Neural Network Learning Predictive Cooperative Control for Uncertain Heterogeneous Marine Vehicles With Unfavorable Communication Factors

This paper addresses the issue of cooperative control for heterogeneous marine vehicle systems (HMVSs) subjected to uncertain dynamics, multiple unfavorable communication factors, and velocity constraints. A state predictor is constructed to compensate for the missing information caused by communication delays, packet loss, replay attacks, and denial-of-service (DoS) attacks, thereby improving the convergence speed of the system. To address the impact of uncertain dynamics and deception signals on motion control and the resulting prediction errors, neural networks with identification capability are developed. Subsequently, a distributed cooperative controller comprising two parts is proposed for HMVSs, based on the prediction results and neural network outputs. The unknown compensation and cooperative control problems are then transformed into respective error stability problems, for which several sufficient conditions are derived to ensure finite-time boundedness of the errors. Finally, several numerical simulations are presented to validate the performance of the entire control architecture.