Some Notes on the Integration of Planning and Reactivity in Autonomous Mobile Robots

Abstract

> none of the above types of knowledge is accurate. More to the point, most of the inaccuracy is inherent to the problem, and cannot be overcome by building a more sophisticated machine. The accuracy of the agent's knowledge about its environment is bounded by the noise and limited range of its sensors, and by the precision and granularity of the prior information. This is true, in particular, of agent's knowledge about its own location, and about the location of the target position. The complexity of the environment and the presence of other agents may cause unpredictable changes, and knowledge of the current state cannot be safely projected, whatever its quality. The effect of each action is muddled by errors in the On leave from Iridia, Universit'e Libre de Bruxelles, Brussels, Belgium. effectors, as well as by the weakness of the knowledge about the situation where it is performed. Finally, the agent's knowledge about its very goals may be inaccurate: e.g.,