Bewertung der Sicherheit automatisierten Fahrens mittels des Fahrerverhaltenmodells „Stochastic Cognitive Model“

Autoren

A. Fries, F. Fahrenkrog - BMW AG

Zusammenfassung

The development of highly automated driving (AD) systems leads to the question of the systems’ impact on traffic safety. Answering this question in a prospective manner – i.e. before bringing the systems to market – is a key element for automotive manufacturers to prove the anticipated safety benefits of the technology which is requested by the German and the European Ethics Commission. In this process virtual simulations play a key role to investigate the safety implication of the technology before the market introduction of the AD system. The safety assessment of AD requires a time- as well as scenario-wise larger scope than the traditional approach for driver assistance system which relies on the re-simulation of reconstructed accidents. Therefore, a Monte-Carlo approach is applied that considers the stochastic variation of the scenarios parameters. Since the Monte-Carlo approach cannot rely on predefined accident trajectories as the traditional approach, this approach requires a highly sophisticated driver behavior model that is capable of covering the entire range from normal driving to handling critical situations in a realistic manner. These considerations and requirements lead to the development of the driver behavior model SCM (Stochastic Cognitive Model) by BMW. SCM combines stochastic approaches with cognitive ones in order to achive such a holistic driver behaviour model for motorways. In this paper, the SCM driver behaviour model is described in detail as well as the validation and verification (V&V) of the model is discussed.