A multiple surrounding point set approach using Theta* algorithm on eight-neighbor grid graphs

Abstract

Path planning on grid graphs has been studied for decades in several applications such as robotics and computer games. However, the process for determining short paths can still be inefficient on large grid graphs. The SPS (Surrounding Point Set) method has made it efficient by considering only a few unblocked cells. Since eight-neighbor grid graphs do not provide sufficient connectivity for SPS, it found a path on a graph where two cells are connected with an edge within a given density bound. However, verifying whether the resulting edges are blocked is time-consuming; further, it is unclear which density bound to use and restrictive to allow only edges within the bound. To address these limitations, we propose a multi-SPS with Theta* that uses eight-neighbor grid graphs, allows distant connectivity by Theta*, and improves the SPS by considering more obstacles between the start and goal cells. Our experimental results demonstrate that compared with the existing path planning algorithms, multi-SPS with Theta* and variants of Theta* can run faster and obtain similar path lengths. The results also demonstrate the beneficial effect for Theta* in that the expensive runtime of Theta* can be reduced by the SPS approach.