Inversion of Vehicle Steering Dynamics with Modelica/Dymola

Abstract

The task of steering a vehicle is an exercise which is usually considered hierarchically in terms of the two subtasks path planning and path following. With the driver in the loop some essential man dependent tasks such as sensing, information processing, and motor function affect the steering quality. In case of simulations, the same applies correspondingly for driver models. In this paper the aim is to investigate vehicle steering dynamics independent of any driver-related properties. The path is therefore assumed given by a reference trajectory together with a speed profile. The steering angle which is necessary for exact or at least approximate path following is sought after. This allows for plausible comparative assessment of different vehicle’s steering dynamics in terms of the demanded steering effort for a certain maneuver. On the other hand, this approach requires dynamic inversion of vehicle steering dynamics which represents the main focus of this paper. Two vehicle models, the common single track model and a detailed model from the Modelica vehicle dynamics library are investigated. Since exact inversion of the detailed vehicle model turns out not to be feasible, approximate inversion is accomplished by means of a novel control structure called inverse disturbance observer. Simulations of a double lane change maneuver are conducted for illustration. Finally, wavelet power spectra of the steering angle signal are used for steering effort assessment.