A two-stage adaptive identification framework of the tire road friction coefficient considering the effect of multiple unknown measurement noises

The tire-road friction coefficient (TRFC) directly determines the available traction and braking forces of the tires, which in turn has a significant impact on vehicle stability control, particularly for commercial vehicles such as heavy-duty trucks. However, onboard sensors typically cannot directly measure the exact TRFC. To obtain an accurate TRFC, estimation algorithms are used, which rely on data from onboard sensors combined with vehicle and tire models. Since the signals required for estimation come from various types of sensors, in practice accurately obtaining the noise statistical characteristics of all sensors is highly challenging. Additionally, due to the complex and variable nature of vehicle operating conditions, noise tends to be time-varying as a result of environmental factors, which inevitably affects the accuracy of the estimation. To address these problems, we propose a two-stage adaptive identification framework that combines the extended H-infinity Kalman filter (EHKF) with the adaptive unscented Kalman filter (AUKF). First, in situations where the noise statistical characteristics are unknown, EHKF and the tire model are used to accurately estimate forces on the front and rear axles. Second, considering the time-varying nature of the noise, the AUKF, along with the vehicle model and axial force information, is employed to estimate the TRFC for the front and rear wheels. Finally, simulation tests on various road surfaces demonstrate that the two-stage adaptive identification method outperforms the unscented Kalman filter in terms of accuracy and stability.