Navigation autonome en environnement dynamique : une approche par déformation de trajectoire

Vivien Delsart 1
1 E-MOTION - Geometry and Probability for Motion and Action
Inria Grenoble - Rhône-Alpes, LIG - Laboratoire d'Informatique de Grenoble
Abstract : This thesis presents a navigation method in uncertain and dynamic en- vironment. More precisely, it consists in determining the motion of a robot from an initial position to a goal one, while preventing the robot to collide with the other agents evolving in its environment. Between deliberative approaches - consisting in determining a priori a complete motion to the goal - and reactive approaches - computing a new motion to execute at each time step during the robot navigation - have arisen the motion deformation approaches, combining a motion planning method with a reactive obstacle avoidance process. Their principle is simple : A priori complete motion is planned up to the goal and provided to the robotic system. During the course of the execution, the remaining part of the motion to execute is continually deformed in response to information provided by the sensors. The robot is consequently able to adapt its motion to the behaviour of the moving obstacles or to the incompleteness of its environment knowledge. Most of the existing motion deformation methods only deform the geometric path followed by the robot. We propose thus to extend the previous approaches to a trajectory deformation approach that modify the followed motion either in space or time. To do it, trajectory deformation reason on an estimation of the future motion of the obstacles. By preventing the trajectory followed by the robot to collide with a forecast model of the future motion of the obstacles, the robotic system may anticipate their motion. As the deformed trajectory is arbitrarily modified in time and space, one of the major difficulties of the approach is to keep the motion constraints of the robot satisfied along the trajectory. In that aim, a trajectory generation approach with a final time constraint has thus been developed. By discretizing the deformed trajectory in a sequence of state-times, the trajectory generation process allows to check if a feasible motion exists between each triplet of successive state-times, and should the opposite case occur they are modified to restore the connectivity of the deformed trajectory. The trajectory deformation and trajectory generation with final time constraints have been illustrated by simulation results, and a few experiments have been proceeded on an automated wheelchair.
Document type :
Theses
Liste complète des métadonnées

https://tel.archives-ouvertes.fr/tel-00592259
Contributor : Vivien Delsart <>
Submitted on : Wednesday, May 11, 2011 - 5:57:00 PM
Last modification on : Thursday, October 11, 2018 - 8:48:02 AM
Document(s) archivé(s) le : Friday, August 12, 2011 - 2:53:29 AM

Identifiers

  • HAL Id : tel-00592259, version 1

Collections

Citation

Vivien Delsart. Navigation autonome en environnement dynamique : une approche par déformation de trajectoire. Automatique / Robotique. Université de Grenoble, 2010. Français. ⟨tel-00592259⟩

Share

Metrics

Record views

621

Files downloads

3668