Practical Prescribed Time Control Framework for Decentralized Robust Steering of Connected Automated Vehicles Under Deception Attacks
提出一种预设性能控制方法，处理车辆间距离和角度约束，在欺骗攻击下实现稳定性，并利用自适应神经网络逼近未知非线性函数，仿真验证了比现有方法更快的收敛速度。
Effective vehicle control contributes to the safety and efficiency of connected automated vehicles (CAVs). Many existing solutions do not consider the maximum effective communication distance and bearing angle constraints between vehicles. This article proposes a novel prescribed performance method to handle distance and angle constraints to achieve vehicle stability under deception attacks. A key aspect is that the above two constraints are successfully transformed from inequality-constrained form to equivalent equation unconstrained form through introducing error transformations, and we prove that the errors of distance and angle are strictly contained within the boundary of the performance function. Another key aspect is to use adaptive bias radial basis function neural network (RBFNN) to approximate unknown nonlinear functions and deception attacks in the system and integrate the approximated results into recursive construction to design adaptive laws and multilane merging control laws. Analysis shows that all signals in a closed-loop system are practical prescribed time stable. Simulations validate that our control method has a faster convergence time than the existing advanced two-dimensional (2-D) vehicle approach and can adaptively adjust convergence to predefined sets under different attack intensities.