Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces
0 likes112 views
The document discusses a proposed solution for positioning IoT devices in dense wireless environments using ultrasound technology, which addresses the limitations posed by radio signals in close proximity. It suggests a hybrid approach that combines WiFi for coarse positioning and ultrasound for precise relative positioning, leveraging distributed inference models for angle rotation and distance. Experiments indicate the feasibility of this method, though challenges remain in improving accuracy and managing context variations.
Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces
- 1. Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces PDF → bit.do/180828 #ultrasound #positioning #virtualcoordinates #densewireless #iot Marat Zhanikeev maratishe@gmail.com maratishe.github.io Tokyo Univ. of Science SCAI/RTCSA 2018 @ Hakodate
- 2. Problems / Solutions • too much congestion in dense wireless spaces (like IoT swarms) so alternatives like ultrasound are helpful • distances are too close for radio signal, while utrasound is sufficiently slow to provide 10cm precision • fixed beacons (as in indoors positioning) are not feasible in IoT swarms, instead relative positioning is preferred M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 2/22 2/22
- 3. Proposal M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 3/22 3/22
- 4. Proposal : WiFi + Ultrasound Positioning Sound Range Wireless Range x y Central Node a • WiFi is used for coarse positioning local context (a kind of virtual beacon?) • WiFi context (SSIDs + signal strength) is encoded into ultrasound codes • relative positions are inferred from ultrasound codes • inteded virtual coordinates: X, Y → angle of rotation .. i.e. one-dimensional M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 4/22 4/22
- 5. Proposal : WiFi→Sound Coding • note: WiFi SSIDs can be anything (existing infra), in which case one can use MD5 prefixes as numbers M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 5/22 5/22
- 6. Proposal : Inference • overlap among neighbor ultrasound codes is expected • network coding can help by converting A ⊕ B into relative angle of rotation • note: method is completely distributed, each nodes does its own inference M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 6/22 6/22
- 7. Experiments M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 7/22 7/22
- 8. Ultrasound Processing on Android • minor diffs, but otherwise, can be done on an Android app • setup: 44100Hz sampling freq., 1 frame per 20-30ms, FFT for each frame is duable • otherwise, can aggregate = one FFT per unit of time M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 8/22 8/22
- 9. Capturing Complex Codes over Mic • experiments show that 10kHz precision is possible, 28 ..29 codes can be stuffed into the ultrasound band (17kHz1～22.5kHz) M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 9/22 9/22
- 10. Capturing Codes with High Noise • noisier background, but can still distinguish the code • the fact that background noise is rare in the ultrasound band, helps! M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 10/22 10/22
- 11. Inference of Relative Positions M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 11/22 11/22
- 12. Infering Angle of Rotation • a string tension-based model for optimizing the mapping Collect data Client Infer relative positioning String-based optimization of relative distance Positioning rings can be open -ended (path walking, etc.) Central Node (inferer) M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 12/22 12/22
- 13. First Attempts ... Utterly Fail no 7 10 no 7 7 no 7 5 yes 7 10 yes 7 7 yes 7 5 no 5 10 no 5 7 no 5 5 yes 5 10 yes 5 7 yes 5 5 no 3 10 no 3 7 no 3 5 yes 3 10 yes 3 7 yes 3 5 M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 13/22 13/22
- 14. Infering Angle of Rotation (v2) Overlapping node/area Overlapping node/area Central Node (inferer) • basic unit is a triangle • for each node A, try to pick the best triangle BAC with it in the middle (from a large combination of possible triangles) • for this proposal, only binary difference is calculated, ignoring the signal strength M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 14/22 14/22
- 15. Inference : Perfect Synthetic Trace • a non-trivial inference logic • fails sometimes even for perfect synthetic conditions • this test: 200 nodes spread randomly 8 9 7 8 7 4 9 7 2 8 9 7 8 10 3 9 10 4 8 7 4 8 6 3 7 6 7 67 7 6 8 3 7 8 4 67 7 6 5 2 7 5 1 M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 15/22 15/22
- 16. The Omnipresence Effect xyzvw.170622.ishigaki.airport.json (54 binaryon 10 0.25) omnieffect 10/34 • happens when the same SSID is found in majority of samples • since binary diffs are used, it is difficult to distinguish nodes from each other • ... resulting in mangled results • solution: short of the case in the visual, the problem is resolved by removing omnipresent SSIDs from data M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 16/22 16/22
- 17. Perforamnce M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 17/22 17/22
- 18. Inference : Synthetic vs Real (1) synth.001.json (26 binaryoff 10 0.25) omnieffect 100/100 xyzvw.170621.haneda.json (47 binaryon 10 0.25) omnieffect 20/33 M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 18/22 18/22
- 19. Inference : Synthetic vs Real (2) xyzvw.170622.ishigaki.airport.json (54 binaryon 10 0.25) omnieffect 10/34 xyzvw.170913.akiba.tsukuba.kaisatsu.json (22 binaryon 10 0.25) omnieffect 13/22 M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 19/22 19/22
- 20. Wrapup and Future Work • it works!, but there is room for improvement in part of context Presence/absence of SSID Signal Strength (RSSI) M1: Binary context deltas Binary M2: Binary delta only for absence SSIDs added to absolute RSSI delta Binary absence penalty Diffs of matching SSIDs M3: Take absolute RSSI delta, then add max RSSIs of missing SSIDs Plus RSSI of absent SSIDs Diffs of matching SSIDs M4: Use absolute RSSI delta only of matching SSIDs Diffs of matching SSIDs Additional heuristics based on visual inspection of raw inference data AM1: Use M2, but pay less attention to deltas with adjacent nodes Binary absence penalty Diffs of matching SSIDs , halved for center nodes in triangles AM2: Use M3, but pay less attention to deltas with adjacent nodes Plus RSSI of absent SSIDs Diffs of matching SSIDs , halved for center nodes in triangles AM3: Use M4, but pay less attention to deltas with adjacent nodes Diffs of matching SSIDs , halved for center nodes in triangles M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 20/22 20/22
- 21. That’s all, thank you ... M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 21/22 21/22
- 22. Relation to Interference Coding 01 S.Bhadra+2 ”Network-Coding in Interference Networks” IEEE Symposium on Information Theory (2006) M.Zhanikeev – maratishe@gmail.com Ultrasound Relative Positioning for IoT Devices in Dense Wireless Spaces – bit.do/180828 22/22 22/22