A probabilistic framework for dynamic motion planning in partially known environments

Current approaches to robot motion planning are limited in their ability to deal with an uncertain and dynamically changing environment. The author discusses the difficulties involved in modeling the situation and proposes a probabilistic model based on discrete events that abstracts the dynamic interaction between the mobile robot and the unknown part of the environment. The resulting framework makes it possible to design and evaluate motion planning strategies that consider both the known portion of the environment and the portion that is unknown but satisfies a probability distribution. Studies of instances of the general model that have yielded useful results in designing efficient motion planning algorithms vis-a-vis parameters representing a robot's environment and the reactive component of its behavior are summarized.<<ETX>>