Research on Motion Planning of Seven Degree of Freedom Manipulator Based on DDPG

For the motion control of the seven degree of freedom manipulator, there are many problems in the traditional inverse kinematics solution, such as high modeling skills, difficulty in solving the equation matrix, and a huge amount of calculation. In this paper, reinforcement learning is applied in seven degree of freedom manipulator. In order to cope with the problem of large state space and Continuous action in RL, the neural network is used to map the state space to the action space. The action selection network and the action evaluation network are constructed with the Actor-Critic framework. The action selection policy is learned by the training of RL based on DDPG. Finally, test the effectiveness of the method by Baxter robot in Gazebo simulator.