关键词: 双机械臂协同, 路径规划, 快速扩展随机树, 动态窗口法, 动态避障, ROS 2平台

Abstract: Aiming at the problems of low search efficiency,local optimality and obstacle avoidance hysteresis in high-dimensional dynamic scenarios when traditional algorithms are applied to dual industrial manipulators,a global-local two-layer path optimization framework was constructed. Specifically,the Rapidly exploring Random Tree (RRT) controlled by a gravitational function was adopted and combined with the Dynamic Window Approach (DWA). Meanwhile,a task scheduling state pool was designed to realize path mutual avoidance and beat synchronization of cooperative manipulators in the shared space. A simulation evaluation of dual industrial manipulators cooperative operation in high-density obstacle and dynamic interference environments was conducted based on the ROS 2 and Gazebo platforms. Compared with the unoptimized path,after the dynamic path planning and optimization algorithm was enabled,the average path length was shortened by approximately 18.0 %,the average energy consumption was reduced by about 18.0 %,the number of conflicts decreased from 3 to 0,and the scheduling delay was also significantly shortened. Subsequently,the algorithm was deployed in an actual metal component assembly unit to complete a two-week continuous grasping and insertion task verification. The average planning time was maintained within 85 ms,the average pose error was controlled within ±2.3 mm,and no shutdown conflicts occurred. These results demonstrate that the algorithm has good engineering feasibility and stability. Studies have shown that the coupled strategy of gravity bias and dynamic obstacle avoidance can significantly improve the path quality and system robustness of dual industrial manipulators collaborative operations,and provide effective technical support for the flexible scheduling of intelligent manufacturing production lines.

Key words: dual manipulator collaboration, path planning, rapidly-exploring random tree, dynamic window approach, dynamic obstacle avoidance, ROS 2 platform