Personal Navigation System

1
Personal Navigation System
Jari Saarinen, Jussi Suomela, Seppo Heikkilä,
Mikko Elomaa and Aarne Halme Helsinki University
of Technology Automation Technology
laboratory Finland
Project funded by the European Community under
the IST programme Future and Emerging
Technologies
2
Partners

• CTU Czech Technical University, Gerstner
  Laboratory (CZ) Coordinator
• Certicon CertiCon a.s. (CZ)
• HUT Helsinki University of Technology,
  Automation Technology Lab.(FI)
• JMUW Bayerische Julius-Maximilians Universität
  Würzburg (DE)
• ARS Steinbeis GmbH Co. für Technologietransfer
  (DE)

3
IntroductionPeLoTe Building Presence through
Localization for Hybrid Telematic Systems

• Scenario
• Human and Robotic entities (HE, RE) explore
  common area with the help of operator
• Both provide continuous mapping data from
  environment
• Mapping information is processed to a common
  presence for both entities
• Case example Fire fighting

4
Related Work

• Infrastructure based tracking for indoors
• Ekahau Positioning Engine (WLAN)
• PinPoint Asset Tracking System (RFID)
• Space Systems Finland Pseudolite GPS
• GSM based localisation
• Standalone systems
• Pedometers (distance/steps taken)
• PointResearch DRM-III (Position based on steps
  and compass)
• Nike Triax V-10 (Speed and distance)

5
Personal Navigation System - PeNa

• Standalone dead reckoning system for human indoor
  positioning
• Stride length measurement unit
• Compass
• Gyro
• Laser odometry
• Laser also used for metric mapping and a priori
  map localisation
• WLAN, UI, Power, Camera, Audio

6
Heading Estimation

• Absolute heading from compass (3DM-G)
• Fibre optic gyro for rapid movements
• The compass is very sensitive to electrical and
  magnetic fields in indoors
• The gyro drifts approx. 2 deg/min
• Fusion with standard Kalman filter

7
Step Length estimation

• SiLMU (Stride length measuring unit, lab-made)
• US based ankle distance meter
• 60Hz continuous measurement
• Microcontroller calculates the stride based on
  Time-Of-Flight
• Provides accurate information of the leg movement
  
• Used with heading information the result is
  comparable to the odometry

8
Using Laser For Localisation

• The laser scan matching provides not only
  information about movement but also environmental
  data (Map)
• Laser odometry provides accurate estimate of
  movement
• Main problems are swinging and placement in human
  body, especially floor and ceiling echoes

9
Correlation Algorithm for Scan Matching

• Generate a set of poses, relative to reference
  scan.
• Transform current scan according to pose.
• For every point in current scan find nearest
  neighbour in reference scan. If neighbour closer
  than a threshold increase hit count.
• After going through all the poses, select the
  pose that has the biggest hit value.
• Improvements
• Coarse-to-fine search
• Not changing the reference scan while there is
  enough information between current scan and
  reference scan
• Least squares estimation for best estimate

10
Dead Reckoning results

• Test result using the integrated system
• Green path is integrated using SiLMU and Heading
• Red path is integrated position after scan
  matching
• Total Path 404,4m
• Time Elapsed 475,5s
• End-Start Error 9,9m
• Relative Error 2,45
• Heading error 18 deg

11
Dead Reckoning results - II
12
Current Development

• Faster and more robust Scan Matching
• Angle histogram matching
• Better 2D correlation for position
• Map Based methods
• Using Correlation
• Monte Carlo Localisation
• SLAM
• Cooperative Localisation
• Using beacons
• Operator assisted localisation
• Integration to whole system

13
Current Development - II
SLAM
Operator Tools
MCL
Testing (with end users)
Beacon system
14
Conclusions and future work

• Standalone indoor localisation system for human
  was presented
• Dead reckoning using SiLMU, compass, gyro and
  laser
• Results are similar to dead reckoning of robots
• Used already for map based localisation and
  initial tests with SLAM
• Currently the whole system is in integration
  phase
• First results with integration in the end of
  October
• End user testing in the end of November

15

• PENA VIDEO

16
(No Transcript)
17
(No Transcript)