Online Reinforcement Learning and Sliding Mode Cooperative Control Strategy for Overhead Cranes

This study proposes a control strategy combining online reinforcement learning and sliding mode for trolley positioning and payload swing suppression of overhead cranes. First, an actor–critic reinforcement learning framework is built. And then an adaptive algorithm is designed to address three goals: trolley positioning, swing suppression, and energy consumption reduction. Next, a robust discrete-time sliding mode control law is designed in accordance with reaching law. Finally, the above two parts of control are integrated to form a composite control structure. In addition, the stability of the crane system under the composite control effect is guaranteed. The composite control system fully leverages the advantages of reinforcement learning and sliding mode control, endowing it with strong robustness and adaptability. Extensive validation and comparisons across multiple scenarios demonstrate that the proposed control strategy achieves up to 48% reduction in payload swing compared to sliding mode control. Simultaneously, it enables overshoot-free positioning and exhibits enhanced robustness against online reinforcement learning.