Neural network-based collision-free optimal formation control for unmanned surface vehicles with the gain iterative disturbance observer

Abstract

Current maritime operations require the application of unmanned surface vehicles (USVs), which have more reliable path tracking capabilities and can extend the mission duration. To address the practical needs of USV formation control, this paper proposes an intelligent collision-free optimal formation control scheme for USV systems with external disturbances. Firstly, an artificial potential field (APF) function with continuous partial derivatives is developed to avoid collisions between USVs and potential obstacles which include other vehicles and environmental obstacles. When the obstacle exits the detection range, the APF function with smooth and continuous partial derivatives avoids the rotation phenomenon. Secondly, a gain iterative disturbance observer (GIDO) with a gain iterative mechanism is designed under the unfavorable effects of external disturbances. Unlike conventional disturbance observers that employ fixed gain coefficients of the disturbance term, the gain of GIDO can be dynamically adjusted by an iterative mechanism to accurately estimate the disturbance and thus improve the robustness of the USV system. Moreover, an actor-critic reinforcement learning algorithm is employed to balance the control performance and costs, thereby to optimize the energy consumption during USV formation. Finally, the optimized backstepping control strategy is proposed to ensure that USVs move to the specified location without any collision. The feasibility and effectiveness of the proposed control approach are well illustrated by simulation results.