Security Data-Driven Path-Following Control for Unmanned Surface Vehicles Under False Data Injection Attacks

Aiming at the problems of strong model dependence, uncontrollable convergence time of sideslip angle estimation, and single-step-only control optimization in the path following control of unmanned surface vehicles under false data injection attacks, this paper proposes a data-driven secure control method based on online learning. A fuzzy logic system is adopted to design an attack detection and estimation mechanism, realizing real-time compensation of the attack signals. A prescribed-time guidance framework consisting of a prescribed-time sideslip angle observer and an adaptive look-ahead distance line-of-sight guidance law is constructed, which can predefine the convergence time of sideslip angle estimation. Combining the infinite-horizon multi-step optimization capability of adaptive dynamic programming (ADP) and the data-driven characteristic of model-free adaptive control (MFAC), an ADP-MFAC optimal secure control law is designed. Simulation results show that the integral of time-weighted absolute error of yaw error of the proposed method is reduced by about 45.8% compared with the classic method, and the sideslip angle estimation performance is superior. This method requires no prior dynamic information, and can effectively mitigate the impact of short-duration attacks and guarantee path following performance.