Abstract: For the casting sorting robot can quickly plan a stable and reasonable path when sorting castings, effectively avoid obstacles, improve the efficiency of motion planning, and carry out motion planning research through ROS. Use the C++ interface provided by ROS to complete the Descartes designated trajectory planning verification. In order to further improve the sorting efficiency and obstacle avoidance ability, the improved algorithm introduces a local iterative artificial potential field on the basis of the GB-RRT^* algorithm, and adds an adaptive target bias strategy and a direct connection strategy. Simulation experiments and casting sorting experiments have been designed and completed through Matlab and ROS. The experimental results show that the use of improved algorithms can effectively reduce the generation of invalid points, shorten the planning time, reduce the length of the path, and improve the planning efficiency. At the same time, when planning the movement, the movement of each joint is stable and the trajectory is smooth, which greatly improves the casting sorting efficiency, improves the sorting environment, and ensures its safety.