Focused Refinement in the RRT*: Trading Optimality for Improved Performance
Beth Boardman, Sonia Martı́nez, Troy A. Harden · 2015
This paper investigates how to limit the exploration property of the RRT* algorithm in order to decrease the computation time needed to produce a low-cost, but good enough, path. We aim to do this by (i) focusing the attention of the RRT* algorithm on paths that are found quickly by RRT*, and by (ii) reducing the number of nodes in the obtained paths. The latter is achieved by an online smoothing process that aims to connect added nodes directly to their grandparents. Extensive two-dimensional simulation results are provided to examine how the number of obstacles in an environment affects the proposed extensions. Simulations for a Dubins’ vehicle are presented to show how the modifications perform for vehicles with differential constraints.