Docstoc

Wi-Fi Assisted Multi-sensor Personal Navigation System for Indoor

Document Sample
Wi-Fi Assisted Multi-sensor Personal Navigation System for Indoor Powered By Docstoc
					GEOIDE Workshop

   Wi-Fi Assisted Multi-sensor
 Personal Navigation System for
      Indoor Environments




                       Xing Zhao, Dr. Naser El-Sheimy
                  Mobile Multi-Sensor System Research Group
                         University of Calgary, Canada

                          Dr. Aboelmagd Noureldin
                        Royal Military College , Canada


                                June 14, 2010
    Table of Contents

      Background: opportunities and challenges for
       personal navigation systems

      Indoor Radio Propagation Modeling

      Wi-Fi Positioning Approach

      Wi-Fi/GPS/IMU/Mag Integrations

      Field Test

                                                                          m
2                        Mobile Multi-Sensor Systems Research Group   m
                                                                          s
                                                                              s
    Problem Statement




                  Indoors and dense urban
                  areas are the biggest
                  challenges for GNSS

                  Where signals < -160dBm
                  Better than 50m accuracy

                  Portable, reliable, accurate
                  and low-cost personal
                  navigator.
                                                                     m
3                   Mobile Multi-Sensor Systems Research Group   m
                                                                     s
                                                                         s
    Candidate Positioning Technologies
       Other GNSS
       Vision Hybrids
    -   Camera/laser
       Radio Hybrids
    -   Cellular
    -   TV/RFID/UWB/Bluetooth
    -   Wi-Fi
       Sensor Hybrids
    -   Gyro and accelerometers
    -   Magnetometer, barometers, etc…

                                                                         m
4                       Mobile Multi-Sensor Systems Research Group   m
                                                                         s
                                                                             s
Taking Advantage of Low-cost Radio
and Sensors
     Wi-Fi has reached ubiquitous levels in devices such
      as laptops , netbooks, and smartphone

     Wi-Fi hotspots (>100 million) have good coverage,
      some with known coordinates
       GPS + Wi-Fi


     Further integrate self-contained sensors
       GPS+ Wi-Fi +sensors (IMU/Mag)


                                                                         m
5                       Mobile Multi-Sensor Systems Research Group   m
                                                                         s
                                                                             s
    Wi-Fi Access At University of Calgary

      An AP with known position broadcasts its service
       set identifier (SSID) every 100 ms in the 2.4 GHz
       frequency band.

      >1280 access points all
      over the school (2008,Oct)
      Aruba AP 70 access point
     - 802.11a/b/g max TX  Rx Sen.
       54Mbps +17.0dBm -73.0dBm


                                                                        m
6                      Mobile Multi-Sensor Systems Research Group   m
                                                                        s
                                                                            s
    The Prototyping Test System


                                       Laptop w/ WLAN card
                                       Garmin CS60X GPS
                                       ADI ADIS16405 IMU
                                      - Tri-axis, digital gyroscope
                                         Tri-axis, accelerometer
                                         and magnetometer




                                                                       m
7                   Mobile Multi-Sensor Systems Research Group     m
                                                                       s
                                                                           s
Wi-Fi Data Collection

                                          Apply NetStumbler
                                           to collect received
                                           signal strength
                                           with GPS




                                                                   m
8                 Mobile Multi-Sensor Systems Research Group   m
                                                                   s
                                                                       s
    Wi-Fi Positioning: Fingerprinting
    Approach
                          K
       Dn =| Vn − Vu |=   ∑ (Vn − Vu )
                          i =1
                                  i    i
                                           2


       Vn : Signal strength vector at the surveyed positions
       Vu: Signal strength vector at the user positions
       K: number of effective access points,
       n : number of surveyed positions in the database with the
       known coordinates.

       - Approximate user nth position with pre-surveyed position
       when Dn is min.
       - e.g. Skyhook deploys a fleet of data collection vehicles to
       conduct a comprehensive access point survey

      Technically renders best performances but
       practically not quite economical, scalable
                                                                                  m
9                                Mobile Multi-Sensor Systems Research Group   m
                                                                                  s
                                                                                      s
 Wi-Fi Positioning: RSS Modeling
 Approach




       PL(d ) = A + B log(d )

                                                                    m
10                 Mobile Multi-Sensor Systems Research Group   m
                                                                    s
                                                                        s
 RSS Impact Factors
    Physical obstruction
    Tx power / Rx front-end
    Antenna orientation
    Signal fading (Doppler, Multipath etc.)
    Radio interference

 -> ~3dB STD in signal strength variations




                                                                        m
11                     Mobile Multi-Sensor Systems Research Group   m
                                                                        s
                                                                            s
 LOS Wireless Propagation Modeling




  LOS
                                                                 m
12              Mobile Multi-Sensor Systems Research Group   m
                                                                 s
                                                                     s
 NLOS Radio Propagation Modeling
                                               WiFi Propogation Modeling (NLOS)
                                   100




                  Path Loss (dB)
                                   90    y = 2.7*x + 37


                                   80

                                   70


                                   60

                                   50

                                   40
  Hidden above the
   ceiling, NLOS                   30
                                     0             5               10            15   20
                                                          10*log(distance) (m)
                                                                                          m
13                                   Mobile Multi-Sensor Systems Research Group       m
                                                                                          s
                                                                                              s
 Other Testing Scenarios




     Open, LOS                      Narrow, NLOS

                                                                  m
14               Mobile Multi-Sensor Systems Research Group   m
                                                                  s
                                                                      s
 Summary Table on Propagation
 Modeling

                       P1              P2              P3              P4

                                                                     Narrow
      Scenario    Wide Hallway    Lecture Room    Above Ceiling
                                                                     Corridor
      Position      ENE238Z         ENE239         ENE228ZS         ENF212
       Type           LOS             LOS            NLOS            NLOS


     Path Loss                                    2.7*10logd+3
                  2.3*10logd+34   2.6*10logd+41                   2.6*10logd+40
     Model (dB)                                           7



          Compromised Model: 2.6*10log(d)+40
          Determine maximum range (with acceptable path loss) ~40m
          Reliable Positioning range ~10m

                                                                                    m
15                            Mobile Multi-Sensor Systems Research Group        m
                                                                                    s
                                                                                        s
 Wi-Fi Positioning Algorithm
  Define a reliable received signal strength threshold
   for received AP. (SNR 49dB~ 10m )

  Estimate the user position with based on the known
   AP coordinates (when # of AP=1)

  Weighted the estimated positions based on SNR
   (when # of AP>1)

  Assign proper STD of the position errors from SNR
                                                                      m
16                   Mobile Multi-Sensor Systems Research Group   m
                                                                      s
                                                                          s
 Field Test Introduction

  Tested at Engineering build, Univ. of Calgary in Jan
   2010
  ~10mins walk
 ~ 280 steps outdoors
 ~ 335 steps indoors
  Bad GPS
  Many Wi-Fi hotpots (>250 APs but only choose 12
   APs (6 LOS) for positioning)


                                                                      m
17                   Mobile Multi-Sensor Systems Research Group   m
                                                                      s
                                                                          s
 Field Test- Reference Trajectory




                                                    Backward
                                                    smoothing:
                                                    accuracy <
                                                    4m




                                                                   m
18                Mobile Multi-Sensor Systems Research Group   m
                                                                   s
                                                                       s
 GPS-Only Solution




                                                                 m
19              Mobile Multi-Sensor Systems Research Group   m
                                                                 s
                                                                     s
 Chosen Access Points




                                                                 m
20              Mobile Multi-Sensor Systems Research Group   m
                                                                 s
                                                                     s
 Selected Access Points for Positioning
                                                                   Max SNR
           Mac Addr.        Room/AP Type   E(m)    N(m)     Z(m)
                                                                    (dB)
 AP1    00:0B:86:C8:8E:60   ENC201/NLOS    -2.5    16.5     7.6      69
 AP2    00:0B:86:CA:B0:60   ENC210Y/LOS    -9.6     5.8     4.4      64
 AP3    00:0B:86:CA:DF:20   ENE239/NLOS    -25     -16.8    8.0      49
 AP4    00:0B:86:CA:D8:60   ENE241/NLOS    -24.8   -33.6    8.0      52
 AP5    00:0B:86:CE:A5:E0   ENE238Z/LOS    -16.1   -42.4    7.9      59
 AP6    00:0B:86:CB:C5:C0   ENE229Z/NLOS   -7.4    -68.5    7.5      64
 AP7    00:0B:86:CF:9C:E0   ENE228/NLOS    -17.1   -76.4    7.5      62
 AP8    00:0B:86:D6:D4:80   ENE227Z/NLOS   -7.4    -82.2    7.5      61
 AP9    00:0B:86:D0:A5:41   ENE228S/NLOS   -5.4    -91.2    8.1      64
 AP10   00:0B:86:D6:90:21   ENE221Z/LOS    -13.4   -92.2    7.8      56
 AP11   00:0B:86:CE:62:62   ENF212/NLOS    -35.1   -99.7     7       54
 AP12   00:0B:86:D0:B4:C2   ENF275/NLOS    -36.3   -110.3    7       52

                                                                              m
21                           Mobile Multi-Sensor Systems Research Group   m
                                                                              s
                                                                                  s
     GPS + Wi-Fi after innovation testing
      GPS only
      Wi-Fi only                                 •65% available
                                                 •Cannot
                                                 distinguish
                                                 rooms, turns or
                                                 floors




                                                                    m
22                 Mobile Multi-Sensor Systems Research Group   m
                                                                    s
                                                                        s
 The Multi-sensors Integration


                                       Step        Step
                            Heading   Detection   Length



                                  Pedestrian       P
                                Dead Reckoning
     Accel.                                                                               GPS
              Measurement




                                                                            Measurement
               Validation




                                                                             Validation
                                                              EKF       P
     Gyro
                                                          Integration                     Wi-Fi
                                                  P,V,A
     Mag.                        Mechanization




                                                                                                      m
23                                    Mobile Multi-Sensor Systems Research Group                  m
                                                                                                      s
                                                                                                          s
 Real-time and backwards smoothed
 solutions
      Post-mission                Next Slide
      Real-time
                                                   • Real-time
                                                   accuracy < 13m
                                                   (mean = 5m).
                                                   • Heading
                                                   accuracy < 20
                                                   degrees all the
                                                   time




                                                                      m
24                   Mobile Multi-Sensor Systems Research Group   m
                                                                      s
                                                                          s
 Finding The Lost Information Indoors

                                               Post-mission   •8 turns missed
                                               Real-time      by GPS/Wi-Fi
                                               GPS only       •All turns
                                               Wi-Fi only     captured by real-
                                                              time and post-
                                                              mission
                                                              solutions
                         Stairs up

     2nd floor hallway




                                            Entering
                                            building

                                                                                 m
25                              Mobile Multi-Sensor Systems Research Group   m
                                                                                 s
                                                                                     s
 Conclusion
  Wi-Fi /sensors integration has great potential for
   mass mobile devices

  Wi-Fi positioning based on radio propagation
   models can circumvent pre-survey

  Implement EKF to integrate gyro/magnetometer
   Wi-Fi with GPS

  For a decent AP environment, ~10m accuracy can
   be obtained
                                                                      m
26                   Mobile Multi-Sensor Systems Research Group   m
                                                                      s
                                                                          s
 Acknowledgments
  This study was supported in part by:
        Geomatics for Informed Decisions (GEOIDE)
        Research fund from the Tecterra
        Natural Science and Engineering Research Council of Canada (NSERC)
        Network Centers of Excellence (NCE) to Dr. Naser El-Sheimy


  Dr. Goodall and Dr. Syed for the technical support



              Thanks for your attention!


                                                                              m
27                           Mobile Multi-Sensor Systems Research Group   m
                                                                              s
                                                                                  s

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:21
posted:2/16/2011
language:English
pages:27