While technology companies and automotive manufacturers are hard at work developing fully autonomous vehicles, the legal framework remains problematic. Who is liable for damages if the owner of the vehicle has no influence whatsoever on how the car is controlled? Amnon Shashua, CEO of the robotic auto company Mobileye, part of the Intel group, recently presented a solution.
He and his team presented a mathematical model that is said to ensure the safety of autonomous vehicles. Shashua makes clear, however, that absolute safety will never be achievable as long as human drivers are on the road. He therefore defines safety in terms of autonomous vehicles that never cause an accident, themselves.
This is where mathematics comes into play. A mathematical model is used to clarify who is at fault in case of an accident. This model is then used as the basis for software designed to control vehicles. The model would be obligatory for self-driving cars, thus ruling out any possibility that they might make a decision that could lead to them causing an accident.
Shashua considers his mathematical model to be the most practical way of preventing accidents. Instead of evaluating enormous volumes of data, each driving command would only need to be checked against predefined mathematical rules. This, alone, is supposed to be capable of keeping autonomous vehicles from causing accidents.
Even the head of Mobileye is aware, however, that any technical glitches – e.g. the failure of a sensor or a mechanical component – could still lead to accidents. He therefore suggests the use of three independent safety systems based on cameras, high-resolution maps, radar and lidar.
According to Mobileye's calculations , their system would significantly reduce the occurrence of fatal accidents from one for every million hours driven to one for every billion hours on the road. Based on U.S. accident statistics for 2016, this would mean only 40 traffic-related deaths compared with the actual number of 40,000.