Fixed-time adaptive RBF neural network controller via minimum learning parameter for ship roll stabilization

Abstract

In this study, a fixed-time adaptive radial basis function (RBF) neural network controller is proposed for ship roll stabilization, considering fixed-time convergence, the computational burden reduction, unknown dynamics, and external disturbances. First, the fixed-time stability theory is integrated with the backstepping method to design a controller for the ship's anti-roll fin stabilizers. With this controller, the errors of the closed-loop system are ensured to converge into the origin with faster convergent time. Moreover, the settling time of the system is independent from the initial states. Second, the unknown dynamics of the ship rolling model are estimated by the RBF neural network. To reduce the computational burden of the neural control system, the minimum learning parameter (MLP) technique is incorporated into the adaptive law of the RBF neural network. Based on the Lyapunov theory, the stability of a closed-loop system is proven to be guaranteed within a fixed time. Finally, numerical simulations and comparison analyses are performed to demonstrate the effectiveness and superiority of the proposed controller.