ParkNav Complex Dynamic Scenes Interpretation and Reactive Motion - - PowerPoint PPT Presentation

parknav complex dynamic scenes interpretation and
SMART_READER_LITE
LIVE PREVIEW

ParkNav Complex Dynamic Scenes Interpretation and Reactive Motion - - PowerPoint PPT Presentation

ParkNav Complex Dynamic Scenes Interpretation and Reactive Motion Planning Robea Project [October 2002- September 2005] Scientific Coordination: Thierry Fraichard eMotion, Gravir Inria Rh one-Alpes Robea Workshop, Toulouse (FR), January 28,


slide-1
SLIDE 1

ParkNav Complex Dynamic Scenes Interpretation and Reactive Motion Planning

Robea Project [October 2002- September 2005] Scientific Coordination: Thierry Fraichard eMotion, Gravir Inria Rhˆ

  • ne-Alpes

Robea Workshop, Toulouse (FR), January 28, 2004 – p.1/4

slide-2
SLIDE 2

Project Summary

Automate the driving of a vehicle. . . . . . moving amidst moving objects (vehicles, pedestrians) in a site equipped with video-cameras Joint integrated demonstrator using the Inria Rhône-Alpes testbed

Robea Workshop, Toulouse (FR), January 28, 2004 – p.2/4

slide-3
SLIDE 3

Scientific Problems

√ Interpretation of complex dynamic scenes with cameras ⋆ Static environment modelling ⋆ Detection, tracking and identification of moving objects in outdoor environments ⋆ Speed, robustness √ Reactive motion planning integrating environment dynamicity and uncertainty on moving objects future behaviour ⋆ On-line motion adaptation ⋆ Speed, safety

Robea Workshop, Toulouse (FR), January 28, 2004 – p.3/4

slide-4
SLIDE 4

Partners

√ eMotion (ex Sharp), Gravir √ Movi, Gravir √ Prima, Gravir √ RIA, Laas-CNRS √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-5
SLIDE 5

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development √ Movi, Gravir √ Prima, Gravir √ RIA, Laas-CNRS √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-6
SLIDE 6

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development ⋆ Reactive motion planning in highly dynamic environments (incremental trajectory computation, moving objects future behaviour prediction) √ Movi, Gravir √ Prima, Gravir √ RIA, Laas-CNRS √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-7
SLIDE 7

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development ⋆ Reactive motion planning in highly dynamic environments (incremental trajectory computation, moving objects future behaviour prediction) √ Movi, Gravir ⋆ Multiple camera calibration, moving object tracking with a pan-tilt-zoom camera √ Prima, Gravir √ RIA, Laas-CNRS √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-8
SLIDE 8

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development ⋆ Reactive motion planning in highly dynamic environments (incremental trajectory computation, moving objects future behaviour prediction) √ Movi, Gravir ⋆ Multiple camera calibration, moving object tracking with a pan-tilt-zoom camera √ Prima, Gravir ⋆ Moving object tracking with multiple fixed cameras √ RIA, Laas-CNRS √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-9
SLIDE 9

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development ⋆ Reactive motion planning in highly dynamic environments (incremental trajectory computation, moving objects future behaviour prediction) √ Movi, Gravir ⋆ Multiple camera calibration, moving object tracking with a pan-tilt-zoom camera √ Prima, Gravir ⋆ Moving object tracking with multiple fixed cameras √ RIA, Laas-CNRS ⋆ Reactive motion planning in weakly dynamic environments (nominal trajectory deformation) √ Vista, Irisa

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-10
SLIDE 10

Partners

√ eMotion (ex Sharp), Gravir ⋆ Project coordination, experimental testbed development ⋆ Reactive motion planning in highly dynamic environments (incremental trajectory computation, moving objects future behaviour prediction) √ Movi, Gravir ⋆ Multiple camera calibration, moving object tracking with a pan-tilt-zoom camera √ Prima, Gravir ⋆ Moving object tracking with multiple fixed cameras √ RIA, Laas-CNRS ⋆ Reactive motion planning in weakly dynamic environments (nominal trajectory deformation) √ Vista, Irisa ⋆ Moving object tracking with a pan-tilt-zoom camera

Robea Workshop, Toulouse (FR), January 28, 2004 – p.4/4

slide-11
SLIDE 11

First Year Activity Report

√ Experimental testbed development (specification, deployment): ⋆ Two fixed cameras, preliminary version of the “Dynamic Map Server” ⋆ Active collaboration of eMotion, Movi and Prima under the supervision

  • f Frédéric Hélin (eMotion)

√ “Visible” collaboration items ⋆ Plenary meeting: Gravir, January 2003 ⋆ Visit of Cedric Pradalier (eMotion) to RIA: experimental vehicle (Cycab) presentation ⋆ Stay of Olivier Lefèbvre (RIA) with eMotion: first live experiments on trajectory deformation

Robea Workshop, Toulouse (FR), January 28, 2004 – p.5/4

slide-12
SLIDE 12

Experimental Testbed Architecture

Vidéo, synchronisation commandes Réseau Routeur hertzien Client Parkview mobile Serveur de carte dynamique Client Parkview Station de travail Capteur 3 Capteur 1 Capteur 2 Bus

Robea Workshop, Toulouse (FR), January 28, 2004 – p.6/4

slide-13
SLIDE 13

Experimental Testbed

Robea Workshop, Toulouse (FR), January 28, 2004 – p.7/4

slide-14
SLIDE 14

Dynamic Map Server

Carte dynamique Éléments statiques Eléments mobiles Eléments semi-mobiles Composition

Robea Workshop, Toulouse (FR), January 28, 2004 – p.8/4

slide-15
SLIDE 15

Moving Object Tracking

Caméra homographie Robust Tracker

1

' X

2

' X

1

' Y

2

' Y

Caméra Robust Tracker

1

' X

2

' X

1

' Y

2

' Y

Fusion

  • Corr. distorsion

homographie

  • Corr. distorsion

Robea Workshop, Toulouse (FR), January 28, 2004 – p.9/4

slide-16
SLIDE 16

Preliminary Dynamic Map Server Results

Robea Workshop, Toulouse (FR), January 28, 2004 – p.10/4

slide-17
SLIDE 17

Perspectives

√ Increase the number of fixed cameras so as to have better coverage and precision √ Improve target fusion and tracking by taking uncertainty into account (Bayesian reasoning) √ Install a pan-tilt-zoom camera so as to allow the integration of Movi and Vista’s work √ Explore possible cooperation between fixed and pan-tilt-zoom cameras √ Identify moving object (vehicle, pedestrian) √ Couple on-board and “off-board” perception √ Experiment both reactive motion planning schemes using the Dynamic Map Server data

Robea Workshop, Toulouse (FR), January 28, 2004 – p.11/4