Intelligent observer-based road surface condition detection and identification

Road surface condition is greatly dependent on the surface´s friction coefficient. The abrupt change of the coefficient results in variation of wheel slip which likely leads to vehicle instability. Although the vehicle on-board sensors can measure the vehicle´s velocities and yaw rate, the measurements, often containing noise and drift, are limited to the surface that the vehicle is engaged. In contrast, an effective observer can be used to estimate the vehicle dynamics for all possible surface conditions. This paper proposes a new observer, called extended state observer (ESO) to estimate the three quantities, and more importantly an additional quantity known as system dynamics. With the aid of the ESO, the following three tasks are performed: (1) noise filtering from the measurement data (2) detection and classification of surface condition change, and (3) identification of the road surface. Fuzzy logic was employed to quickly detect the change of road surface condition and further classify the surface; a neural network was employed to help determine the friction coefficient. The dynamic model used in this study can be applied to four-wheel independent drive vehicles. The presented methods were simulated when a vehicle encountered a significant change from a uniform-mu (i.e. uniform friction coefficient) surface to a split-mu surface (i.e. different friction coefficient on each side of the wheels) during cornering.