Live Working Manipulator Control Model based on DPPO-DQN Combined Algorithm

To realize autonomous obstacle avoidance and navigation of live working manipulator, a DPPO-DQN combined algorithm is proposed. Firstly, the live working process of the manipulator is studied and the kinematics analysis of the 2D model is carried out to guide the coordinate transformation calculation of each node and trajectory of the manipulator. Secondly, the principle of the combined algorithm, the design of the environment state space, the action strategy and the reward function are studied, followed by realization of DQN, DPPO and DPPO-DQN algorithm training. Meanwhile, comparison the performance of obstacle avoidance and route navigation in two-dimensional. Finally, the combined algorithm is proposed and simulated. The experimental results show that the combined algorithm has a significant improvement in navigation success rate compared with the single model.