Synergistic Feedforward and Optimal Feedback Control for Lateral Trajectory Tracking of High-Speed Autonomous Vehicles
Abstract
To address trajectory tracking degradation and instability in autonomous vehicles during extreme obstacle avoidance, this paper proposes a synergistic lateral control algorithm combining kinematic feedforward compensation and a Linear Quadratic Regulator (LQR). First, a fourth-order error state-space model is constructed using two-degree-of-freedom vehicle dynamics in the Frenet frame. To mitigate phase lag under high-speed and large-curvature conditions, a curvature-based feedforward module is integrated with an optimal feedback gain matrix derived via the Discrete Algebraic Riccati Equation (DARE). The algorithm is validated through a high-fidelity 72 km/h Double Lane Change (DLC) simulation incorporating tire friction limits and actuator delays. Comparative results demonstrate that the proposed method drastically reduces the lateral Root Mean Square Error (RMSE) to 0.042 m—an approximate 65% accuracy improvement over optimized Pure Pursuit and Stanley algorithms—while strictly clamping lateral acceleration within physical safety boundaries. Furthermore, error phase portrait analysis mathematically confirms the global asymptotic stability of the closed-loop system, achieving Pareto optimality between tracking fidelity and chassis smoothness.