An Adaptive Rapidly-Exploring Random Trees Algorithm Based on Cross-Entropy Optimization

In this paper a novel adaptive rapidly-exploring random trees algorithm based on cross-entropy optimization (CE-RRT) is proposed. We seek to provide a low-cost, fast, and effective solution for path planning of robots in various complex environments. Firstly, an adaptive sampling strategy is introduced to make the search directional. Then, an adaptive step adjustment strategy is proposed to improve the search efficiency of the algorithm. Finally, the cross-entropy algorithm is introduced to optimize redundant nodes in feasible paths and improve path quality. In order to verify the feasibility and effectiveness of the proposed algorithm, it is used to solve path planning problems in two two-dimensional environments and one three-dimensional environment. The RRT and RRT* algorithms are used as benchmarks to measure the effectiveness of the three optimization strategies. The simulation demonstrates that the proposed CE-RRT algorithm can effectively improve search efficiency and path quality. Particularly (path shortened by 26%, 22.70%, and 49.11%), the CE-RRT algorithm exhibits stronger robustness in three-dimensional environments. In addition, the proposed CE-RRT algorithm can be used to plan a reasonable path for the dual robot based on the dual Sawyer simulation platform.