Reinforcement learning-based distributed cooperative sliding mode control for unmanned surface vehicles

The complex marine environment poses a huge challenge to the ocean operations of unmanned surface vehicles (USVs). In this article, we will utilize sliding mode control theory and reinforcement learning (RL) method to consider collaborative control problem for USVs during the ocean operations. Firstly, by using the Kronecker product property, the USVs are further rewritten as an equivalent extended dimensional system, which can be implemented by the distributed control operations. For this extended dimensional system, we formulate the controller with sliding mode gain and reinforcement learning update rate. On this basis, we solve the collaborative control problem for USVs, and then, the corresponding algebraic criteria are also formulated in detail. Subsequently, the reachability is demonstrated for the designed sliding surface. To ends of the paper, an USV numerical example is given to verify the effectiveness of the collaborative control method and algorithm design proposed in this article.