A Comprehensive Survey and Analysis on Path Planning Algorithms and Heuristic Functions

Abstract

This paper explores the performance of four commonly used path planning algorithms of A*, D*, LPA* and D* Lite in both static and dynamic environments. To assess the effect of heuristic functions for path planning algorithms, four commonly used heuristic functions of Manhattan distance, diagonal distance, Euclidean distance and squared Euclidean distance are used based on three simulation environments with different map size and complexity. Experimental results show that the heuristic function significantly affects the computing performance and the final path of path planning algorithms, and Manhattan distance is the most efficiency heuristic function for all the four path planning algorithms. We apply A*, D*, LPA* and D* Lite algorithms to static and dynamic environments, respectively with different map size and complexity. We find that A* has the best performance in relatively simple static environments. When the map size and complexity of the static environment is highly increased, D* Lite is the best choice. In contrast, D* Lite has the best performance for path planning in dynamic environments. However, we also find that A* also has an excellent performance in dynamic environments when the map size and complexity are decreased.