Privacy-Protected Formation Control for Unmanned Surface Vehicles: A Neural Predictor-Based Noncooperative Game Framework

Abstract

This paper proposes a formation control strategy based on noncooperative game framework for unmanned surface vehicles (USVs) in the presence of eavesdroppers. Firstly, the interactions among USVs during formation control are modeled as a noncooperative game. A privacy-protected estimator is then developed for each USV by employing a twin-network structure, so as to effectively estimate the actions of other USVs while preventing information leakage incurred by eavesdropping attacks. Additionally, a novel neural predictor is designed using an accelerated learning-boosted echo state network (ALESN) to approximate unknown system parameters with enhanced convergence rate and prediction accuracy. On the basis of the estimator and predictor, a distributed formation control method is subsequently devised via seeking the Nash equilibrium of the formulated game. The implementation of the control method only requires local information exchanged between neighboring USVs, and thus the desired computational efficiency and scalability can be achieved. A simulation example is finally provided to validate the effectiveness of the reported privacy-protected formation control approach.