SpinLoc: Spin Around Once to Know Your Location Souvik Sen Romit Roy Choudhury, Srihari Nelakuditi
2 Context Advances in localization technology = Location-based applications (LBAs) (iphone AppStore: 6000 apps, Android: 1000 apps)
2 Context Advances in localization technology = Location-based applications (LBAs) (iphone AppStore: 6000 apps, Android: 1000 apps) Mostly outdoor apps
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost GPS ineffective indoors
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost GPS ineffective indoors WiFi based: Horus, Place Lab, Horus Most need war-driving
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost GPS ineffective indoors WiFi based: Horus, Place Lab, Horus Most need war-driving Infrastructure based: Cricket, Pinpoint High deployment cost
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost GPS ineffective indoors WiFi based: Horus, Place Lab, Horus Most need war-driving Infrastructure based: Cricket, Pinpoint High deployment cost WiFi is ubiquitous
3 Indoor Localization Technology n Limited by accuracy, wardriving and deployment cost GPS ineffective indoors WiFi based: Horus, Place Lab, Horus Most need war-driving Infrastructure based: Cricket, Pinpoint High deployment cost WiFi is ubiquitous Can we improve WiFi localization without wardriving?
Angle of Arrival (AoA) based Localization AP1 AP2 R1 4
Angle of Arrival (AoA) based Localization AP1 AP2 R1 4
Angle of Arrival (AoA) based Localization AP1 AP2 R1 4
Angle of Arrival (AoA) based Localization AP1 AP2 R1 4
Angle of Arrival (AoA) based Localization AP1 AP2 AP with multiple antennas. Enterprise Settings. R1 4
Angle of Arrival (AoA) based Localization AP1 AP2 AP with Multiple antennas Enterprise Settings R1 5
Angle of Arrival (AoA) based Localization AP1 AP2 AoAwith localization phone. AP Multiple on antennas Enterprise Settings R1 5
Angle of Arrival (AoA) based Localization AP1 AP2 AoAwith localization phone. AP Multiple on antennas Enterprise Settings R1 5
Angle of Arrival (AoA) based Localization AP1 AP2 AoAwith localization phone. AP Multiple on antennas Enterprise Settings Privacy preserving. R1 5
We propose SpinLoc AoA based localization aided by user spins 6
Indoor Multipath 7
Indoor Multipath Direct Path Reflected Path 7
Direct Path Energy Direct Path 8
Direct Path Energy Direct Path 8
Direct Path Energy Direct Path Energy Direct Path 8
Direct Path Energy Direct Path Energy Direct Path 9
Direct Path Energy Direct Path Energy Direct Path 9
Direct Path Energy - Human body predominantly water - 2.4Ghz is water s resonant frequency Direct Path Energy Direct Path 9
Direct Path Energy - Human body predominantly water - 2.4Ghz is water s resonant frequency Direct Path Energy Direct Path Hypothesis: Direct path energy minimum when human perfectly blocks phone 9
9 Direct Path Energy Direct Path Human
Direct Path Energy Direct Path Human 10
Direct Path Energy Direct Path Energy Compass Orientation Direct Path Human 10
Direct Path Energy Direct Path Energy Compass Orientation Direct Path Human 10
Direct Path Energy Direct Path Energy Compass Orientation Direct Path Human 11
Direct Path Energy Direct Path Energy Compass Orientation Direct Path Human 11
Direct Path Energy Direct Path Energy Minimum when user blocks phone Compass Orientation Direct Path Human 11
Direct Path Energy 1.4 Normalized Direct Path Energy Direct Path Energy Normalized Direct Path Energy 1.6 1.4 1.2 1 0.8 0.6 0.4 0 60 120 180 Angle 240 300 360 1.2 1 0.8 0.6 0 50 Compass Orientation 100 150 200 Angle 250 300 350 Direct Path Human 12
Direct Path Energy 1.4 Normalized Direct Path Energy Direct Path Energy Normalized Direct Path Energy 1.6 1.4 1.2 1 0.8 0.6 0.4 0 60 120 180 Angle 240 300 360 1.2 1 0.8 0.6 0 50 Compass Orientation 100 150 200 Angle 250 300 350 Direct Path Human 13
Direct Path Energy Normalized Direct Path Energy 1.6 1.4 1.2 1 0.8 0.6 0.4 0 AoA = ( Compass orientation corresponding to minimum direct path energy 60 120 180 Angle + 180 240 300 360 ) mod 360 Direct Path Human 13
RSSI as Direct Path Energy? Direct Path Reflected Path 14
RSSI as Direct Path Energy? Direct Path Reflected Path RSSI 14
Facing
Facing RSSI = Direct + Reflected
Facing RSSI = Direct + Reflected 20 db + 5 db
Facing RSSI = Direct + Reflected 20 db + 5 db = 25 db
Facing RSSI = Direct + Reflected 20 db + 5 db = 25 db Blocking Direct + Reflected
Facing RSSI = Direct + Reflected 20 db + 5 db = 25 db Blocking Direct + Reflected 10 db + 15 db = 25 db
But how do we estimate direct path energy? Using off-the-shelf WiFi cards? 17
Phase Phase of H(f) Amplitude Amplitude of H(f) Channel Frequency Response 60 40 20 0 1 0 1 0 5 10 15 Subcarrier f 20 25 30 Subcarrier Channel Frequency Response 18
Channel Frequency Response Phase Amplitude Amplitude of H(f) Phase of H(f) 60 40 20 0 1 0 1 0 5 10 15 20 25 30 Subcarrier f Subcarrier 18
Channel Frequency Response Phase Amplitude Amplitude of H(f) Phase of H(f) 60 40 20 0 1 0 1 0 5 10 15 20 25 30 Subcarrier f Subcarrier IFFT 25 20 Power delay profile SNR in db 15 10 5 0 0 0.5 1 1.5 Delay in microseconds 19
Channel Frequency Response Direct path Energy Phase Amplitude Amplitude of H(f) Phase of H(f) 60 40 20 0 1 0 1 0 5 10 15 20 25 30 Subcarrier f Subcarrier IFFT SNR in db 25 20 15 10 5 Power delay profile Reflected paths 0 0 0.5 1 1.5 Delay in microseconds 19
Localization using AoA 20
Localization using AoA Actual Direction Estimated Direction 20
Localization using AoA Actual Direction Estimated Direction 20
Localization using AoA Actual Direction Estimated Direction 21
Localization using AoA Actual Direction Estimated Direction 21
Localization using AoA Actual Direction Estimated Direction Centroid 21
Localization using AoA and RSSI Actual Direction Estimated Direction 22
Localization using AoA and RSSI Actual Direction Estimated Direction 22
Localization using AoA and RSSI Actual Direction Estimated Direction 22
Localization using AoA and RSSI Actual Direction Estimated Direction 22
Localization using AoA and RSSI Actual Direction Estimated Direction Centroid 22
Localization using AoA and RSSI Actual Direction Estimated Direction Centroid 1. Angle estimation error: Difference between actual and estimated direction 22
Localization using AoA and RSSI Actual Direction Estimated Direction Centroid 2. Localization error: 1. Angle estimation error: Difference between actual Difference between actual and estimated location and estimated direction 22
Performance Evaluation n Used off-the-shelf Intel 5300 cards to obtain direct path energy n Evaluated SpinLoc at various settings: Engineering Building 30 locations 3-5 APs Cafeteria 25 locations 3-4 APs n Compare with RSSI based trilateration scheme 23
Angle Estimation Error 50 Mean angle error 40 30 20 10 0 10 15 16 20 21 25 26 30 31 40 SNR in db 24
Angle Estimation Error 50 Mean angle error 40 30 20 10 0 10 15 16 20 21 25 26 30 31 40 SNR in db Average error = 20 degrees, for APs stronger than 20dB RSSI 24
Localization Error 1 0.75 CDF 0.5 0.25 Using only angle Using angle + RSSI Using only RSSI 0 5 10 15 20 25 30 35 Localization error in meter Average localization error = 5 meters 25
SpinLoc: Closing Thoughts n AoA based localization without wardriving Human body attenuates direct path to APs SpinLoc identifies this direct path, extract AoA Triangulates to estimate location SpinLoc Limitations: n Requires user involvement: spin once to know your location n Phone listens to WiFi traffic while user spins May not be able to transmit data to the internet while spinning n Depends on existence of direct path from AP May be heavily attenuated n Require extensive testing in crowded environments 26
Questions, comments? Duke SyNRG Research Group http://synrg.ee.duke.edu