Aiming at the operational problems of robots in dynamic and complex environments， the motion control method of manipulator tracking moving targets and avoiding obstacles is studied. The traditional PID control method is combined with DDPG algorithm. PID control is used to make the working plane of the manipulator approach the target quickly and coincide with it. Then DDPG algorithm is used to make the manipulator autonomously learn to track the projection of target and avoid the projection of obstacles in the plane， and finally achieve tracking and obstacle avoidance in the three-dimensional space. This method takes the manipulator as an agent which perceives the distance deviation between the target and the obstacle to learn obstacle avoidance and tracking strategies automatically. This method is applied to the path planning and obstacle avoidance task of multi-degree-of-freedom manipulator. Compared with DDPG algorithm， which only takes manipulator as agent to make decision control in space， the simulation results show that the proposed method has a better convergence effect and control performance， and the manipulator can stably track the target and avoid obstacles after training.