Event

There is currently a lot of focus on how to do safety assurance and how to estimate the safety impact of Autonomous Driving Systems (ADS). One approach that has been suggested is to use traffic simulations to “generate” crashes by simulating millions to billions of kilometers of driving in virtual environments. The approach requires driver models for human drivers, as well as reasonable implementations of the ADSs and several other components (infrastructure, vehicles models, etc). There is however quite a bit of critique and call for caution of using traffic simulation for safety benefit assessment – in large parts arguing that today’s driver models (for normal everyday driving, but also models of driver responses to critical situations and how to extrapolate to future drivers) are not good enough, and that just minor difference in the models can have large impacts on the safety benefit assessment. It is not obvious that the crashes that are “generated” are realistic, nor that the rate of scenario occurrence is good enough for ADS benefit assessment. How should you validate the methods? What are traffic simulations useful for? Also, alternative methods to assess the safety impact of ADS will be briefly discussed, but focus will be on traffic simulations.

In summary, this seminar will discuss the pros and cons of the use of traffic simulations for safety benefit assessment of ADS. We will likely not come to a consensus, but it will bring many of the pieces of the puzzle on the table, after which we can start putting the pieces (incl. experiences across the SAFER partners and beyond) of the puzzle together – maybe initiating further collaboration or inspiring new research.   

The workshop is facilitated by Dr. Linus Wågström, Research area director for Safety performance evaluation at SAFER and Dr. Jonas Bärgman, Chalmers University of Technology. The workshop will be conducted as a hybrid event and you can participate either online or onsite at SAFER, you register your preferred alternative in the sign-up form below.

Welcome!