . thus help WINLAB IAB, May 2006 WINLAB IAB, May 2006

Contributions Impact of landmark placement on localization Impact of landmark placement on localization performance performance Analytic Model Analytic Model Experimental Results Experimental Results Compute upper bound upper bound on the maximum location error on the maximum location error Compute given the placement of landmarks. given the placement of landmarks. Find optimal patterns optimal patterns for landmark placement for landmark placement Find maxL- -minE minE Novel algorithm maxL Novel algorithm Generic analysis works for a variety of: analysis works for a variety of: Generic algorithms, , networks networks, and , and ranging modalities ranging modalities. . algorithms WINLAB IAB, May 2006 WINLAB IAB, May 2006

Outline Background and motivation Background and motivation Theoretical Analysis Theoretical Analysis Finding an Optimized Landmark Deployment Finding an Optimized Landmark Deployment Experimental Study Experimental Study Conclusion Conclusion Related work Related work WINLAB IAB, May 2006 WINLAB IAB, May 2006

Analysis with Least Squares in Localization Ranging step: Ranging step: Distance estimation between unknown and Distance estimation between unknown and landmarks landmarks Various methods available Various methods available Focus on RSS RSS and and TOA TOA Focus on Lateration step: step: Lateration Traditional: Non- -linear Least squares method linear Least squares method Traditional: Non WINLAB IAB, May 2006 WINLAB IAB, May 2006

Error Analysis Reduce to Linear Least Squares: Reduce to Linear Least Squares: Localization result: ideal ideal Localization result: Localization result: actual actual Localization result: Location estimation error: Location estimation error: With With WINLAB IAB, May 2006 WINLAB IAB, May 2006

Error Analysis The landmark deployments with equal eigenvalues eigenvalues The landmark deployments with equal minimize errors! minimize errors! , where are the singular values of , A where are the singular values of A A T T A A are the squares of the singular The eigenvalues eigenvalues of of A are the squares of the singular The A values of A values of T A A T A can be found as: The eigenvalues eigenvalues of of A can be found as: The WINLAB IAB, May 2006 WINLAB IAB, May 2006

Patterns for Optimal Landmark Placements 5 landmarks 3 landmarks 4 landmarks (square plus (equilateral (square) center of mass) triangle) ? ? ? 7 landmarks 8 landmarks 6 landmarks (square plus (nested squares) (nested triangles) nested triangle) ? ? ? WINLAB IAB, May 2006 WINLAB IAB, May 2006

Finding the Optimal Deployment Analytic analysis gives us shape shape Analytic analysis gives us Length of sides unknown Length of sides unknown Physical constrains of a building of a building Physical constrains MaxL- -MinE MinE Algorithm: MaxL Algorithm: Get optimal pattern optimal pattern based on geometry based on geometry Get Fit optimal pattern into maximum floor size Fit optimal pattern into maximum floor size Stretch/shrink the deployment shape until such the deployment shape until such Stretch/shrink movements stop reducing localization errors movements stop reducing localization errors An iterative search iterative search algorithm algorithm An WINLAB IAB, May 2006 WINLAB IAB, May 2006

Outline Background and motivation Background and motivation Theoretical Analysis Theoretical Analysis Finding an Optimized Landmark Deployment Finding an Optimized Landmark Deployment Experimental Study Experimental Study Conclusion Conclusion Related work Related work WINLAB IAB, May 2006 WINLAB IAB, May 2006

Experimental Study Networks: Networks: 802.11 (WiFi WiFi) ) 802.11 ( 802.15.4 (ZigBee ZigBee) ) 802.15.4 ( Localization algorithms: Localization algorithms: Point- -based: based: RADAR RADAR Point Area- -based: based: ABP (Area Based Probability) ABP (Area Based Probability) Area Lateration: : Lateration BN (Bayesian Networks) BN (Bayesian Networks) LS (Least Squares) LS (Least Squares) Ranging modalities: Ranging modalities: RSS (Received Signal Strength) RSS (Received Signal Strength) TOA (Time of Arrival) TOA (Time of Arrival) WINLAB IAB, May 2006 WINLAB IAB, May 2006

Experimental Setup - 802.11 network 802.11 network - 802.15.4 network 802.15.4 network - - - 4 landmarks in two deployments: 4 landmarks in two deployments: - 4 landmarks in two deployments: 4 landmarks in two deployments: - - Colinear case Colinear case Horizontal case Horizontal case Square case Square case Square case Square case - 115 training points 115 training points - 70 training points 70 training points - - WINLAB IAB, May 2006 WINLAB IAB, May 2006

Evaluation Metrics Error CDF Error CDF Provide statistical specification of the localization Provide statistical specification of the localization accuracy accuracy Average error Average error Average of the distances between the estimated Average of the distances between the estimated location to the true location location to the true location Hö ölder metrics lder metrics H Relates the magnitude of the perturbation in signal Relates the magnitude of the perturbation in signal space to its effect on the localization results space to its effect on the localization results WINLAB IAB, May 2006 WINLAB IAB, May 2006

Localization Accuracy RSS 802.11 Network RSS 802.11 Network Error CDF across algorithms Error CDF across algorithms Square case Square case Colinear case case Colinear WINLAB IAB, May 2006 WINLAB IAB, May 2006

Localization Accuracy RSS 802.15.4 Network RSS 802.15.4 Network Error CDF across algorithms Error CDF across algorithms Square case Square case Horizontal case Horizontal case WINLAB IAB, May 2006 WINLAB IAB, May 2006

Using Time of Arrival Distance estimation based on round trip time between a Distance estimation based on round trip time between a node and a landmark node and a landmark Distance error analysis: TOA vs. RSS Distance error analysis: TOA vs. RSS TOA error modeling: TOA error modeling: RSS TOA WINLAB IAB, May 2006 WINLAB IAB, May 2006

Localization Accuracy TOA 802.11 Network TOA 802.11 Network Error CDF across algorithms Error CDF across algorithms Square case Square case Colinear case case Colinear WINLAB IAB, May 2006 WINLAB IAB, May 2006

Localization Accuracy Optimized landmark deployment Optimized landmark deployment TOA RSS TOA RSS WINLAB IAB, May 2006 WINLAB IAB, May 2006

Conclusion Derived an upper bound an upper bound on the maximum location on the maximum location Derived error given the placement of landmarks error given the placement of landmarks Developed a novel algorithm a novel algorithm, , maxL maxL- -minE minE, for , for Developed finding the optimal landmark placement finding the optimal landmark placement Significant performance improvement of a wide of a wide Significant performance improvement variety of algorithms variety of algorithms ABP and RADAR: > 20% > 20% ABP and RADAR: LS: > 30% > 30% LS: BN: ~ 10% ~ 10% BN: Tension between optimized landmark deployment for between optimized landmark deployment for Tension localization vs. deployments that optimize for signal localization vs. deployments that optimize for signal coverage coverage WINLAB IAB, May 2006 WINLAB IAB, May 2006