This paper was produced for the 2019 NAFEMS World Congress in Quebec Canada
Over the last decade, automotive OEMs have increased their investments towards the development of Advanced Driver Assistance Systems (ADAS). The intention is to augment driver behaviour towards increased safety and comfort. During the development cycle of an ADAS, a simulation model of the vehicle is required to be inserted in the loop with the controller under design. Therefore, since the beginning of an ADAS design process, the adopted vehicle simulation model, and its fidelity, influences the results of the controller. Very often, control engineers make use of simplified models, in which several physical phenomena and dynamics effects are not included , meaning that from the analysis of the results it is not possible to evaluate the interaction between the vehicle and the controller that are due to the neglected dynamics.
This study aims at underlining the effects of the different vehicle model fidelity on the results of typical simulations performed during the development of an ADAS. The benchmark model used in this analysis is a high fidelity multibody model. The model comprises of rigid bodies connected together by means of ideal joints. The total number of degrees of freedom of a typical vehicle model is greater than 150. This model is combined with the semi-empirical tyre model  (MF-Tyre/MF-Swift) for realistic modelling of the contact forces and torques between the tyre patch and the road. The high fidelity benchmark model is compared with simplified models. The first model considered is a planar model for simulating the longitudinal dynamics of a vehicle. The other simplified vehicle models are dedicated to the simulation of the lateral motion and are the single track model and the double track model. These simplified model cannot fully reproduce the complex dynamics of the vehicle, but they are very commonly adopted by engineers in the design phase of a control algorithm due to their simplicity and computational speed.
The scenarios investigated are derived from typical simulations needed during the development of ADAS. The first scenario is a pure longitudinal dynamics manoeuver, in which the vehicle proceeding at constant speed, performs a braking manoeuver until standstill. The second scenario is a purely lateral dynamics manoeuvre, the vehicle takes a turn while moving at constant speed. Finally a third scenario is considered, in which both longitudinal and lateral dynamics are combined. The vehicle starts from standstill and accelerates while turning.
The results show that the simplified models perform quite well for the reproduction of the vehicle trajectories and velocity profiles. The estimation of the roll and pitch angles, although less accurate, are still comparable with the results of the multibody model. From the analysis of the results it can be concluded that, as long as the scenarios to simulate are within the boundary conditions of the simplified models, the results are comparable. Further investigation will be done in a range of more advanced manoeuvers, in which the non-linear dynamics of the vehicle will be more relevant and the differences between high fidelity and simplified models will be more evident.