<p>Traditional model predictive control (MPC) suffers from insufficient adaptability under complex dynamic conditions due to inadequate consideration of the time-varying characteristics of tire lateral stiffness. To address this issue, this paper proposes a dual-layer adaptive closed-loop collaborative MPC path tracking control method. The first layer operates on the vehicle model dimension: a segmented recursive least squares algorithm with limited memory (RLS-LM) is employed for real-time estimation of tire lateral stiffness across linear and nonlinear regions, mitigating data saturation and improving adaptability to nonlinear tire conditions. The second layer operates on the controller parameter dimension: an enhanced Particle Swarm Optimization (PSO) algorithm incorporating piecewise nonlinear inertia weight adjustment, condition-adaptive learning factors, and a Gaussian mutation mechanism is designed for offline optimization of prediction and control horizons under various speed-curvature conditions, with a multi-objective fitness function using tolerance-based weight allocation. These two layers are integrated into a collaborative closed-loop architecture: the RLS-LM estimator continuously corrects the MPC prediction model, while the PSO-optimized time-domain parameters ensure appropriate controller configuration, and the resulting control actions generate new vehicle states that in turn refine the stiffness estimation. Simulation results demonstrate that the proposed method significantly enhances path-following accuracy and maneuver stability under complex conditions.</p>