Improving RRT's Efficiency through Motion Primitives Generation Optimization

Abstract

Rapidly-exploring Random Tree (RRT) algorithm has been used for motion planning in numerous and diverse robotic applications. For applications which demands higher motion resolution, the computational cost increases together with the number of motion primitives used to expand the RRT. In this paper we present a method based on optimization by elimination which is applied to the Rapidly-exploring Random Tree algorithm to reduce its computational cost. This method optimizes the efficiency of motion primitives generation. It identifies tree expansion in promising areas by initially generating and analyzing only few motion primitives. Then it increases the number of primitives for motion resolution enhancement in those promising areas. The results achieved by applying this method evidence a substantial decrease in the computational cost.