Modelling of road roughness for full-car models: a spectral factorization approach

Abstract

In this paper, we study modelling of road roughness for full-car models from power spectrum and coherence measurements of parallel tracks. The modelling problem is cast as a spectral factorization problem and solved via subspace-based identification algorithms. Two different spectral factorization problems are formulated. In the first formulation, four road inputs are directly identified in innovation form while in the second formulation, an excitation model for the front axle is first identified and then, the excitation model for the rear axle is derived from the front axle model via Pade approximation of a time-delay. The numerical study performed on real data shows that both factorization formulations have similar performances despite the direct approach yields less complex models.