Design of 3D bridge crane controller based on TD3 algorithm

This paper addresses the issues of high coupling and nonlinearity in the underactuated system of a crane by drawing inspiration from the application of deep reinforcement learning in the field of robotic arm control. The approach of deep reinforcement learning is applied as the controller for the crane. The crane is first modeled using Simscape Multibody, and then the TD3 algorithm is chosen as the controller. Suitable state variables and a reward function are selected to describe the crane's state and objectives, enabling precise control while satisfying constraints. A comparison with traditional PID control methods demonstrates that the proposed approach achieves higher work efficiency and control performance.