On the use of randomized low-discrepancy sequences in sampling-based motion planning

Abstract

This paper shows the performance of randomized low-discrepancy sequences compared with others low-discrepancy sequences. We used two motion planning algorithms to test this performance: the expansive planner proposed in [1], [2] and SBL [3]. Previous research already showed that the use of deterministic sampling outperformed PRM approaches [4], [5], [6]. Experimental results show performance advantages when we use randomized Halton and Sobol sequences over Mersenne-Twister and the linear congruential generators used in random sampling. © Springer-Verlag Berlin Heidelberg 2005.