Calibration and validation of a tire–snow interaction model

Abstract

We address the challenge of the validation of models for a vehicle interacting with a natural snowy terrain by applying a rigorous statistical framework. Gaussian process-based stochastic metamodels were used to fit noisy test data in drawbar pull and traction as a function of slip, and to transform the deterministic physically-based tire–snow interaction model into a stochastic one. Important parameters such as the mechanical properties of snow, the coefficient of friction between the tire and snow, and the depth of snow were obtained using a Gaussian maximum likelihood method. The uncertainties of parameters, and prediction using calibrated parameters for front and rear wheels were quantified and assessed using interval-based local and global validation metrics between models and test data. Overall agreement between models and test data is good.