spotfi: decimeter level localization using wifinetseminar.stanford.edu/seminars/10_15_15.pdf ·...

SpotFi: Decimeter Level Localization using WiFi

Manikanta Kotaru, Kiran Joshi, Dinesh Bharadia, Sachin KattiStanford University

Indoor localization platform providing decimeter-level accuracy could enable a host of applications

Targeted Location Based Advertising

Indoor Navigation (e.g. Airport Terminals)

Real Life Analytics(Gym, Office, etc..)

Applications of Indoor Localization

2

Easily Deployable

3

• Commercial WiFi chips

Easily Deployable• Commercial

WiFi chips• No hardware

or firmware change

4

Easily Deployable• Commercial

WiFi chips• No hardware

or firmware change

• No User Intervention

5

Easily Deployable, Universal

6

• Localize any WiFi device

• No specialized sensors

Easily Deployable, Universal, Accurate

7

1 m

• Error of few tens of centimeters

State-of-the-art

8

System Deployable Universal AccurateRADAR, Bahl et al, ’00 9 9 8

HORUS, Youssef et al, ’05 8 9 9

ArrayTrack, Xiong et al, ’13 8 9 9

PinPoint, Joshi et al, ’13 8 9 9

CUPID, Sen et al, ’13 9 8 8

LTEye, Kumar et al, ’14 8 9 9

Phaser, Gjengset et al, ’14 8 9 9

Ubicarse, Kumar et al, ’14 9 8 9

State-of-the-art

9

System Deployable Universal AccurateRADAR, Bahl et al, ’00 9 9 8

HORUS, Youssef et al, ’05 8 9 9

ArrayTrack, Xiong et al, ’13 8 9 9

PinPoint, Joshi et al, ’13 8 9 9

CUPID, Sen et al, ’13 9 8 8

LTEye, Kumar et al, ’14 8 9 9

Phaser, Gjengset et al, ’14 8 9 9

Ubicarse, Kumar et al, ’14 9 8 9

SpotFi, Kotaru et al, ’15 9 9 9

System Overview

10

Localization - Overview

11

Localization - Overview

12

Challenge - Multipath

13

Solving The Multipath ProblemState-of-the-art

14

Model signal on antennas alone Model signal on both antennas and subcarriers

SpotFi

Subc

arrie

rsAntennas

𝒇𝟏𝒇𝟐𝒇𝟑𝒇𝟒

Step 1: Resolve Multipath

15

𝜽𝟏

𝜽𝟐

Signal Modeling

Equal Distance

Line

16

Phase

Distance travelled by the WiFi signal

Phas

e 1 / frequency

0

17

18

Equal PhaseLine

Signal Modeling – AoA (Angle of Arrival)

Signal Modeling - AoA

Define Φ = e

𝜽𝟏

1

Γ is complex attenuation of the path.Φ depends on AoA

Phase at the antenna 1: 𝑥 = Γ

Phase at the antenna 2: 𝑥 = Γ Φ

Phase at the antenna 3: 𝑥 = Γ Φ

19

23

Say There Are Two Paths…

20


21

𝑥 = Γ

𝑥 = Γ Φ

𝑥 = Γ Φ


𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

22

Problem Statement

23

CSI - Known

𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

Problem Statement

24

𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

Parameters - Unknown

Problem Statement

25

Number of paths (or AoAs) < Number of antennas (or equations)

𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

Typical Indoor Multipath

26

That’s A Problem

State-of-the-art Commodity WiFi chips

Number of antennas/equations should be atleast 5

27

How To Obtain More Equations?

28

Model signal on both antennas and subcarriers

Subc

arrie

rs

Antennas

𝒇𝟏𝒇𝟐𝒇𝟑𝒇𝟒

𝒇𝟏

𝒇𝟐

29

Each Subcarrier Gives New Equations

30

Define Ω = e

Γ is complex attenuation of the path.Ω depends on incoming signal ToF

Phase at first subcarrier: 𝑥 = Γ

Phase at second subcarrier: 𝑥 = Γ Ω

Signal Modeling – ToF (Time of Flight)

Estimate both AoA and ToF

31

More number of equations in terms of parameter of our interest


32

At first subcarrier, for 3 antennas

𝑥 = Γ

𝑥 = Γ Φ

𝑥 = Γ Φ

At second subcarrier, for 3 antennas

𝑦 = Γ Ω

𝑦 = Γ Φ Ω

𝑦 = Γ Φ Ω

Say There Are Two Paths…At first subcarrier, for 3 antennas

𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

At second subcarrier, for 3 antennas

𝑦 = Γ Ω + Γ

𝑦 = Γ Φ Ω + Γ Φ Ω

𝑦 = Γ Φ Ω + Γ Φ Ω33

𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

𝑦 = Γ Ω + Γ



Problem Statement

34

Subc

arrie

r 1Su

bcar

rier 2

CSI - Known

Problem Statement

35

Subc

arrie

r 1Su

bcar

rier 2

Parameters - Unknown

𝑦 = Γ + Γ

𝑦 = Γ Φ + Γ Φ

𝑦 = Γ Φ + Γ Φ

𝑦 = Γ Ω + Γ



𝑥 = Γ + Γ

𝑥 = Γ Φ + Γ Φ

𝑥 = Γ Φ + Γ Φ

𝑦 = Γ Ω + Γ



Problem Statement

36

Subc

arrie

r 1Su

bcar

rier 2

Number of equations =

Number of Subcarriersx

Number of Antennas

AoA, ToF Estimates

37

𝜽𝟏, 𝝉𝟏

𝜽𝟐, 𝝉𝟐

Step 2: Identify Direct Path

38

𝜽𝟏, 𝝉𝟏

𝜽𝟐, 𝝉𝟐

𝜽𝟏, 𝝉𝟏

AoA, ToF Estimates

39

𝜽𝟏, 𝝉𝟏

𝜽𝟐, 𝝉𝟐

Use Multiple Packets

40

𝜽𝟏, 𝝉𝟏

𝜽𝟐, 𝝉𝟐


41


42


43

Direct Path Likelihood

Higher weight

Higher weight

Higher weight

Lower weight

Lower weight

44

• Smaller ToF


Higher weight

Lower weight

45

Lower weight

Lower weight

• Smaller ToF

• Tighter Cluster

Lower weight


Higher weight

46

Higher weight

Lower weight

Lower weight

• Smaller ToF

• Tighter Cluster

• More Packets

Lower weight

Highest Direct Path Likelihood

47

9

Step 3: Localize The Target

48

𝜽𝟏, 𝝉𝟏

𝜽𝟐, 𝝉𝟐

𝜽𝟏, 𝝉𝟏

Use Multiple APs

Direct Path AoA = 45 degreesSignal Strength = 10 dB

Direct Path AoA = 10 degreesSignal Strength = 30 dB Find location that best explains

the AoA and Signal Strengthat all the APs

49

Direct Path AoA = -45 degreesSignal Strength = 20 dB

Use Different Weights

Use different weights for different APs

50

Direct Path AoA = 45 degreesSignal Strength = 10 dBDirect Path Likelihood

Direct Path AoA = 10 degreesSignal Strength = 30 dBDirect Path Likelihood

Direct Path AoA = -45 degreesSignal Strength = 20 dBDirect Path Likelihood

Evaluation

51

52 m

Testbed

52

Access point Target

AP Locations Target Locations

0

0.2

0.4

0.6

0.8

1

0.05 0.5 5

Empi

rical

CDF

Localization Error (m)

Indoor Office Deployment

52 m

16 m

0.4 m

53

ArrayTrack Ubicarse SpotFi

0.3 m 0.4 m 0.4 m

AP Locations Target Locations

Stress Test – Obstacles Blocking The Direct Path

54AP Locations Target Locations

52 m

0

0.2

0.4

0.6

0.8

1

0.05 0.5 5

Empi

rical

CDF


Stress Test – Obstacles Blocking The Direct Path

55

1.3 m

AP Locations Target Locations52 m

Effect of WiFi AP Deployment Density

56

0

0.2

0.4

0.6

0.8

1

0.05 0.5 5

Empi

rical

CDF


3 APs4 APs5 APs

0.8 m

Conclusion

• Deployable: Indoor Localization with commercial WiFi chips

• Accurate: Accuracy comparable to state-of-the-art localization systems which are not suitable for wide deployments

• Universal: Simple localization targets with only a WiFi chip

57

References• J. Xiong and K. Jamieson, “Arraytrack: A fine-grained indoor location system,” NSDI ’13. • S. Kumar, S. Gil, D. Katabi, and D. Rus, “Accurate indoor localization with zero start-up cost,” MobiCom ’14. • P. Bahl and V. N. Padmanabhan, “Radar: An in-building rf-based user location and tracking system,”

INFOCOM 2000. • S. Kumar, E. Hamed, D. Katabi, and L. Erran Li, “Lte radio analytics made easy and accessible,” SIGCOMM ’14. • J. Gjengset, J. Xiong, G. McPhillips, and K. Jamieson, “Phaser: Enabling phased array signal processing on

commodity wifi access points,” MobiCom ’14. • M. Youssef and A. Agrawala, “The horus wlan location determination system,” MobiSys ’05. • S. Sen, J. Lee, K.-H. Kim, and P. Congdon, “Avoiding multipath to revive inbuilding wifi localization,” MobiSys

’13. • K. Joshi, S. Hong, and S. Katti, “Pinpoint: localizing interfering radios,” NSDI ’13. • M. Kotaru, K. Joshi, D. Bharadia, S. Katti, "SpotFi: Decimeter Level Localization Using WiFi," ACM SIGCOMM

2015.• All the icons are from the Noun Project https://thenounproject.com/

BackFi: High Throughput WiFi Backscatter for IoT

Dinesh Bharadia*, Kiran Joshi*, Manikanta Kotaru, Sachin KattiStanford University

*co-primary authors

The Internet of Things (IoT) Vision

2


Sense

2

Collect & Analyze


Sense

2

Collect & Analyze Control


Sense

2

Collect & Analyze Control


Sense

BackFi

2

Ubiquitous connectivity

Low power

High uplink rate

Sufficient range

What do we need for IoT Connectivity?

Sense

3

BackFi: Ubiquitous, low power, high throughput connectivity for IoT sensors using ambient WiFi

4

Sense


4

Sense


Decodedata

4

Technical spec Key enabling technique

Same as WiFi Backscatter ubiquitous ambient signals

Less than 50 uW Passive backscatter radios

Up to 6.67 Mbps Maximal ratio combining

Up to 7m Self-interference cancelation

Ubiquitousconnectivity

Low power

High uplink rate

Sufficient range

BackFi’s Contributions

5

Related Work

6

Related Work


Low power

High uplink rate

Sufficient range

6

WiFi Backscatter: H. Ishizaki, et. al. “A Battery-less WiFi-BER modulated data transmitter with ambient radio-wave energy harvesting” B. Kellogg et. al. “Wi-Fi Backscatter: Internet Connectivity for RF-Powered Devices”

WiFi-Backscatter

Related Work


Low power

High uplink rate

Sufficient range

6

RFID-based

WiFi Backscatter: H. Ishizaki, et. al. “A Battery-less WiFi-BER modulated data transmitter with ambient radio-wave energy harvesting” B. Kellogg et. al. “Wi-Fi Backscatter: Internet Connectivity for RF-Powered Devices”RFID based: J.F. Ensworth et. al. “Every smart phone is a backscatter reader”, P. Zhang et. al. “Ekhonet”

WiFi-Backscatter

Related Work


Low power

High uplink rate

Sufficient range

6

RFID-based BackFi’15

WiFi Backscatter: H. Ishizaki, et. al. “A Battery-less WiFi-BER modulated data transmitter with ambient radio-wave energy harvesting” B. Kellogg et. al. “Wi-Fi Backscatter: Internet Connectivity for RF-Powered Devices”RFID based: J.F. Ensworth et. al. “Every smart phone is a backscatter reader”, P. Zhang et. al. “Ekhonet”

WiFi-Backscatter

Related Work


Low power

High uplink rate

Sufficient range

6

BackFi’s Overview

Sense

7

BackFi’s Overview

IoT Sensor

Sense

7

BackFi’s Overview

IoT Sensor BackFi AP

Sense

7

Sense

IoT Sensor Design

8

Sense

IoT Sensor Design

…10101010…

Sensor data

8

Sense

IoT Sensor Design

…10101010…

Sensor data 0 to -11 to +1

8

Sense

IoT Sensor Design

…10101010…

Sensor data Data modulation0 to -11 to +1

8

BackFi AP Design

Sense

9

BackFi AP Design

Sensor&backscatter

Received&signal =&& +Environmental&reflections

Sense

9

Challenge 1: Strong Environmental Reflections

10


-100

-80

-60

-40

-20

0

20

2.45

Pow

er in

dBm

Frequency in GHz

10

40 MHz&


-100

-80

-60

-40

-20

0

20

2.45

Pow

er in

dBm

Frequency in GHz

transmitted&signal

10

40 MHz&


-100

-80

-60

-40

-20

0

20

2.45

Pow

er in

dBm

Frequency in GHz

sensor&backscatter

transmitted&signal

10

40 MHz&


-100

-80

-60

-40

-20

0

20

2.45

Pow

er in

dBm

Frequency in GHz

sensor&backscatter

transmitted&signal

environmental&reflections

10

Sensor&backscatter

Received&signal =&& +Environmental&reflections

40 MHz&

Why not use Self-Interference Cancelation?

11


Transmitted signal

Received signal

11


Transmitted signal

Received signal

Sense

11


Transmitted signal

Received signal

Σ

After&cancelation = 0

Cancelation filter

Sense

11


Transmitted signal

Received signal

Σ


Cancelation filter

Sensorbackscatter

Received&signal = + Environmental&reflections

Sense

11


Transmitted signal

Received signal

Σ


Cancelation filter

Sensorbackscatter



Sense

11

Eliminating environmental reflections

Transmitted signal

Received signal

Σ


Cancelation filter

Sense

Sensorbackscatter


12


Turn off the backscatter

Transmitted signal

Received signal

Σ


Cancelation filter

Sense

Sensorbackscatter


12



Estimate the environmental reflections

Transmitted signal

Received signal

Σ


Cancelation filter

Sense

Sensorbackscatter


12




Transmitted signal

Received signal

Σ


Cancelation filter

Sense

Sensorbackscatter


12

Cancelation filter


After&cancelation = Sensor&backscatter


Turn on the backscatter

Transmitted signal

Received signal

Σ


Cancelation filter

Sense

Sensorbackscatter


12

Cancelation filter

Challenge 2: Inferring IoT Sensor Data

!!!

13


Sensor backscatter is function of:!!!

Sense

13


Sensor backscatter is function of:• Transmitted signal!!

Sense

13


Sensor backscatter is function of:• Transmitted signal• IoT sensor data!

Sense

13


Sensor backscatter is function of:• Transmitted signal• IoT sensor data• Wireless channel distortions

Sense

13

Modeling Sensor Backscatter

Sense

14


Transmitted signal&= :

Sense

14


Transmitted signal&= :

Sense: ∗ ℎ

14


: ∗ ℎTransmitted signal&= :

15



Sensor data = =

15


: ∗ ℎ . = : ∗ ℎTransmitted signal&= :

Sensor data = =

15


Rx = sensor&backscatter =


Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

)∗ ℎ: ∗ ℎ . =(

= =

15




Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

�)∗ ℎ: ∗ ℎ . =(

= =

15

Estimating Backscatter Channel


16



16

Use predefined sequence of sensor data&= to estimate channel



Sensor data = 1

16



: ∗ ℎ . 1 : ∗ ℎTransmitted signal&= :

Sensor data = 1 )∗ ℎ: ∗ ℎRx=(

16




Sensor data = 1

Rx = sensor&backscatter =�: ∗ (ℎ ∗ ℎ)

)∗ ℎ: ∗ ℎRx=(

16




Sensor data = 1

Rx = sensor&backscatter =�: ∗ (ℎ ∗ ℎ)

Estimate h

)∗ ℎ: ∗ ℎRx=(

16





Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(= =

17




Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

�)∗ ℎ: ∗ ℎ . =(

= =

17




Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

�)∗ ℎ: ∗ ℎ . =(

= =

� �

17




Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

�)∗ ℎ: ∗ ℎ . =(

= =

� �Incoming signal z = : ∗ ℎ

17




Sensor data

)∗ ℎ: ∗ ℎ . =Rx=(

�)∗ ℎ: ∗ ℎ . =(

= =

� �Incoming signal z = : ∗ ℎ

sensor&backscatter = )∗ ℎz. =(17

Demodulating = from Sensor BackscatterSensor Backscatter = {E. =} ∗ ℎ

19

Demodulating = from Sensor Backscatter

Incoming z

Sensor Backscatter = {E. =} ∗ ℎ

19


sensor data&=

Incoming z

-1

1


19


sensor data&=

Incoming z

sensorbackscatter

-1

1


19


sensor data&=

Incoming z

sensorbackscatter

-1

1


19

WiFi AP sampling rate

40 Msps

IoT sensor Information switching rate

2 Msps


sensor data&=

Incoming z

sensorbackscatter

-1

1


19

WiFi AP sampling rate

40 Msps

IoT sensor Information switching rate

2 Msps20


sensor data&=

Incoming z

sensorbackscatter

-1

1

redundancy


19


sensor data&=

Incoming z

sensorbackscatter

-1

1

redundancy


19

GH:IJHK&LHMIN&ONJPIQIQR =STUIRℎMV& ∗ VHJWKUV =STX &∗ VX&


sensorbackscatter

redundancy


19

GH:IJHK&LHMIN&ONJPIQIQR =STUIRℎMV& ∗ VHJWKUV =STX &∗ VX&

BackFi Prototype

24

BackFi Prototype

24

Modulator

Antenna

Digital control board

WiFi Backscatter radio with BPSK, QPSK & 16 PSK

BackFi Prototype

24

Modulator

Antenna



PA

TX RX

Cancelation filter

Antenna

BackFi Prototype

24

Modulator

Antenna



PA

TX RX

Cancelation filter

Antenna

Built using WARP SDR platform, designed for 802.11, BW 20MHz, 20dBm TX power

BackFi Prototype

24

Modulator

Antenna


BackFi Prototype

25

Testbed & Performance Metrics

!

!

!

!!

6m

8m

26


Indoor office environment:• AP and IoT sensor are

placed in LOS!

!

!!

6m

8m

26



placed in LOS• WiFi clients are placed

nearby!

!!

6m

8m

26




nearby• Varied the placement of IoT

device, client and WiFi AP.

!!

6m

8m

26




nearby• Varied the placement of IoT

device, client and WiFi AP.

Performance metrics• Throughput• Energy per bit

6m

8m

26

What is the range and throughput ?

012345678

0.5 1 2 4 5 6 7

Thro

ughp

ut in

Mbp

s

Range in meters

27

Throughput of IoT sensor with distance from AP


012345678

0.5 1 2 4 5 6 7

Thro

ughp

ut in

Mbp

s

Range in meters

100Kbps

27


Three order of magnitude better throughput than prior WiFi backscatter


012345678

0.5 1 2 4 5 6 7

Thro

ughp

ut in

Mbp

s

Range in meters

100Kbps

27


Throughput in Mbps

EPB inpJ/bit

Total Power Consumption in uWfor continuous mode

.1 12.66 1.27

.5 5.04 2.52

1 4.10 4.10

2 3.62 7.24

6.67 5.97 39.92

What is the power consumption of BackFi?

28

Throughput in Mbps

EPB inpJ/bit

Total Power Consumption in uWfor continuous mode

.1 12.66 1.27

.5 5.04 2.52

1 4.10 4.10

2 3.62 7.24

6.67 5.97 39.92

What is the power consumption of BackFi?

Two order magnitude better EPB than prior work 28

!

!

!

Conclusion

29

BackFi provides high throughput, low power, ubiquitous connectivity using ambient WiFi signals

!

!

!

Conclusion

29


• Not restricted to WiFi, can use other ambient signals such as LTE, Bluetooth

!

!

Conclusion

29



• Vision: Build a pervading layer of connectivity over all ambient communication signals

!

Conclusion

29



• Vision: Build a pervading layer of connectivity over all ambient communication signals

• Next step: go from a link to a network

Conclusion

29

spotfi: decimeter level localization using wifinetseminar.stanford.edu/seminars/10_15_15.pdf ·...

Documents