Cooperative Localization Algorithms for Improved Road Navigation - - PowerPoint PPT Presentation

cooperative localization algorithms for improved road
SMART_READER_LITE
LIVE PREVIEW

Cooperative Localization Algorithms for Improved Road Navigation - - PowerPoint PPT Presentation

TECHNISCHE Professorship of UNIVERSITT Communications Engineering CHEMNITZ Cooperative Localization Algorithms for Improved Road Navigation Eric Richter , M. Obst, V. Leonhardt, R. Schubert, N. Mattern, and G. Wanielik Chemnitz University


slide-1
SLIDE 1

TECHNISCHE UNIVERSITÄT CHEMNITZ

Eric Richter eric.richter@etit.tu-chemnitz.de Professorship of Communications Engineering

Chemnitz University of Technology, Germany

Cooperative Localization Algorithms for Improved Road Navigation

11/14/08 1

Eric Richter, M. Obst, V. Leonhardt, R. Schubert, N. Mattern, and G. Wanielik

slide-2
SLIDE 2

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 2

Cooperative Localization - Introduction

  • Using smoothed pseudo range

double differences or

  • ther sensors like radar or laser
  • Intended to be used in

automotive field (but not limited to it)

  • e.g. helps for

lane precise localization

GPS Sat. GPS Sat. GPS Sat.

slide-3
SLIDE 3

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 3

Distributed Data Fusion - Example

slide-4
SLIDE 4

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 4

Distributed Data Fusion - Example

… …

slide-5
SLIDE 5

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 5

Centralized Data Fusion – One way out

Additional infrastructure needed! Central Data Fusion Unit Estimation

slide-6
SLIDE 6

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 6

Distributed Data Fusion – Covariance Intersection

slide-7
SLIDE 7

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 7

Distributed Data Fusion - Example

Common Information

slide-8
SLIDE 8

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 8

Distributed Data Fusion – Removing Common Information

  • Data fusion of remote and local information considering common information
slide-9
SLIDE 9

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 9

Cooperative Localization – The System

slide-10
SLIDE 10

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 10

Cooperative Localization –Network Topology

… … … … … … …

Tree structured network architecture tracks the common information between two nodes

slide-11
SLIDE 11

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 11

Cooperative Localization – Estimated State

  • State space consists of the vehicle’s position and heading
  • non-linear state transition
  • Movement model

Constant Turn Rate & Velocity (CTRV)

vxk vyk

  • y

k

  • x

k

slide-12
SLIDE 12

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 12

Cooperative Localization – Filter Technology

  • Particle filter represent an arbitrary PDF by samples drawn from it
slide-13
SLIDE 13

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 13

Cooperative Localization – Filter Technology

  • Kullback-Leibler divergence is used for resampling
  • keeps bandwidth low as the number of particles

is adaptive to the current estimation

  • judging particles on the measurement likelihood function
slide-14
SLIDE 14

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 14

Cooperative Localization – Common Information and PF

  • Transformation of received distribution into local state space:

Shifting the received distribution by the smoothed pseudo range double difference distribution

  • Converting one distribution into continuous

representation for support equality

slide-15
SLIDE 15

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 15

Cooperative Localization – Simulation Results

Without communication With communication

slide-16
SLIDE 16

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 16

Cooperative Localization – Simulation Results

RMSE for different number of communicating vehicles after 300 samples

slide-17
SLIDE 17

Eric Richter eric.richter@etit.tu-chemnitz.de

TECHNISCHE UNIVERSITÄT CHEMNITZ

Professorship of Communications Engineering 11/14/08 17

Conclusion

  • Decentralized networks provide a scalable way of fusing information produced

by moving groups of vehicles

  • Improved position estimates
  • Avoiding data incest by removing common information using a general method
  • Extendable by arbitrary sensors and further a priori knowledge (like maps)

by using particle representations

  • Future work: Evaluation the proposed system with real data
slide-18
SLIDE 18

TECHNISCHE UNIVERSITÄT CHEMNITZ

Eric Richter eric.richter@etit.tu-chemnitz.de Professorship of Communications Engineering 11/14/08 18

Thank you

www.carai.de