Dynamic curvature-adaptive MPCC with physical feasibility constraints for high-speed racing under frequent curvature variations

In high-speed racing environments, frequent changes in track curvature can compromise the stability of control systems, thereby limiting the racing performance of autonomous vehicles. To address this challenge, an enhanced model predictive contouring control algorithm integrating a curvature-aware dynamic error adaptation mechanism with physical feasibility constraints (CA-PF-MPCC) is proposed. The algorithm employs a dynamic weight adjustment function based on trajectory curvature, adaptively balancing contouring and lag errors to enhance responsiveness in sections with varying geometric characteristics. Additionally, by combining the friction ellipse theory with curvature-velocity coupling constraints, a unified control-speed feasible domain is established, which effectively suppresses physical instability during aggressive maneuvers. Experimental results indicate that, under scenarios with frequent curvature changes, CA-PF-MPCC reduces racing time by 3.263 s, 2.715 s, 4.289 s, and 1.720 s compared to the Feedforward-Feedback, MPCC, MPC-CBF, and LPV-MPC algorithms, respectively. The rate of change of the heading angle decreases by 0.062 rad/s, 0.070 rad/s, 0.139 rad/s, and 0.038 rad/s, correspondingly, thereby effectively suppressing heading fluctuations and improving attitude stability. These findings demonstrate that the proposed method significantly enhances racing performance while improving dynamic adaptability and high-speed stability on complex tracks, highlighting its practical relevance.