Estimation-Based Adaptive Online LQI-Stanley Integrated Path Tracking Control for Autonomous Vehicles

This study presents an innovative, integrated control architecture combining adaptive online linear quadratic integral (AOLQI) and Stanley geometric control systems for autonomous vehicle path tracking. A key feature of the proposed framework is the offline optimization of the AOLQI parameters using the Particle Swarm Optimization (PSO) algorithm. To address varying road conditions, a Forgetting Factor Recursive Least Squares (FFRLS) algorithm is employed for real-time tire cornering stiffness estimation, complemented by a Kalman–Bucy filter for high-fidelity vehicle side-slip angle observation. The efficacy of this architecture is validated through MATLAB/Simulink and IPG CarMaker co-simulations across demanding benchmarks, including a 100-m radius circular path, the high-speed Hockenheim race track and the high-curvature Stelvio Pass road profile. Numerical evaluations demonstrate that, compared to the conventional LQI, the proposed approach achieves reductions in root mean square error (RMSE) of 44.85%, 24.76% and 51%, and decreases in integral square error (ISE) of 69.2%, 43.16% and 75.86% respectively, in these different scenarios. These results confirm that using an integrated approach with adaptive online LQI and Stanley control, alongside an estimation layer, ensures superior tracking precision and performance improvements across extreme road geometries.