AEB-RRT*: an adaptive extension bidirectional RRT* algorithm

Due to the probabilistic completeness and asymptotic optimality, the RRT* algorithm can find sub-optimal solutions and solve path planning problems effectively compared with other strategies. The B-RRT* algorithm introduces the bidirectional search to obtain faster convergence and shorter paths. To solve the problems of path planning in complex environments with concavities, convexities, narrow channels, mazes and multiple obstacles, an adaptive extension bidirectional RRT* (AEB-RRT*) algorithm is proposed. The AEB-RRT* algorithm simultaneously extends from both the starting point and the ending point, and adaptively adjusts the sampling probability and different expansion methods according to the number of collision detection failures. After acquiring a feasible path, it is then post-processed by interpolation and the principle of triangular inequality to obtain a shorter collision-free path. And the Manhattan distance replaces the Euclidean distance to calculate the distances between nodes in order to achieve higher computational efficiency. The experimental results show that the AEB-RRT* algorithm outperforms other path planning algorithms in stronger search ability, obtaining near-optimal or best paths with high efficiency and better robustness. Furthermore, it is successfully applied to solve the 6-DOF arc welding robot path planning problem.