Self-regulated robust variable forgetting factor recursive least squares for online vehicle state estimation

Accurate tire parameter identification is essential for reliable vehicle state estimation, as tire characteristics vary with operating conditions and directly affect vehicle dynamics observers. This paper proposes a Self-Regulated Robust Variable Forgetting Factor Recursive Least Squares (SR-VFF-RLS) algorithm for simultaneous online tire parameter identification and vehicle state estimation. The method extends the conventional Recursive Least Squares algorithm by introducing an adaptive forgetting factor regulated according to prediction-error variation and excitation level. In addition, robust adaptation mechanisms are incorporated to limit the influence of low-excitation conditions and undesired parameter updates. The proposed approach is first evaluated through numerical simulations under ideal conditions and subsequently validated using a high-fidelity vehicle model. Experimental validation is finally conducted using proving-ground measurements and benchmarked against the Simulator-enhanced Estimation Loop and the Dynamic-Kinematic Extended Kalman Filter methods. Results demonstrate faster convergence and more stable tire parameter identification than its competitors. The proposed approach also reduces lateral velocity and sideslip estimation errors, while its tire force estimates closely match wheel force transducer measurements and capture the nonlinear tire force–slip angle relationship. These results demonstrate the potential of the proposed approach for real-time tire parameter identification and vehicle state estimation in advanced vehicle dynamics applications.