Skip to Main content Skip to Navigation
Theses

Fondation d'un planificateur robotique intégrant le symbolique et le géométrique

Abstract : In this thesis report, we address the problem of planning for robotic systems by facing its full complexity : we take into account geometric, kinematic and symbolic constraints. For instance, we are interested in problems where several mobile manipulators robots have to manipulate and assemble objects in a constrained environment. Whereas task planning and motion planning have been solved independently so far, we propose to consider these problems as a whole and to combine specific techniques in a single planner. To reach this goal, we introduce a new search space and propose original search algorithms. This work has been validated through the development of a system called "aSyMov" and we report some encouraging results on complex problems. The report consists in five main chapters. First we describe state-of-the-art motion planning methods. Then a new search space and new search techniques are introduced to generalize and combine the previous algorithms. This is followed by a summary of the most recent techniques in task planning and by the description of an original way to bind geometric and symbolic domains. Then we turn to the presentation of the architecture and the algorithms of our planner. ASyMov is a forward search planner in the state space. A state in aSyMov actually stands for all the possible geometric instances defined by a given set of symbolic facts. At each step of the search, the chosen action is validated through a procedure carefully designed to reduce the number of geometric instances to check. This is achieved by triggering back-propagation of geometric constraints only for geometric actions. A strong advantage of our solution is the possibility to interwine the learning of the environment topology (through motion planning techniques) and the search of a solution. Finally we show several problems solved by aSyMov: these original examples necessitate to consider both geometric and symbolic aspects and thus illustrate the strength of our co ntribution.
Document type :
Theses
Complete list of metadata

Cited literature [61 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00010249
Contributor : Emilie Marchand <>
Submitted on : Thursday, September 22, 2005 - 3:01:12 PM
Last modification on : Friday, January 10, 2020 - 9:08:09 PM
Long-term archiving on: : Friday, April 2, 2010 - 10:10:07 PM

Identifiers

  • HAL Id : tel-00010249, version 1

Citation

Fabien Gravot. Fondation d'un planificateur robotique intégrant le symbolique et le géométrique. Automatique / Robotique. Université Paul Sabatier - Toulouse III, 2004. Français. ⟨tel-00010249⟩

Share

Metrics

Record views

322

Files downloads

303