Online estimation of friction coefficients of winter road surfaces using the unscented Kalman filter

An innovation to a previously proposed method for estimating friction coefficients of winter road surfaces was achieved through the introduction of a relatively new algorithm, the unscented Kalman filter. Its use, instead of a generic algorithm, made estimating friction coefficients in real time possible while keeping the core vehicular motion model unchanged. The problem of estimating such coefficients was too complicated to apply conventional feedback techniques, such as an extended Kalman filter, because of the presence of not only a nonlinear algebra equation but also a set of multiple differential equations. The unscented Kalman filter did not require any explicit function of state and observation equations in deriving Kalman gain. This paper describes the usefulness of this new filter in solving the problem and describes numerical experiments that validated the effectiveness of the proposed new method in terms of computational efficiencies. The friction coefficients estimated with the new technique were in fairly good accordance with those measured in the field.