Navigation and World Modeling for a Mobile Robot: A Progress Report

Cet articleprésente les progr-ès récents clans le développementdes robots mobiles capablesde comportement intelligent dirige par un but. Sa contribution essentielte est la description d'une architecture logicielle qui permet au systeme robotique do coordonner action et perception en utilisant la connaissance. Cette architecture so decompose en deux architectures hiérarchiques pour le contrôle de la perception et des actions.

'L'article resume les dévelopements récents en navigation et en perception. Parmi les themes abordés, un point important est la presentation d'un protocole de controle pour le niveau '.véhicule. Ce protocole définit un véhicule virtuel qui facilite le transport des systémes de • navigation entre véhicules. L'article traite egalement Ies probtémes d'estimation de l'incertitude de la position estimée du véhicule ainsi que les avancés récentes en matière de modélisationcomposite.

Factory robots function in highly engineered environments and thus can be made to function with very limited sensing and intelligence. The introduction of robots into such unconstrained environments such as a construction site or a mine requires a radically different approach. Such systems must sense and model their environment and base their actions on such a model. They must also apply intelligence so that they do not simply repeat actions, but accomplish goals.

This paper presents recent progress towards the development of mobile robots which intelligently accomplish goal oriented behavior. A primary-contribution is the introduction of an architecture which permits a robotic system to apply knowledge to the coordination of action and perception. Hierarchical architectures are presented for both perception and navigation.

discussion of recent developments in navigation and perception are presented. Among the most important points is the presentation of a general purpose vehicle-level control protocol. This protocol defines a virtual vehicle which can permit navigation systems to be easily transported between vehicles. The problems of estimating the uncertainty in a vehicle's estimated position, and recent advances in composite modeling are also discussed.