As robotaxi pilots expand across cities, proving that autonomous vehicles are safe is becoming a more complex task than simply counting crashes. Crash statistics remain essential, but they capture only one part of how an autonomous vehicle interacts with the road environment and the wider transport system.

Recent comparisons of robotaxi crash rates provide encouraging evidence. One US study found that Waymo vehicles had a 68% lower crash rate than human drivers. However, such figures need to be interpreted carefully because autonomous fleets currently operate within limited geographical areas, specific speed ranges and defined operating conditions. A result from one deployment cannot automatically be generalised to every road or situation.

A broader safety assessment should also consider how autonomous vehicles behave during unusual road events. Interactions with emergency responders, police, school buses and temporary traffic restrictions can create situations that are not necessarily recorded as crashes but can still affect public safety. For an autonomous system, recognising that an ambulance or fire truck needs access—and responding appropriately—can be just as important as avoiding a collision.

Another emerging issue is the cost of false alarms. Robotaxis may sometimes call emergency services because a passenger appears unresponsive, even when the person is simply asleep or impaired. At larger deployment scales, repeated false alarms could consume emergency-service resources and create costs for cities.

This points towards a more comprehensive concept of roadworthiness. Safety metrics could eventually need to include not only collisions and injuries, but also traffic disruption, emergency-response interactions, false alarms and other forms of operational behaviour. Such indicators could help regulators understand the real-world consequences of autonomous mobility before systems are deployed at scale.

For Europe, this broader perspective is particularly relevant as international regulatory work on automated driving develops. The future framework for autonomous vehicles will need to evaluate not just whether they can drive safely, but whether they can behave predictably within the complex social and operational environment of a city.

GOSWAMI, Partha. OPINION: Safety metrics and autonomous vehicle deployment. Autonomous Vehicle International [online]. 2026 [cited 2026-09-18]. Available from: https://www.autonomousvehicleinternational.com/news/safety/opinion-safety-metrics-and-autonomous-vehicle-deployment.html

Obrázek: chat gpt