Fast approximation to gaussian obstacle sampling for randomized motion planning

Abstract

We propose a new optimization of randomized motion planning via a local directed strategy. The basic motion planning problem is to find a collision free trajectory for a moving object (rigid, articulated or deformable) in a static or dynamic environment. We propose an improvement of the Rapidly-exploring Random Tree (RRT) method in associating the concepts of visibility and Gaussian sampling. This improvement focuses on the random sampling and its localization in free spaces. The new Gaussian sampling is described by a set of geometrical primitives and permits to define the random sampling behavior in the entire free space. In this paper, we first consider the existing alternatives for random sampling. Then we propose our localized random sampling that refines the environmental possibilities such as free space evaluation according to the mover's dynamic constraints. The environmental possibilities are identified during the RRT development. The experiments and results validate that our method improve the mover's trajectory in static environments.