Even if self-driving vehicles can become much safer than vehicles with human drivers, in a possible future of self-driving cars, still traffic accidents cannot be ruled out entirely. In the aftermath of an accident, many questions should be answered in order to be able to answer liability issues. Probably even more importantly, we should learn from such a situation in order to avoid accidents in the future.
Event Data Recorders (EDRs), popularly known as “black boxes”, are already commonly used in self-driving vehicles. However, in most cases only basic observational (sensory) information is stored. This information cannot not tell us how information is used by the vehicle to make a decision and whether the decided action has been performed correctly. TNO has recommended which additional information should be stored in EDRs to be able to assess the full chain of events including observation, processing, performance and context.
The recent study TNO conducted suggests that possible future legislation of EDRs should take into account cost-effectiveness of the implementation and the rapid advances in AI-based systems. Exceedingly high demands on the amount- and type of data to be collected may hinder future advancement. Also, questions surrounding ownership of in-car data and privacy of passengers are still unclear and require better regulation.
As part of the ‘Study on Safety of non-embedded software’ TNO, VVA, and SSSA analysed the legal and business landscape related to new advanced technologies associated with digitisation and AI. Among others, TNO provided a prospective foresight study on specifications of Event Data Recorders (EDRs) in road vehicles and other AI-based systems in fields such as industrial robots and autonomous shipping. The recorded data can be used for multiple purposes, for instance, training, safety assessment, surveillance, causation, vehicle diagnostics, testing and development.
SOURCE: TNO
