Comparing vehicle dynamics models of different complexity for comfortable and emergency steering in virtual assessment of active safety systems

Abstract

The World Health Organization (WHO) states that every year, 1.35 million people die as a result of road traffic crashes [1]. According to the National Highway Traffic Safety Administration (NHTSA), rear-end collisions are the type of crashes that occur most frequently. Actually, 29 percent of all collisions are rear-end crashes [2]. It is the responsibility of the automotive sector to make use of the technologies available to the full extent, to reduce the number of fatalities caused every year due to road traffic crashes. The development of Advanced Driver Assistance Systems (ADAS) has led to the prevention or mitigation of many collisions. Therefore, testing of active safety systems plays a vital role in improving these systems. Rigorous analysis under different conditions can provide insight to further development of these systems. Development of active safety and automated driving systems requires safety evaluation during the development stage by using different tools for testing and evaluation. One such tool for evaluation of active safety systems is performed using computer simulations virtually, as evaluation of these systems in real traffic scenarios is not only dangerous but also expensive. Since vehicle models take part in simulations, it is important to study the impact of vehicle models of different complexities for evaluating active safety systems. The main goal of the thesis is to select different vehicle models of varying complexity levels and study what impact the selection has on the last point of steering to avoid a stationary obstacle while performing evasive manoeuvring. In the increasing order of their complexities, a single-track model, a two-track model without load transfer, and a two-track model with load transfer, were selected. These models were further subjected to changes by considering different longitudinal speed, different cornering stiffness at the front and rear axles of the vehicle, and different tire models. Analysis was performed in terms of comparing the timing of the point of no return (last time to avoid the obstacle) where no substantial difference was noticed among the different vehicle models. Finally, based on considering the deviations in the last point of steering in seconds and computational time required for 10,000 iterations, the conclusion is made that the most complex model is not always necessarily the best model that can be used in simulations.