Vehicle modeling and parameter estimation using adaptive limited memory joint-state UKF

Vehicle modeling is an essential part of controller design and validation. This is especially true for model-based control design approaches, such as model-predictive control (MPC), which require an accurate model for predicting the vehicle motion. In this paper we propose a new adaptive joint-state unscented Kalman filter (JUKF) to estimate the unknown vehicle parameters using experimentally collected data. We test the proposed algorithm using three nonlinear vehicle models of increased fidelity: a single-track model, a double-track model and a full 11-dof vehicle model. Simulation results validate the proposed algorithm.