Post on 07-Aug-2020
INDOOR POSITIONING VIA SMARTPHONE SENSING
Associate Professor Tao Gu, RMIT University, Australia
Outline
• Indoor Localization using smartphone magnetometer
• Floor localization
• Localization in Metro Trains
Applications • Shopping malls, office buildings, airports, train
stations, hospitals, trade shows, museums, warehouse, etc.
Smartphone Built-in Sensors
• Use smartphone sensors • magnetometer
• accelerometer
• compass
• gyroscope
• Do not rely on Wi-Fi
• Battery efficiency
• Accurate (1~2.8m )
A Magnetic Fingerprinting Approach to Indoor Localization
Magnetic Fingerprint
Magnetometer Readings
𝑩 = [𝐵𝑥, 𝐵𝑦, 𝐵𝑧] 40 43 44 45 49 48
44 43 46 47 51 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
Magnetic Fingerprint: Not Unique!
||𝑩||
Particle Filter
•Particle Motion Tracking
•Particle Weight Evaluation
•Particle Resampling
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
--
0
-90
Heading changes
Magnetic readings
Initialization
Default Particle Step Length=1
Motion Tracking & Weight Estimation
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
--
0
-90
Heading changes
Magnetic readings
1
1
0
0
0
0
0
0 0
0
0
0
Resampling
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
42
--
0
-90
Heading changes
Magnetic readings
Motion Tracking & Weight Estimation
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
--
0
-90
Heading changes
Magnetic readings
1
0
Resampling
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
--
0
-90
Heading changes
Magnetic readings
Get you!
P1: Particle Step Length & User Step Length
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
47
--
0
-90
0
Heading changes
Magnetic readings
Default Particle Step Length=1
User Actual Step Length=2
User Actual
Position
Estimated Position
P2: Localization Failure - Kidnapped Robot Problem
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
47
--
0
-90
0
Heading changes
Magnetic readings
We cannot realize the localization has failed until all particles’ weight become 0.
User Actual
Position
Estimated Position
1
P3: Heading Offset Change
40 43 44 45 49 48
44 43 46 47 47 47
40 45 48 50 52 48
39 42 44 47 49 43
37 38 42 43 45 40
35 36 40 41 41 38
45
43
46
47
47
--
0
-90
0
180
Heading changes
Magnetic readings
User’s Heading direction
Phone’s Heading direction
User answer a Call
User Actual
Position
Estimated Position
P4: Calibrating Different Smartphones ?
Our Contributions
• Dynamic step length estimation
• Kidnapped robot problem
• Heading offset change
• Calibrating different smartphones
Real Experiments in an Office Building
Accuracy
1-2.8 m
Smartphone Battery
Floor Localization
• Motivation: • Do not rely on Wi-Fi
• Leverage on smartphone sensing and crowdsourcing
• Applications: • Location based services
• Emergency services
Traveling Time
Step Count
Acceleration
Bluetooth
Encounter
Acceleration
User Trace
Cloud Server
Stairs
From Floor
To Floor
Traveling Time
Step Count
1 2 8 s 22
1 3 12 s 44
Elevator
Mapping table
Download
User Trace
From Floor
To Floor
Traveling Time
Step Count
1 2 8 s 22
1 3 12 s 44
The Initial floors of users
Escalators
Localization in Metro Trains
• Problem definition: how to locate mobile users in a metro train?
• Motivation: • Do not rely on Wi-Fi
• No need to war-drive all the metro lines
• Smartphone sensing • accelerometer
• magnetometer
• barometer
• Crowdsourcing
Structure Infrastructure Movement
Metro Line
Traditional Indoor Spaces
Different from Traditional Indoor Spaces
Limited infrastructure
Low speed Random
High speed Predictable
Abundant
Simple
Complex
11/2/2015
Logging accelerometer, barometer reading. Do train stop event Recognition.
Localization map
Trace merging Pattern map generation
Cloud Server
Metro Train
A B C
Crowdsourcing
User traces 1
location
Stop event Recognition. Running event matching.
2
More Information • My website
• https://sites.google.com/site/gutao98/home
• Related Publications • Hongwei Xie, Tao Gu, Xianping Tao, and Jian Lu. A Reliability-Augmented Particle Filter for
Magnetic Fingerprinting based Indoor Localization on Smartphone, IEEE Transaction on Mobile Computing (TMC), 2015.
• Hongwei Xie, Tao Gu, Xianping Tao, and Jian Lu. MaLoc: A Practical Magnetic Fingerprinting Approach to Indoor Localization using Smartphones, In Proc. of the 2014 ACM International Joint Conference on Pervasive and Ubiquitous Computing (UbiComp 2014), Seattle, Washington, September 13-17, 2014.
• HaiboYei, Tao Gu, Xianping Tao, and Jian Lu. SBC: Scalable Smartphone Barometer Calibration through Crowdsourcing, In Proc. of the 11th International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services (MobiQuitous 2014), London, December 2-5, 2014.
• HaiboYei, Tao Gu, Xianping Tao, and Jian Lu. B-Loc: Scalable Floor Localization using Barometer on Smartphone, In Proc. of the 11th IEEE International Conference on Mobile Ad hoc and Sensor Systems (MASS 2014), Philadelphia, Pennsylvania, October 28-30, 2014.
• HaiboYei, Tao Gu, Xianping Tao, and Jian Lu. Crowdsourced Smartphone Sensing for Localization in Metro Trains, In Proc. of IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks (WoWMoM 2014), Sydney, Australia, June 16-19, 2014.
Opportunities
• R&D Centers • Joint research and development
• Joint funding
• Industry Partners • Commercial Product development
• Australian Government Funding • Collaboration between an Australian university and an industry
partner (local or overseas).
• http://www.arc.gov.au/linkage-projects
• Venture Capital for creating a start-up
11/2/2015