Safe motion planning for a robot in a dynamic, uncertain environment

Abstract

The problem of efficiently planning a path for a robot between two points when the path is forced to change dynamically by the occurrence of certain events in the environment is addressed. An event, for example, may be the discovery of another moving object on a collision course with the robot. The robot is forced to take evasive action whenever such an alarm occurs. A probabilistic model that represents the robot's dynamic behavior in response to alarms with a Poisson distribution and safety rules that assume that some regions are safe are presented. A provably optimal expected solution for the problem is presented, and the role of the probabilistic parameter of the dynamic environment on the optimal path, and the role of vision (or time to collision) on the planned paths are studied. The simplicity of the model used is justified by the difficulty of analyzing a more complicated (unknown) dynamic environment, and by the generality of the results obtained using this simple model.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>