Real-time identification of the tyre–road friction coefficient using an unscented Kalman filter and mean-square-error-weighted fusion

Vehicle active safety systems can benefit significantly from a priori knowledge of the road conditions. This paper proposes a real-time tyre–road friction estimation algorithm based on an unscented Kalman filter and mean-square-error-weighted fusion, using measurements related to the electric vehicle and in-wheel motors. Because the modules are mutually independent, the dimensionality of the matrix used in the filtering process can be reduced to shorten the computational period. The approach can also work more effectively in various conditions than those techniques which focus on a specific direction of the vehicle dynamics. First, a modified Dugoff et al. tyre model is considered to express the non-linear characteristics of the tyres. Then, to observe the longitudinal tyre forces and the lateral tyre forces, a parameter identification technique that utilizes the information from an inertial sensor and a gyroscope is introduced. An unscented Kalman filter rather than the traditional linear filter is adopted to obtain higher accuracy in the saturation region. After the relationship between the mean square error and the side-slip angle or the slip ratio is taken into account, the final identified value is set by combining the results from the longitudinal unscented Kalman filter and the lateral unscented Kalman filter in order to improve the accuracy and the robustness further. Experiments conducted on both a dry road and a slippery road verify that the method based on an unscented Kalman filter can identify different tyre–road friction levels.