关键词: 机械臂, 路径规划, 双向RRT算法, 目标偏置, 自适应步长

Abstract: In response to the issues of poor target orientation, long convergence time, and low path quality in traditional rapidly-exploring random tree （RRT） algorithms, an improved algorithm is proposed. The algorithm first reads the map information and determines an adaptive initial step size, while restricting non-obstacle spaces to a random sampling space. A bidirectional goal bias strategy is incorporated to enhance the target orientation of the sampled points. Next, an adaptive step size strategy is employed, where the surrounding space of the nodes to be expanded is discretized. The expansion step size is adaptively adjusted based on the complexity of the obstacles, thus accelerating the algorithm’s convergence. After the search is completed, the path is pruned and optimized by removing redundant nodes and adding new nodes at path turning points to improve the path quality. Finally, the improved algorithm is tested in various two-dimensional and three-dimensional environments through simulation experiments. The results indicate that the improved algorithm outperforms the standard RRT algorithm, the GB-RRT algorithm, and the RRT-Connect algorithm in terms of planning time, path length, number of path nodes, and collision detection counts during the search process. Additionally, the obstacle avoidance simulation experiment with the AUBO-i5 robotic arm further validates the effectiveness of the proposed algorithm.

Key words: robotic arm, path planning, bidirectional RRT algorithm, goal bias, adaptive step size