1 Teng Wei and Xinyu ZhangGyro in the Air: Tracking 3D Orientation of Batteryless Internet-of-Things Teng Wei and Xinyu Zhang Department of Electrical and Computer Engineering University of Wisconsin - Madison

2 Batteryless Orientation TrackingOrientation is an essential input Mobile gaming Virtual reality Indoor navigation Battery-powered Many applications require passive orientation sensing Ad-hoc setup Least maintenance Low cost Gaming Smart Home Supply Chain Production Line

3 Technique Review Image-based orientation detectionSensitive to background variation Require good ambient light condition Fail during the occlusion Privacy concern

4 Tagyro: RFID-based Orientation TrackingBasic idea RFID Reader Tag Array Object Standard phase-distance model in the RFID 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑡𝑎𝑔_𝑡𝑜_𝑟𝑒𝑎𝑑𝑒𝑟 𝑃ℎ𝑎𝑠𝑒 2π 𝑤𝑎𝑣𝑒𝑙𝑒𝑛𝑔𝑡ℎ 2

5 estimated orientationTagyro: RFID-based Orientation Tracking Phase Difference of Arrival (PDoA) of a Tag Array Tag Array Tag1 Tag2 θ = 0 θ > 0 θ < 0 θ PDoA (Tag 1 and Tag 2) = 0 𝜑 2 − 𝜑 1 = 0 > 0 𝜑 2 − 𝜑 1 > 0 < 0 𝜑 2 − 𝜑 1 < 0 Tag2 Tag2 RFID Antenna Tag1 Tag1 Tracking algorithm estimated orientation Input: Array layout and phase Process: Compute intensity map that gauges the similarity between measured and theoretical PDoA Output: Orientation

6 Conditions Underlying the Phase-based TrackingTag backscatters like an isotropic point source RFID Antenna Phase changes linearly with tag-to-reader distance Layout of the tag array is known

7 Imperfect Tag Radiation Patternsignal +X +Y +Z RFID Antenna Rotation of a single tag Experiment Setup RSS Phase blind direction antenna polarity

8 Measured and Theoretical PDoA of A Two-Tag ArrayImperfect Tag Radiation Pattern Rotation of a tag array signal Tag2 r Tag1 RFID Antenna Experiment Setup Measured and Theoretical PDoA of A Two-Tag Array Electromagnetic coupling deviates the measured PDoA

9 A Summary of Problems and SolutionsAntenna polarity Solution: Polarity Alignment Not Aligined Aligned Polarity Electromagnetic coupling between tags Solution: Array Layout Sensing Blind direction Cannot track orientation in 3D Solution: Multi-array Tracking

10 Dealing With the CouplingKey observation Tag coupling scales PDoA by a constant factor scaling Scaling factor depends on effective tag-to-tag distance Effective distance Geometry distance Tag1 Tag2 6 cm Geometry distance Effective distance Effective distance Tag2’ 9 cm

11 Array Layout Sensing Sense the effective layout of a tag arrayRandomly rotate tag array and measure the phase t PDoA PDoA scaling Track PDoA scaling, and map it to the effective distance X Y Z Tag 1 … Tag i Tag 1 (ref.) d11 d1i Tag j (ref.) dj1 dji Compute the effective layout from distance matrix Construct a matrix of effective distance for each tag pair

12 Toward 3-DoF Orientation TrackingDual-array and dual-antenna setup Reader antenna 2 Array 1 Array 2 Antenna 1 Y-axis, Z-axis Blind direction Antenna 2 X-axis, Z-axis Array 2 Reader antenna 1 Array 1 X Y Z Algorithmic extension Combo Validator OrientationTracking Orientation Spectrum Synthesis Phase RSS Orientation (𝜃 𝑥 , 𝜃 𝑦 , 𝜃 𝑧 ) Determine which array is out of the blind direction Combine the redundant DoFs

13 Calibrating Reader Frequency-HoppingRandom frequency hopping in RFID reader 902~928 MHz; 50 channels; stay at each channel for 200 ms Observation: Different frequencies induce different initial phase-offsets at the reader Frequency-hopping calibration algorithm Objective: Calibrate the phase so that it looks like output from a single frequency Key: cancel the initial phase-offset and calibrate phase shift of different frequencies

14 Implementation and Evaluation SetupHardware Impinj RFID Reader RFID Antenna Tag Array RFID Tags Object Impinj R420 RFID reader Circular-polarized antenna Three types of tags Evaluation setup Tagyro GUI Real-time implementation

15 Orientation Tracking AccuracyAccuracy vs. DoF Accuracy over Distance Accuracy vs. Size of Tag Array Average tracking error ~ 5-degree

16 Case Study 12-side dice Smart home 91.7% detection accuracyFine-grained record

17 Conclusion Tagyro: 3D passive orientation sensing by using RFID tags, based on phase-distance model Solve key challenges that break the ideal phase-distance model Blind direction Antenna polarity Electromagnetic coupling Extensive evaluation using COTS RFID reader, antenna and tag Demonstrate accurate passive orientation tracking https://www.youtube.com/watch?v=sxTKrBZXP7k

18 Questions? Thank you