ZHOU Zijian, LI Lixia, LI Zhimin, LIU Tao, XU Jijin, SHI Ruxing
To address the limitations of traditional Rapidly-exploring Random Tree (RRT) algorithms in robotic path planning—such as low sampling efficiency, poor path quality, and insufficient consideration of joint torque and motion smoothness constraints—we propose an Adaptive RRT with Mechanical Constraints (ARRT-MC). This algorithm enables efficient, safe and smooth path generation for six-degree-of-freedom robotic arms in joint space. The algorithm employs Sobol sequence to generate initial samples, combines a collision feedback-driven adaptive bias mechanism to dynamically adjust sampling density and direction, thereby improving coverage and convergence speed. During parent node selection phase, a multi-objective cost function is introduced to comprehensively balance path length, maximum joint torque, motion smoothness, and environmental guidance, achieving global optimization of path quality. The expansion phase incorporates a collision-and-moment-feedback-based joint expansion strategy for starting and target points, along with dynamic step-length adjustment, to enhance path guidance and robustness. By integrating bidirectional asynchronous path pruning with local optimization, adaptive smoothing, and local optimization mechanisms, the algorithm effectively eliminates redundant paths, improves trajectory continuity, and suppresses joint torque fluctuations. Numerical results demonstrate that the proposed algorithm significantly outperforms traditional methods in path quality, convergence efficiency, and motion performance, showing promising engineering application potential.
