Robot Arm Control Method of Moving Below Object Based on Deep Reinforcement Learning

The existing robot arm control system has long commissioning time and the control system has poor scope of application. In this paper, the Deep Deterministic Policy Gradient (DDPG) algorithm has been adopted and adapted to control the robot arm to move below the object at any position, thereby enhancing the flexibility of the control algorithm and shortening the adjusting time. In addition, to address the problem that physical production line cannot be utilized directly or provide sufficient data for training deep reinforcement learning agent, this paper constructs a virtual model containing both the robot arm and the object as training environment for the agent. Simulation experiment has been performed with state variables and reward properly designed. As is shown by the results, the control agent trained in this paper show good performance in controlling the robot arm, which in turn confirms the effectiveness of the training algorithm with effective data support of the constructed simulation environment.