Runtime reduction in optimal multi-query sampling-based motion planning

Abstract

Sampling-based motion planning algorithms have been successfully applied to various types of high-dimensional planning tasks. Recently an extension of PRM algorithm called PRM* planner has been proposed which guarantees asymptotic optimal solutions in terms of path length. However, the high runtime of sampling-based algorithms is still a serious disadvantage. In this paper, a new extension of PRM planner is proposed which incorporates the variable neighborhood radius feature of PRM* and the sampling radius of low-dispersion sampling in order to improve the cost of the generated solutions in terms of path length and runtime. The performance of the proposed algorithm is tested in different planning environments. Furthermore, the proposed planner is compared to the original PRM and the PRM* approaches and shows significant improvement.