The recent death of a pedestrian caused by a test vehicle in Arizona brought to light the unique challenge that safety presents to the autonomous vehicle community. To date, no universally acceptable methodologies for implementing and validating safety or safe-by-design approaches have emerged. The AV market must develop safety solutions or it will flounder. Europe’s strict safety and privacy laws have catalyzed leading safety research at European universities and industry in new directions. While traditional functional safety standards such as ISO26262 focus on technical safety (i.e., faults do not cause malfunction that threaten safety), there are no standards that define the safety of intended functionality -the overall behavior of a vehicle is safe in the absence of faults.
Promising approaches are emerging that focus on assuring safety in vehicles using real-time, in-line, safe-by-design techniques, combined with offline validation of software and systems, both from connected vehicle and infrastructure perspectives. They build on dependable, near-optimal perception and cooperative decision making in a multi-agent and multi-modal transport scenario, anticipating alternative European concepts. However, it is unclear how to translate these concepts into actual vehicles that reliably work in unpredictable real-world road situations.