Autonomous UK? Maybe not.

The UK automated vehicles act is intended to position the UK at the centre of the AV revolution. But how well thought through is it? And will it actually work?

Turning base regulations into gold

On May 20, 2024, the UK Automated Vehicles Act 2024 received Royal Assent in the Kings Speech. The Act is intended to boost the UK's economy by making it a leader in the self-driving vehicle industry. The measures will come into effect in stages ​with full implementation by 2026. Many are buoyant about the economic opportunities here. The act has led to confident statements that fully autonomous vehicles will be a regular sight on the UK’s roads within two years. But will they? What alchemy has this Act performed to enable this leap of technology to happen imminently on the UK’s shores?

The sky’s the limit

The Act establishes a framework for the deployment and regulation of self-driving vehicles in the UK​. The intent here is to foster public trust as advanced autonomous technologies merge into the everyday UK transport network and take over the driving task from the UK’s populous.

The primary question to ask here is how people’s safety will be practically and effectively assured. This is an issue of science and engineering. A secondary, more subtle question then follows: How will public trust in safety be gained?

We all know that people are error-prone and therefore logically it should be possible to argue that effective AV technology is safe. However people often aren’t logical. Drivers tend to value the sense of control they have in their car, even when they rationally understand that driving on a road exposes them to risks beyond their control. Also drivers don’t generally believe that they’re the ones who are likely to cause an accident, as they consistently overestimate their driving ability.

In terms of winning the public trust, the thinking seems to go like this:
Modern planes are highly automated too, but - despite the obvious risks - passengers put their lives in the hands of the plane manufacturers and maintainers all the time. The new laws proposed are therefore modelled heavily on those in place in the aviation sector.

Globally at least as good as

Those of us who are students of risk acceptance will note that Section 2(2) of the Act states the principle that:

authorised automated vehicles will achieve a level of safety equivalent to, or higher than, that of careful and competent human drivers.

As I read it, this is a ground-breaking statement. Fundamentally it appears to be a departure from the UK risk acceptance principle that safety must be managed ‘so far as is reasonably practicable’ (SFAIRP) which ultimately means that risk reduction is made until the ‘time, trouble and effort’ is too much to reduce it further. This revised risk acceptance criteria is actually something more akin to the French GAMAB (globalement au moins aussi bon) principle of establishing whole system safety risk equivalence.

This causes my first stumbling block. It’s not immediately clear to me how this equivalence will be established. Estimating the failure rate of a complex engineered system is fundamentally different - qualitatively and quantitatively - from estimating the human error associated with a driver. In particular people are not prone to systematic failures and autonomous vehicles would be.

Risk estimation for automotive safety currently benefits from extensive experience, and a large, relevant store of applicable data. However, engineered systems could fail in significant, novel and unpredictable ways, making the assessment of risk complicated and prone to uncertainty. Large scale failures - to multiple vehicles - also become possible through software errors or cyber attacks. Over-the-air software updates of vehicles might be necessary to resolve the former but they could greatly increase the ‘attack surface’ for the latter.

So the safety case argument for an AV won’t be easy to make against this new criteria, particularly in the early days when there is little in service experience. I sense that a lot more people are going to need to become au fait with the automotive software safety standard ISO26262 too.

Another concern with the approach to risk acceptance is that, unlike the ALARP principle, it doesn’t stress the need for continuous improvement. Once AVs have been approved and accepted as equivalent to current levels of safety where will the drive for further safety improvement come from?

Realigning responsibilities

There are two categories of Automated Vehicle (AV) that will be able to be licensed for use on the UK’s roads. A user-in-charge vehicle can flip between driven and self-driving modes. In a ‘no-user-in-charge’ journey no occupant is a driver during the journey. In some cases, the vehicle may not even have a steering wheel. Critically, in either case, when a vehicle is driving itself, the driver wouldn’t be responsible for the vehicles behaviour; the manufacturer would.

This flipping of liability seems to have caused a number of people concern, in particular the motor insurers. One particular concern for them is that, in this new world of accountability, should the vehicle malfunction the driver is a potential claimant - and one who is very exposed to any risk that might arise. The critical question for liability is then: What mode was the vehicle operating in when it crashed? This in turn makes it important that, following an accident the vehicle insurer (as well as the investigator) is able to access system data that proves whether or not the vehicle itself was at fault.

It is easy to imagine how a given manufacturer might not be inclined to provide this transparency. How easy would it be for UK insurers to get key data and design information relating to a specific technical fault from a vehicle made in, say, mainland Europe. Or China?

The reality is that much of the underlying design information might not be readily available in advance either. Manufacturers will want a degree of ‘grandfather rights.’ They will also want the ability to minimise assurance costs on the basis that their technology is approved in another country, as happens in aviation where there are global standards to facilitate this.

Another way of thinking about the issue is that where vehicles are designed and manufactured overseas this will most likely offshore the causal factors for major accidents. So, enforcement would also need to go beyond UK borders - and the regulations describe the proposed framework for this.

Will other countries welcome UK regulators and investigators digging into the systems of their ‘national champion’ automotive manufacturers to establish the liability for accidents? The manufacturer’s concern might also spread to intellectual property or cyber security. Would the UK regulators and inspectors be able to develop and maintain the highly technically competent staff they would need to conduct these investigations in any case?

Planes, trains and automobiles

As regards overseas manufacturers, the key question for me is whether or not there are manufacturers out there who are both technically capable enough to take on the practical, scientific and engineering challenges here whilst also being prepared to take on the significant costs and liabilities that the regulation would impose on them. In a parallel sector, the railway trade-body UNIFE recently wrote a public letter stating that it was seeking exemptions for railway network extensions from the EU cyber resilience act on the basis that compliance:

would increase the investment costs (that are already high) to an extent the projects would no longer be economically viable.

If that’s the view of transport technology providers, regarding EU wide laws, relating to known risks, would similar manufacturers be likely to take on more uncertain risks, to meet the local laws of a European ‘third country’?

Some of these challenges might be headed off by a homegrown British AV company, for whom the regulatory and assurance requirements would be more meaningful. But a review of UK AV start-ups appears to me to show that they are, in general, developing their technology to be built into the various, existing manufacturers’ vehicle platforms, as ‘software as a service.’ This creates another critical interface, and raises additional questions about the boundaries of responsibility and liability.

And while I am in a challenging mood: who is going to invest in the development of the competent maintainers - of the type that exist in the aviation sector - to look after these new highly complex, safety critical machines? This has the potential to be a drag on upscaling, much like the lack of charging infrastructure has hit the roll out and uptake of electric vehicles.

The transition to self-driving private vehicles is not just a legal leap forward, it’s a scientific, philosophical and behavioural one. There is also a certain amount of inevitability to it. As election fever grips the nation all UK political parties seem to be promoting the idea that economic growth fuelled by technology is the only way to repair Britain’s flailing economy. That is, of course, creating a drive to find the Brexit dividend that was promised to the UK electorate, by using its new law making ability to innovate. Speaking as a Brit, I really hope we can do so. We do have cutting edge scientific capabilities at our disposal.

But fools rush in where angel’s fear to tread. Innovation is linked intrinsically to risk-taking. And as the risks - both commercial and safety - will ultimately fall on UK taxpayers and drivers, we all need to be careful that the desperate push here does not create unintended consequences.

I’m keen to build the network for Tech Safe Transport. If you know anyone who is interested in the safety of modern transport technology, and who likes a thought provoking read every few weeks, please do share a link with them.

Thanks for reading

All views here are my own. Please feel free to feed back any thoughts or comments and please do feel free to drop me an e-mail. My particular area of professional interest is practical risk management and assurance of new technology. I’m always keen to engage on interesting projects or research in this area.