ACB-RRT*: Adaptive Companion Points Bidirectional RRT* Algorithm
摘要
To solve the problems of long search time, poor convergence and many redundant points in the RRT* algorithm, an adaptive companion point bidirectional RRT* algorithm (ACB-RRT*) is proposed. This algorithm adopts a hybrid strategy which including target-biased, dynamic step, and companion point generation. The target-biased strategy takes the target node as a sampling point according to the size of the random sampling probability to enhance the guidance. The dynamic step adopts different step sizes for expansion according to the random sampling probability to accelerate the convergence of the algorithm. The companion point generation determines multiple corresponding companion points based on the obtained expanded point to reduce the number of iterations. It also dynamically adjusts the angle of generating companion points based on the number of failed expansions. After obtaining a feasible path, trajectory optimization is performed on it. Greedy algorithm and cubic B-spline curve fitting are used to optimize nodes and smooth the trajectory, and finally an optimal collision-free path is obtained. Compared with RRT, RRT*, RRT-GoalBias, and B-RRT* algorithms, the results show that ACB-RRT* algorithm outperforms them in search time, path length, and number of iterations, indicating the superiority of this algorithm. Additionally, the algorithm has been successfully applied to welding scenarios.