« back to the Hub

Will we ever trust computers to drive us safely?

Much is claimed for the “driverless” cars that are making news around the world. On 16th March, Venturefest Bristol and Bath (VFBB17) held its first Roundtable to discuss the key issues around these Connected and Autonomous Vehicles (CAVs). Insiders and experts from the major industries, insurance, vehicle and cyber security, as well as government, debated if and how these vehicles can gain widespread public trust.

One possibly surprising theme that emerged was that it currently isn’t clear how to test and certify these autonomous vehicles are safe for road use. It seems the global standard used to test cars’ functional safety, ISO26262, doesn’t yet cover the type of decision-making software used to control CAVs. Certifying that this type of software not only correctly controls the car, but that it makes the right decisions, is a serious challenge, made even more complex if each car “learns” as it is used over time, or if continual learning is shared across several vehicles.

Fortunately, the VENTURER and FLOURISH projects based in the region are working on solutions, and the assembled group provided a cascade of suggestions, some hotly debated. For the software itself, very deep and extensive simulation testing would significantly shorten timescales. But a capability completely independent of any manufacturer or supplier would likely be mandatory. Such a centre of excellence would need an international standing to bring in best practice from all manufacturers and countries.

Of course, the whole vehicle must be certified, not just the software, which suggests that the future maintenance and repair of these vehicles will be substantially different from today, prompting the suggestion of an “MOT on Steroids” capable of testing and re-certifying both software and electronic systems annually or after repair.

Although connected cars offer many benefits from ride sharing to dynamic traffic management, the current vulnerability to cyber threats is well-known. Manufacturers are rushing to patch issues, and pushing to develop “cyber hardened” systems, an absolute requirement for software that is a critical part of the safety system, and driving the car. However, in-depth experience of computer security suggests that making cars tamper-proof won’t be enough. These vehicles will be part of a wide network that must be resilient against threats even from nodes within it.

A safety rating for autonomy was a great suggestion, similar to the NCAP ratings for insuring new cars. There were numerous ideas for aspects that need to be tested, and how, but the most debated questions was… who?

Is it the government’s responsibility to define how these vehicles, and their software, must be tested? Won’t that stifle innovation and technology businesses? Would it be down to the motor insurance community to lead this, as they do for NCAP? Can manufacturers and technology suppliers gain a market advantage from a reputation for safety and security? Although some cited evidence that this is be possible, others were more sceptical and thought that legislation would be needed, probably driven by consumer pressure. A similar group will debate which business models can succeed, and which will fall by the wayside, at the third Roundtable in September: “Who will be the winners and losers in the future world of connected and automated vehicles”.

However, prompted by the VFBB17 theme of Smart City Challenges and the tech solutions, the next Roundtable in June will take on the thorny question of: “Will these connected and autonomous vehicles really reduce congestion and pollution in cities and urban areas?”

With limited places, please contact us if you’d like to be part of the debate.

invest@venturefestbristolandbath.com / www.venturefestbristolandbath.com / @venturefestbb / #VFBB17

John McNicol of Nova Modus, working in partnership with Venturer, Flourish and Venturefest Bristol and Bath

Category: news