Gaussian sampling for probabilistic roadmap planners

Abstract

Probabilistic roadmap planners (PRMs) have become a popular technique for motion planning that has shown great potential. A critical aspect of a PRM is the probabilistic strategy used to sample the free configuration space. In this paper we present a new, simple sampling strategy, which we call the Gaussian sampler, that gives a much better coverage of the difficult parts of the free configuration space compared with the standard uniform sampler. This results in much smaller roadmaps that can be computed faster. The approach uses only elementary operations which makes it suitable for many different planning problems. Experiments indicate that the technique is indeed very efficient in practice.