Tire model identification and development of a tire-road friction observer

Abstract

In order to improve vehicle safety systems and autonomous control of vehicles, the real time knowledge of tire forces and friction coefficient is desirable. The tire interaction with the road through the tire contact patch is the only means to control the movement of the vehicle. The goal of this thesis is to develop a nonlinear observer for estimation of the potential of the road friction coefficient during various driving maneuvers. Maneuvers will be performed by a nonlinear 14 degree of freedom vehicle model with independent suspension, which serves as a substitute for a physical vehicle. An Extended Kalman filter is chosen as the observer. Within the observer, the reference vehicle is represented by a nonlinear single-track model. The tire model used is a modified version of the widely known Pacejka Magic Formula tire model. To identify the parameters for the tire model, a gradient based minimization problem is solved to find the tire parameters such that the dynamic characteristics of the single-track model closely match those of the reference vehicle. Once the tire parameters are determined, a similar minimization problem is set up and solved to obtain the parameters for the Extended Kalman Filter. Parameters of the EKF are chosen such that the state estimation closely matches a reference measurement from the complex vehicle simulation model. The resulting observer shows good performance in estimating the friction coefficient during highly nonlinear maneuvers. Some maneuvers are performed that show the estimation of varying road conditions. The performance is governed by the quality of the - XIV - tire model. When the observer model shows poor performance in capturing the behavior of the reference vehicle, the estimation of the friction coefficient suffers. If the tire is operated in the linear region, the friction coefficient is harder to identify. Improvements in future work can be made by carefully selecting the maneuvers for identifying the tire model, such that a broader range of tire slip and normal load variations is reached.