Enhancing Indoor Inertial Odometry with WiFi Raghav H. - - PowerPoint PPT Presentation

β–Ά
enhancing indoor inertial odometry with wifi
SMART_READER_LITE
LIVE PREVIEW

Enhancing Indoor Inertial Odometry with WiFi Raghav H. - - PowerPoint PPT Presentation

Mar 18, 2023 771 likes β€’1.05k views

Enhancing Indoor Inertial Odometry with WiFi Raghav H. Venkatnarayan, Muhammad Shahzad NC State University, Raleigh, USA UbiComp 2019 Outline 1. Background 2. Motivation 3. Technique 4. Implementation and Evaluation 5. Conclusion 2 of 26

slide-1
SLIDE 1

Enhancing Indoor Inertial Odometry with WiFi

Raghav H. Venkatnarayan, Muhammad Shahzad NC State University, Raleigh, USA UbiComp 2019

slide-2
SLIDE 2

Outline

  • 1. Background
  • 2. Motivation
  • 3. Technique
  • 4. Implementation and Evaluation
  • 5. Conclusion

2 of 26

slide-3
SLIDE 3
  • 1. Background : Inertial Odometry

Odometry : Estimating change in position over time i.e distance Robotics UAVs Fitness VR Several Applications

3 of 26

slide-4
SLIDE 4
  • 1. Background : Inertial Odometry

Inertial Odometry : Odometry using IMUs (Accelerometer + Gyro)

  • Power Efficient
  • Ubiquitous
  • Inexpensive (~2 USD)
  • Scalable

Acceleration Rotation

4 of 26

slide-5
SLIDE 5
  • 1. Background : Inertial Odometry

Inertial Odometry : Odometry using IMUs (Accelerometer + Gyro)

Accelerometer Gyroscope Axis Transformation Orientation Linear Acceleration

ΰΆ± ࢱ𝑒𝑒

Accn.. Position

Error grows cubically in time Solution : Sensor Fusion (e.g GPS)

5 of 26

slide-6
SLIDE 6
  • 1. Background : Inertial Odometry

Error grows cubically in time Outdoors : Sensor Fusion (GPS + IMU) Indoors :

  • PDR

➒ Limited to Step Counts ➒ Learning Stride Lengths ➒ Only Humans

  • Other Modalities (IR, Ultrasound, Vision, LIDAR)

➒ Limited range or LoS only ➒ Reduced Ubiquity ➒ Inconsistent indoor localization accuracy

Is there a more ubiquitous modality for accurate indoor inertial odometry ?

6 of 26

slide-7
SLIDE 7
  • 2. Motivation : WiFi assisted Inertial Odometry
  • Most handhelds : IMU + WiFi NIC.
  • WiFi Communication:

– Power Efficient – Ubiquitous

  • Measurements from WiFi communication : CSI
  • Device Motion => Doppler Shift in CSI
  • CSI =>Doppler Shift => Device Speed => IMU Fusion

Doppler Shift

Challenge : Doppler Shift β‰  Device Speed

CSI : Change in Amplitude + Phase CSI : 0.01β…‡40Γ—2𝜌

Dist. Ampl.

7 of 26

slide-8
SLIDE 8
  • 2. Motivation : WiFi assisted Inertial Odometry

Problem Statement : Derive speed from the Doppler Shifts in WiFi signals from a single AP to correct the drift errors in inertial odometry Requirements:

  • 1. Not require fingerprinting
  • 2. Commodity WiFi Devices
  • 3. Resilient to background human movements
  • 4. Single AP, no hardware/firmware modifications
  • 5. Deployable on robots and humans

8 of 26

slide-9
SLIDE 9
  • 3. Technique : Overview

Idea : Measure device movement speed from WiFi channel measurements and correct IMU Speed Drift

Sensor Fusion Speed Estimation from IMU Speed Estimation from WiFI CSI Distance

CSI Acc.

4 key insights

9 of 26

slide-10
SLIDE 10
  • 3. Technique : WiFi CSI as a speed sensor

Δ𝑒 1 2 3 1 Device moves Δ𝑒 in time Δ𝑒 2 𝑀0 : Signal Path Length @ 𝑒 = 0 3 𝑀1 : Signal Path Length @ 𝑒 = Δ𝑒 Path Length Change Speed : π’˜ =

𝑀1βˆ’π‘€0 Δ𝑒

𝐡 βˆ— cos 2𝜌 π’˜Ξ”π‘’ 𝑑/𝑔 + 2πœŒπ‘€0 𝑑/𝑔 + πœ’π‘‘π‘™ CSI Power :

Insight 1 : Path Length Change => Sinusoid in CSI Power

10 of 26

slide-11
SLIDE 11
  • 3. Technique : WiFi CSI as a speed sensor

Δ𝑒 4 2 5 4 Device moves Δ𝑒 in time Δ𝑒 2 𝑀0β€² : Signal Path Length @ 𝑒 = 0 5 𝑀1β€² : Signal Path Length @ 𝑒 = Δ𝑒 Path Length Change Speed : π’˜β€² =

𝑀1β€²βˆ’π‘€0β€² Δ𝑒

𝐡′ βˆ— cos 2𝜌 𝑀′Δ𝑒 𝑑/𝑔 + 2πœŒπ‘€0β€² 𝑑/𝑔 + πœ’β€²π‘‘π‘™ CSI Power :

Insight 2 : Different Multipaths => Different sinusoids in CSI Power

11 of 26

slide-12
SLIDE 12
  • 3. Technique : WiFi CSI as a speed sensor

Δ𝑒

Insight 3 : If 𝛦𝑒 << length of all 𝑙 multipaths, path length change speed => a relation of 𝛦𝑒

Δ𝑒 90 βˆ’ πœ„π‘™ 90 βˆ’ πœ„π‘™ = 𝜚k Path Length Change Speed : 𝑀𝑙 =

𝑀0

π‘™β€²βˆ’π‘€0 𝑙′

Δ𝑒

=

𝜠 β…† 𝒅𝒑𝒕 πœΎπ’ πš¬π’–

𝑀0

𝑙

𝑀1

𝑙

12 of 26

(𝛦𝑒 <<𝑀0

𝑙) ^ (𝛦𝑒 <<𝑀1 𝑙) βˆ€π‘™

𝜚k Δ𝑒

slide-13
SLIDE 13
  • 3. Technique : WiFi CSI as a speed sensor

Insight 3 : Freq. of sinusoid => 𝑔 ( 𝛦𝑒 , cosπœ„π‘™ ) Insight 1 : Path Length Change => Sinusoid in CSI Power Insight 2 : Different Multipaths => Different sinusoids in CSI Power Challenging to accurately find πœ„π‘™ on Commodity WiFi!

13 of 26

𝜠 β…† 𝒅𝒑𝒕 πœΎπ’ πš¬π’–

slide-14
SLIDE 14
  • 3. Technique : WiFi CSI as a speed sensor

Δ𝑒

Insight 4 : Multipath k most parallel to the direction of motion i.e πœ„π‘™ = 0 or πœ„π‘™ = Ο€ => highest path length change speed

Δ𝑒

𝑀𝑙 =

𝑀0

π‘™β€²βˆ’π‘€0 𝑙′

Δ𝑒

= 𝜠 β…† 𝒅𝒑𝒕 𝟏

πš¬π’–

Freq (Highest Frequency Sine)Γ—Wavelength β‰ˆ Device Speed

𝑀𝑙 β‰ˆ 𝐺𝑙𝝁 => πœ β…† β‰ˆ π’˜π’πœ π’– => πœ β…† β‰ˆ πΊπ‘™ππœ π’–

14 of 26

𝝁 = πŸ”. πŸ‘π’…π’ @πŸ”. πŸ—π‘―π’Šπ’œ!

slide-15
SLIDE 15
  • 3. Technique : WiFi CSI as a speed sensor

Putting it all together: Every πš¬π’– : 1) CSI Power Time Series ∢ * Noise & Human interference removal 2) STFT : 3) WiFi Speed = 1.0166 𝐺𝑙𝝁 e.g 1.0166 * 5 hz* 5.2cm = 26.413 cm/s

15 of 26

slide-16
SLIDE 16
  • 3. Technique : WiFi CSI as a speed sensor

1) Bias Computation 2) Bias Elimination 3) IMU Speed = 𝑀𝑦

2 + 𝑀𝑧 2 + 𝑀𝑨 2

Kalman Filter 1) Process Var : IMU Speed 2) Measurement Var : CSI Speed 3) Compute optimal middle ground estimate

16 of 26

slide-17
SLIDE 17
  • 4. Implementation and Evaluation

Testing Platform : Custom Handheld device ( 10cm x 15cm x 5cm box ) Inside : HummingBoard Pro running Ubuntu 14 + Intel WiFi Chipset Outside: Rear : 3 Omnidirectional Antennas (HalfWave ULA) Front : Arduino Uno + Invensense MPU-6050 IMU + USB

17 of 26

πœ„π‘™

slide-18
SLIDE 18
  • 4. Implementation and Evaluation

Deployments : Humans ( 4M, 2F) + Drone

Drone with Vive Tracker

18 of 26

slide-19
SLIDE 19
  • 4. Implementation and Evaluation

Environments : 4

19 of 26

slide-20
SLIDE 20
  • 4. Implementation and Evaluation

Environments : 4 Other humans

20 of 26

Platform

slide-21
SLIDE 21
  • 4. Implementation and Evaluation

Evaluation Metric : RO Error = Estimated Distanceβˆ’Actual Distance

Actual Distance

  • IMUDR

Distance computed from IMU double integration

  • WIOSpotFi

Distance computed from Most Parallel Path using a state-of-the-art SuperResolution AoA Method (πœ„π‘™ Insight 3)

  • WIO

Distance computed from Insight 4 (HF sinusoid)

21 of 26

slide-22
SLIDE 22
  • 4. Implementation and Evaluation
  • 1. Human Deployments

Straight Paths ROE over time 6% Curved Paths 5% Observed Speeds Changing Speeds 6% <-50dBm AP Placements

22 of 26

70% 42%

slide-23
SLIDE 23
  • 4. Implementation and Evaluation
  • 1. Human Deployments

Multiple Human Odometry 8% 6,7,7,8% 55cm (NLoS) Error across Envs. Gyro Drift

23 of 26

slide-24
SLIDE 24
  • 4. Implementation and Evaluation
  • 2. Drone Deployment

Straight Paths 4% Curved Paths 6% 5% 5-10% AP Placements Changing Speeds

24 of 26

slide-25
SLIDE 25
  • 5. Conclusion
  • Proposed a novel WiFi-assisted inertial odometry technique
  • The key novelty of using the WiFi signals as the auxiliary source of information that

works in indoor environments, w/o fingerprinting, and resilient against changes in environment

  • Median RO error of just 6.87% and 5.7% respectively for human subjects and a

drone across all scenarios, and at least 3x more accuracy compared to pure Inertial Odometry

25 of 26

slide-26
SLIDE 26

Thank You!

26 of 26