Tire-Stiffness Estimation by Marginalized Adaptive Particle Filter

This paper considers longitudinal and lateral tirestiffness estimation for road vehicles, using wheel-speed and inertial measurements. The deviations from nominal stiffness values are treated as disturbances acting on the vehicle, and are included in a nonlinear vehicle model. We formulate a Bayesian approach based on particle filtering, where the tire stiffness as well as the associated uncertainty are jointly estimated together with the vehicle velocity vector, the yaw rate, and the bias components for the inertial sensors. For computational efficiency, we marginalize out the noise parameters, hence do not need to include them in the state vector. Experimental data for a double lane-change maneuver indicate that the stiffness can be estimated within a few percent of the true values.