Skip to Main content Skip to Navigation
Theses

Contrôle adaptatif d'un agent rationnel à
ressources limitées dans un environnement dynamique et incertain.

Simon Le Gloannec 1, 2
1 Equipe MAD - Laboratoire GREYC - UMR6072
GREYC - Groupe de Recherche en Informatique, Image, Automatique et Instrumentation de Caen
2 MAIA - Autonomous intelligent machine
INRIA Lorraine, LORIA - Laboratoire Lorrain de Recherche en Informatique et ses Applications
Abstract : This thesis deals with decision-theoretic autonomous agents. This work consists in constructing a control system for a resource-bounded agent evolving in a uncertain environment. Such agents must be able to control their resources consumption during a mission. The first part of this thesis introduces the concept of planning under uncertainty in general, and Markov decision processes (MDP) in particular, for the control. Solving techniques of large MDPs are presented.
In this control system, we consider resource-bounded agents adopting progressive reasoning as a specific resource-bounded reasoning with anytime behavior. We call progressive processing units (PRU) the task structure which allows the agent to adapt the quality of their accomplishment to the available resources. Each PRU defines a multi-level hierarchy task, to better accomplish the mission.
This thesis presents two extensions of the progressive reasoning : the control of multiple resources and an adaptive control system that faces changes during the mission. Firstly, algorithms are presented to avoid combinatorial explosion due to the multiple resources. Secondly, a value function approximation algorithm permits to quickly obtain a control system when the mission suddenly changes. Promising experimental results have been obtained and illustrated on a real robot.
Complete list of metadatas

Cited literature [12 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00157545
Contributor : Hal System <>
Submitted on : Tuesday, June 26, 2007 - 2:45:39 PM
Last modification on : Friday, February 26, 2021 - 3:28:05 PM
Long-term archiving on: : Monday, September 24, 2012 - 10:30:35 AM

Identifiers

  • HAL Id : tel-00157545, version 1

Citation

Simon Le Gloannec. Contrôle adaptatif d'un agent rationnel à
ressources limitées dans un environnement dynamique et incertain.. Intelligence artificielle [cs.AI]. Université de Caen, 2007. Français. ⟨tel-00157545⟩

Share

Metrics

Record views

431

Files downloads

291