Data-Driven Steering Control for Heterogeneous Autonomous Vehicle Platoons via Interval Excitation-Based Learning Approach
研究了异构自动驾驶车辆队列的路径跟踪问题，提出一种数据驱动转向控制方案，无需精确模型，通过区间激励条件降低计算复杂度，仿真验证了可行性。
This article investigates the path following problem for heterogeneous two-degree-of-freedom (2-DoF) autonomous ground vehicle platoons (AGVPs) with unknown dynamics. The control objective is formulated as an inhomogeneous linear quadratic tracking (LQT) problem, aiming to minimize the lateral relative offset and heading error between the following vehicle and its preceding vehicle by designing the steering angle. To overcome the reliance on accurate system models, a novel data-driven control scheme is proposed. First, a model-based iterative learning algorithm is proposed using the matrix decomposition method. This algorithm removes the requirement for an initial stabilizing control policy while guaranteeing the convergence speed. Furthermore, it avoids the ill-conditioned numerical issues caused by small discount factors in LQT problems. Furthermore, the algorithm is extended to a data-driven implementation via a double-layer integral structure, which relaxes the persistence-of-excitation (PE) condition to the interval excitation (IE) condition. The proposed algorithm can iteratively solve the optimal LQT controller from any initial policy, while reducing the computational complexity and eliminating the requirement for historical data storage. Finally, the simulation results demonstrate the feasibility and superiority of the proposed method.