Parameter Identification of a Planetary Rover Wheel-Soil Contact Model via a Bayesian Approach

Abstract

Soft soil contact models developed for planetary exploration rovers play an important role in the study of rover mobility. Nowadays, most of the existing contact models are based on Bekker theory, which requires the evaluation of several soil parameters usually measured via bevameter tests. However, substantial differences existing between the plate-soil contact scenario and the wheel-soil contact scenario, along with large variability associated with the bevameter experiments, give rise to large uncertainty in the choice of the model parameter values. In this paper, a Bayesian procedure is proposed to deal effectively with the presence of uncertainty. In the proposed approach, model parameters are random variables with prior distributions derived from bevameter measurements. The prior distributions are then enhanced to posterior distributions through single wheel test data. At the end, the procedure identifies a set of possible model parameter configurations that result in high experimental-numerical matching.