Passivity-Based Online Reinforcement Learning for Real Time Model-Free Overhead Crane System Control

In this paper, a novel model-free online Reinforcement Learning (RL) control method is proposed for the real-time overhead crane control problem. The crane control problem is first formulated as an optimal regulation problem with a user-specified objective function. Two neural-networks, namely Actor-Critic networks, are employed to approximate the objective function and the optimal control policy respectively. Then, an improved network updating rule with an additional passivity-based stabilization term is developed to remove the requirement of the initial stabilizing control policy. Unlike other crane control approaches, the proposed online RL algorithm does not rely on prior knowledge of the overhead crane mathematical model. Finally, simulation studies are carried out to demonstrate the effectiveness of the proposed method in the presence of system parameter variations when compared to LQR control.