Hierarchical decision-theoretic robotic surveillance

Abstract

In this paper, we discuss a decision-theoretic strategy for surveillance as a first step towards automating the planning of the movement of an autonomous surveillance robot. We extend a previous proposal by including some heuristics based on an abstract representation of the environment. We show, using a concrete example, how these heuristics allow computationally feasible, finite look-ahead versions of the decision-theoretic strategy to escape local minima, and to better approximate globally optimal solutions. 1 Introduction In several projects, robots are employed to perform surveillance tasks. See, for example, [ Cybermotion, 1990; Murphy and Bott, 1995 ] . However, in most of these projects, the robots are typically used as a kind of flexible sensor platform controlled by some external human operator who makes the high-level decisions on where to go, and how to use the on-board sensors. We are interested in automating this high-level decision making process to allow a...