Reinforcement-Learning-Based Adaptive Optimized Fixed-Time Containment Control for Multiple QUAVs Under Malicious Attacks: A Flexible Tunnel Constraint Approach

초록

This article zeroes in on the adaptive resilient optimized containment control for multiple quadrotor uncrewed aerial vehicle under malicious attacks and input saturation. First, a flexible tunnel performance constraint strategy is presented to reduce the conservative constraints distributed on both sides of traditional prescribed performance methods, where the auto-adjustable envelopes effectively prevent the containment error from exceeding the maximum allowable tight sets. Subsequently, by constructing a modified fixed-time filter and the identifier-actor-critic structure, a reinforcement-learning-based dynamic surface containment control strategy is established to train an approximate optimal control solution. In addition, an adaptive fuzzy wavelet neural resilient optimal fixed-time containment protocol is developed to counteract malicious attacks, effectively removing the conservatism assumption of the boundaries of multiplicative and additive attack signals. Stability analysis demonstrates that all signals of the closed-loop attitude control system are fixed-time bounded, and the output signals of the followers enter into the convex hull formed by the leaders. Finally, the feasibility of the proposed control algorithm is substantiated by illustrative results.

키워드