Training a Robotic Arm Movement with Deep Reinforcement Learning
Xiaohan Ni, Xin He, Takafumi Matsumaru · 2021 IEEE International Conference on Robotics and Biomimetics (ROBIO) · 2021
This paper introduces a general experimental design scheme for conditions and parameter settings of robotic arm control under the specific task when using Deep Deterministic Policy Gradient(DDPG) algorithm to train the robotic arm for completing the control task. Based on the Coppelia simulation tool, this paper builds an interactive reinforcement learning environment for robotic arm control tasks, and designs two different control tasks to verify the validity of experimental design schemes. Conclusions in this paper provide an important reference for finding suitable environmental design and parameter settings for using DDPG to train a manipulator and improving the training effect.