“Accidents will happen” – so let’s learn from them

When developing new technologies such as autonomous vehicles (AVs), there will inevitably be situations where despite the maximum care being taken, accidents happen. These might happen in a variety of situations, but accidents involving autonomous vehicles are likely to be different from those involving only ‘standard’ cars – for example if an AV hits a pedestrian and there are no witnesses. In order to ensure that self-driving cars can be deployed safely, we need to think ahead to how to deal with accidents involving them, and to ensure that we can learn from such accidents in order to try and prevent recurrences.

The RoAD project (Responsible AV Data) is seeking to ensure that when an AV is involved in an accident, the data from its on-board data recorder can be used to determine what happened, help to decide who is responsible, and hopefully prevent similar accidents happening in the future.

Currently, on-board data recorders (black boxes) are not necessarily set up to meet these objectives – instead they may be recording data that is only useful to the car’s manufacturer. Even more problematically, the data may be proprietary and the manufacturer may not have to release it to an insurer or a court, and even if they did so the data might not be useful in determining what happened. There are no universally accepted standards for what they are or should be recording.

Our project seeks to not only examine the type of data that the black box collects, but also to determine who could or should access it, and whether it could be used by an insurer or a court. The project will create simulations of crashes that we can use to test what data should be recorded, and will work with insurance companies and legal experts to establish whether this data will be sufficient to determine what happened and who is responsible. We will also be looking at other issues around data collection, such as members of the public being recorded by car cameras, how long data is kept, and whether data is stored securely. We will create recommendations for standards around data collection, use, and storage, to help ensure that AVs are safe, accountable and trustworthy

If you would like to be part of our research project, we will shortly be gathering views and opinions from members of the public about self-driving cars and the data they collect. To participate, please send an email to roadpublicsurvey@cs.ox.ac.uk and we will send you a link when the survey begins.

Carolyn Ten Holter