RFID Object Localization Gabriel Robins and Kirti Chawla Department of Computer Science University of Virginia
Outline What is Object Localization ? Background Motivation Localizing Objects using RFID Experimental Evaluation Conclusion 02/33
What is Object Localization ? Goal: Find positions of objects in the environment Problem: Devise an object localization approach with good performance and wide applicability 03/33 ObjectsEnvironments
Current Situation 04/33 Lots of approaches and applications lead to vast disorganized research space Inapplicable Not general Mismatched Identify limitations Determine suitability Techniques Signal arrival angle Signal strength Signal arrival time Signal phase Technologies Satellites Lasers Ultrasound sensors Cameras Applications Outdoor localization Indoor localization Mobile object localization Stationary object localization
Localization Type 05/33 SelfEnvironmental Self-aware of position Processing capability Not aware of position Optional processing capability
Localization Technique 06/33 Signal arrival time Signal arrival difference time Signal strength Signal arrival phase Signal arrival angle Landmarks Analytics (combines above techniques with analytical methods)
RFID Technology Primer 07/33 RFID reader RFID tag Passive Semi-passive Active Interact at various RF frequencies Inductive Coupling Backscatter Coupling
Motivating RFID-based Localization 08/33 Low-visibility environments Not direct line of sight Beyond solid obstacles Cost-effective Adaptive to flexible application requirements Good localization performance
State-of-the-art in RFID Localization 09/33 Pure RFID –based localization approaches Hybrid
Contributions 10/33 Pure RFID-based environmental localization framework with good performance and wide applicability Key localization challenges that impact performance and applicability
Power-Distance Relationship 11/33 Reader powerDistanceTag power Cannot determine tag position Empirical power- distance relationship
Empirical Power-Distance Relationship 12/33 Insight: Tags with very similar behaviors are very close to each other
Tag Sensitivity 13/33 Variable sensitivities Bin tags on sensitivity Average sensitiveHigh sensitive Low sensitive Pile of tags Key Challenges Results 25 %54 %8 % 13 %
Reliability through Multi-tags 14/33 Platform design Results Insight: Multi-tags have better detectabilities (Bolotnyy and Robins, 2007) due to orientation and redundancy
Tag Localization Approach 15/33 Setup phase Localization phase
Algorithm: Linear Search 16/33 Linearly increments the reader power from lowest to highest (LH) or highest to lowest (HL) Reports the first power level at which a tag is detected as the minimum tag detection power level Localizes the tags in a serial manner Time-complexity is: O(# tags power levels)
Algorithm: Binary Search 17/33 Exponentially converges to the minimum tag detection power level Localizes the tags in a serial manner Time-complexity is: O(# tags log(power levels))
Algorithm: Parallel Search 18/33 Linearly decrements the reader power from highest to lowest power level Reports the first power level at which a tag is detected as the minimum tag detection power level Localizes the tags in a parallel manner Time-complexity is: O(power levels)
Reader Localization Approach 19/33 Setup phase Localization phase
Algorithm: Measure and Report 20/33 Reports a 2-tuple TagID, Timestamp after reading a neighborhood tag Sorted timestamps identify object’s motion path Time-complexity is: O(1)
Localization Error 21/33 Reference tag’s location as object’s location leads to error Number of selection criteria Error-reducing Heuristics
Experimental Setup 22/ Y-axis X-axis Track designMobile robot design
Experimental Evaluation 23/33 Empirical power-distance relationship Localization performance Impact of number of tags on localization performance
Empirical Power-Distance Relationship 24/33
Localization Accuracy 25/33
Algorithmic Variability 26/33
Localization Time 27/33
Performance Vs Number of Tags 28/33 Diminishing returns
Comparison with Existing Approaches 29/33 Hybrid
Visualization 30/33 Accuracy Work area Antenna control Heuristics
Deliverables 31/33 Patent(s): 1.Kirti Chawla, and Gabriel Robins, Method, System and Computer Program Product for Low- Cost Power-Provident Object Localization using Ubiquitous RFID Infrastructure, UVA Patent Foundation, University of Virginia, 2010, US Patent Application Number: 61/386,646. Journal Publication(s): 2. Kirti Chawla, and Gabriel Robins, An RFID-Based Object Localization Framework, International Journal of Radio Frequency Identification Technology and Applications, Inderscience Publishers, 2011, Vol. 3, Nos. 1/2, pp Conference Publication(s): 3.Kirti Chawla, Gabriel Robins, and Liuyi Zhang, Efficient RFID-Based Mobile Object Localization, Proceedings of IEEE International Conference on Wireless and Mobile Computing, Networking and Communications, 2010, Canada, pp Kirti Chawla, Gabriel Robins, and Liuyi Zhang, Object Localization using RFID, Proceedings of IEEE International Symposium on Wireless Pervasive Computing, 2010, Italy, pp Grant(s): 5. Gabriel Robins (PI), NSF Grant on RFID Pending
Conclusion 32/33 Pure RFID-based object localization framework Key localization challenges Power-distance relationship is a reliable indicator Extendible to other scenarios
33/33 Thank You
34 Backup Slides
Key Localization Challenges 35 RF interference Occlusions Reader locality Tag spatiality Tag sensitivity Tag orientation Back
Single Tag Calibration 36 Constant distance/Variable power Variable distance/Constant power Back
Multi-Tag Calibration: Proximity 37 Constant distance/Variable power Variable distance/Constant power Back
Multi-Tag Calibration: Rotation 1 38 Constant distance/Variable power Back
Multi-Tag Calibration: Rotation 2 39 Variable distance/Constant power Back
Error-Reducing Heuristics 40 Heuristics: Absolute difference Back
Error-Reducing Heuristics 41 Heuristics: Minimum power reader selection Back
Error-Reducing Heuristics 42 Heuristics: Root sum square absolute difference Back
Error-Reducing Heuristics 43 Localization error Root sum square absolute difference Meta-Heuristic Minimum power reader selection Absolute difference Other heuristics Back