Robot motion planning with uncertainty in control and sensing

摘要

Most of the research in robot motion planning has focused on finding a collision-free path connecting two configurations of the robot among obstacles, under the assumptions that geometric models are complete and accurate and that robot control is perfect. In this paper we consider the problem of planning motion strategies in the presence of uncertainty in both control and sensing, using the preimage backchaining approach. Though attractive, this approach raises difficult computational issues. One of them, preimage computation, is the main topic of this paper. We describe two practical methods for computing preimages for a robot having a two-dimensional Euclidean configuration space. We also propose a combination of these two methods which yields the most powerful practical method devised so far for computing preimages. A motion planner based on these methods has been implemented and experimental results are given in this paper. We discuss non-implemented improvements of this planner, including a new way to embed knowledge in a motion command in order to produce larger preimages.