Assuring autonomous operation: how can we know an autonomous vehicle will perform as expected, and what are the challenges to get there?
The notion of human control is central to traditional automotive functional safety models. But as vehicles move up the Society of Automotive Engineers (SAE) levels the validity of this approach is open to question. Machine learning (ML) is recognised as being critical to achieving higher levels of autonomy. But how do we know that the ML (and the vehicle) will perform as expected on the open road?
Safety assurance provides confidence that an autonomous vehicle will perform as expected. This workshop will consider a number of questions: What assurance is required? What assurance can be achieved? How can we achieve this assurance?
Professor McDermid will make suggestions on strategies for solving these challenges and draw out analogies with other domains before talking with delegates about their views and experience of assuring the safety of autonomous vehicles.