Aiming at the path planning problem of pear pollination manipulator in complex environment, this paper proposes a path planning strategy based on deep reinforcement learning. According to the forward and inverse kinematics characteristics of the manipulator, the state space and action space of the manipulator are designed, and the reward function is set based on the distance between the manipulator and the obstacle. The simulation environment is simulated by Deep Deterministic Policy Gradient (DDPG). The simulation results verify that the system can quickly and stably control the manipulator to reach the pollination target. It provides theoretical and technical support for the autonomous operation of the pollination manipulator.