Decimeter-Level Localization with a Single WiFi Access Point Deepak - PowerPoint PPT Presentation

Decimeter-Level Localization with a Single WiFi Access Point Deepak Vasisht Swarun Kumar, Dina Katabi

Indoor Localization is Cool! SpotFi [SIGCOMM’ 15], ToneTrack [Mobicom’ 15], Phaser [Mobicom’ 14], Tagoram [Mobicom’ 14], LTEye [SIGCOMM’ 14], ArrayTrack [NSDI’13], PinPoint [NSDI’13], PinIt [SIGCOMM’13], Zee [MobiCom’12], PinLoc [MobySys’12], EZ [MobiCom’10], …. • Locate off-the-shelf devices • Accuracy of tens of cm

But… They Need 4-5 Access Points Homes and small businesses have ON ONE access point (AP)

Application : Control heating based on occupancy

Application : WiFi Geo-Fencing

Application : Device-to-device Localization Enable device-to-device localization without infrastructure support

Chronos • Enables decimeter-accurate localization using a single off-the-shelf WiFi card • A novel algorithm to estimate propagation time to sub-nanosecond accuracy using a WiFi card • Implemented and evaluated in practical settings

Why past work needs multiple AP’s? θ

Single Access Point? distance θ

Measuring Distance Distance = speed of light x propagation delay

Measuring Distance Distance = speed of light x propagation delay How do we measure propagation delay?

Propagation Delay t Phase of the signal( ) = 𝑛𝑝𝑒 2𝜌 2𝜌𝑔𝑢 ϕ

Propagation Delay: Example 2.41 GHz 0 1 0.5 1.5 t (ns) 𝜚 = 2𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌

Propagation Delay: Example 5.8 GHz 5.18 GHz 2.48 GHz 2.41 GHz 1 0 0.5 1.5 t (ns)

Mathematically 𝜚 - = 2𝜌𝑔 - 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 . = 2𝜌𝑔 . 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 / = 2𝜌𝑔 / 𝑢 𝑛𝑝𝑒 2𝜌

Mathematically 𝜚 - = 2𝜌𝑔 - 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 . = 2𝜌𝑔 . 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 / = 2𝜌𝑔 / 𝑢 𝑛𝑝𝑒 2𝜌 Use Chinese Remainder Theorem to get the propagation delay

Can’t measure propagation delay without detection delay Distance = speed of light x propagation delay Measured delay = propagation delay + detection delay

Packet Detection Delay Detection Delay Detection Decoding • Detection delay >> Propagation delay How do we eliminate detection delay? Detection Delay ≈ 200 ns, Propagation Delay ≈ 20 ns § • Detection delay is unpredictable

Problem: Separate detection delay from propagation delay Solution: Leverage that propagation delay and detection delay happen at different frequencies f f-f c Detection Decoding Detection Propagation Delay Delay

f-f c f Detection Decoding Detection Propagation Delay (t’) Delay (t) 𝜚 = 2𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌 𝜚 = 2𝜌𝑔𝑢 + 2𝜌 𝑔 − 𝑔 4 𝑢′ 𝑛𝑝𝑒 2𝜌

f-f c f Detection Decoding Detection Propagation Delay (t’) Delay (t) 𝜚 = 2𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌 𝜚 = 2𝜌𝑔𝑢 + 2𝜌 𝑔 − 𝑔 4 𝑢′ 𝑛𝑝𝑒 2𝜌 0 Idea: Use OFDM to measure phase at f=f c

But WiFi does not transmit at f=f c

Solution: Leverage OFDM 𝜚 = 2𝜌𝑔𝑢 + 2𝜌 𝑔 − 𝑔 4 𝑢′ 𝑛𝑝𝑒 2𝜌 𝜚 4 = 2𝜌𝑔 4 𝑢 + 0 𝑛𝑝𝑒 2𝜌 Phase Phase at f=f c 𝜚 4 𝑔 𝑔 4 + 2𝜀 𝑔 𝑔 4 − 2𝜀 𝑔 4 − 𝜀 𝑔 4 + 𝜀 4 + 3𝜀 𝑔 4 − 3𝜀 4

Mathematically 𝜚 4,- = 2𝜌𝑔 4,- 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 4,. = 2𝜌𝑔 4,. 𝑢 𝑛𝑝𝑒 2𝜌 𝜚 4,/ = 2𝜌𝑔 4,/ 𝑢 𝑛𝑝𝑒 2𝜌 Chronos eliminates packet detection delay by leveraging OFDM properties

Additional System Components • Initial Phase Offset Compensation • Multipath resolution

Additional System Components • Initial Phase Offset Compensation • Multipath resolution

Initial Phase Offsets t 𝜚 = 2𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌 𝜚 = 2𝜌𝑔𝑢 + Δ𝜚 𝑛𝑝𝑒 2𝜌

Idea: Use Acknowledgements t 𝜚 - = 2𝜌𝑔𝑢 + Δ𝜚 𝑛𝑝𝑒 2𝜌 𝜚 . = 2𝜌𝑔𝑢 − Δ𝜚 𝑛𝑝𝑒 2𝜌 𝜚 - + 𝜚 . = 4𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌

Idea: Use Acknowledgements t 𝜚 - = 2𝜌𝑔𝑢 + Δ𝜚 𝑛𝑝𝑒 2𝜌 𝜚 . = 2𝜌𝑔𝑢 − Δ𝜚 𝑛𝑝𝑒 2𝜌 Chronos eliminates phase offsets by using 𝜚 - + 𝜚 . = 4𝜌𝑔𝑢 𝑛𝑝𝑒 2𝜌 acknowledgements

Additional System Components • Initial Phase offset Compensation • Multipath resolution

Problem: Multipath Effect

Solution: Find delays for each path 10 ns 5.2 ns 16 ns Distance to source corresponds to the smallest delay.

Experimental Evaluation

Implementation • Evaluation with off-the-shelf Intel WiFi 5300 cards • Kernel modifications to the iwlwifi driver in the Ubuntu kernel • Ground truth measurements using laser distance measurement device ( 1mm accurate)

Evaluation Testbed: Office Environment 20 m 20 m

Distance Measurement Accuracy 1 0.8 0.6 CDF LOS 21 cm 0.4 NLOS 14 cm 0.2 0 0 1 2 3 Error (m)

SpotFi (SIGCOMM’ 15) Localization Accuracy 3 AP’s 190 cm 1 4 AP’s 80 cm 0.8 5 AP’s 60 cm 0.6 CDF 98 cm 0.4 LOS 65 cm NLOS 0.2 0 0 2 4 Error (m)

SpotFi (SIGCOMM’ 15) Localization Accuracy 3 AP’s 190 cm 1 4 AP’s 80 cm 0.8 5 AP’s 60 cm 0.6 CDF 98 cm 0.4 LOS 65 cm NLOS 0.2 Chronos can achieve state-of-the-art 0 localization accuracy with a single AP 0 2 4 Error (m)

Applications Smart Homes Device to Device WiFi Geo-fencing localization

Applications Smart Homes Device to Device WiFi Geo-fencing localization

Application: Smart Homes Bedroom1 Bedroom2 Living room 9 m Kitchen Bath 13 m

Application: Smart Homes Bedroom1 Bedroom2 Living room 9 m Chronos detects the correct room with Kitchen Bath accuracy 94%. 13 m

Applications Smart Homes Device to Device WiFi Geo-fencing localization

Application: GeoFencing Coffee Station 7 m 9 m

Application: GeoFencing Coffee Station 7 m Chronos can accurately authenticate WiFi 9 m users with 97% accuracy.

Applications Smart Homes Device to Device WiFi Geo-fencing localization

Application: TakeMyPicture Drone

Application: TakeMyPicture Drone 3 2 1 y (m) 0 User -1 Drone -2 -3 x (m) -2 0 2

Application: TakeMyPicture Drone 1 0.8 0.6 CDF 0.4 4.2 cm 0.2 0 0 5 10 15 Error (cm)

Application: TakeMyPicture Drone 1 0.8 0.6 CDF 0.4 4.2 cm 0.2 Chronos enables a drone to follow the user 0 0 5 10 15 with no infrastructure support. Error (cm)

Related Work • WiFi Localization: SpotFi [SIGCOMM’ 15], ToneTrack [Mobicom’ 15], Phaser [Mobicom’ 14], Tagoram [Mobicom’ 14], …. • Closest Work: SAIL [MobiSys’ 14]

Conclusion • Chronos is the first system to enable accurate localization on off-the-shelf WiFi cards • Its key enabler is a novel algorithm that can estimate accurate propagation delay, by eliminating the detection delay