Supported by results from the recent Drive Me research project, this talk will present Volvo Cars approach to determining the potential safety impact of autonomous vehicles. We identify four main safety challenges, (1) development of a safety impact assessment methodology, (2) development of a pre-crash scenario-based approach, (3) definition of target levels for safety performance, and (4) development of crash avoidance performance testing methods.
Thereafter, we go through how these challenges are addressed by the Volvo Cars Holistic Safety Impact Assessment Framework (see image). We explain how we create representative conflict situations within the ODD, how we determine safety target levels for injury prevention and crash avoidance. We propose a solution to the “what is safe enough” riddle by using the Reference Driver (an attentive, skilled, experienced driver model).
Results from the application of this framework will be presented, showing the performance of an AD vehicle in comparison with the Reference Driver and other references. (We believe that this is the first time this type of crash avoidance results has ever been presented.) Descriptions of new safety methods and procedures will be presented as well.
Our research has points to three myths and we propose solutions for them: Myth of the mileage-driven approach, Myth of driver capabilities, and Myth of unidimensional automation.