Trust in AV: are motorists ready to put their faith in automation?

The evolution of automated vehicle technology is radically changing the game for independent safety testers, so too the rules of engagement. Testing organisations, working in close collaboration with carmakers, provide a crucial check and balance, and the safety assurance required to win consumer trust and drive the uptake of potentially life-saving systems.

Current testing is focused on advanced driver assistance systems (ADAS) and assumes the driver is always alert and ready to take back control of the wheel when required. As soon as the driver is removed from the equation, testing needs to be much more scientific and evaluate potentially millions more real-world scenarios in which the vehicle will be entirely responsible.

It is imperative the testing community is ready for automated vehicle’s imminent arrival. At stake is the one thing that automakers need more than anything else: the trust of the motoring public. Trust that manufacturers have rigorously tested their safety systems. Trust in the data they make available. Trust that testing standards will drive systems that are robust and prevent crashes. And ultimately trust that the automated systems on their car will not put them in danger.

Virtual simulation

Virtual modelling is becoming increasingly sophisticated. This gives huge advantages in terms of car design, stress and safety testing, and huge cost-reductions in respect of the ability to test software and hardware virtually before production and fitment. It is also a useful tool from a research perspective. For example, during physical testing, Thatcham Research’s passive safety crash tests utilise ubiquitous crash-test dummies. They may look sophisticated but are in fact relatively crude.

They use generic rules about bio fidelity and injury tolerance levels and they are basic in terms of assessing the impact of crashes on different morphologiesold and young, male and female and so on. They have served the industry well over the past 50 years and have helped significantly reduce serious injuries and fatalities. However, their simplicity means automotive risks missing injuries and not protecting people that do not look and behave like crash dummies.

The solution to overcoming this challenge lies in virtual testing and has been enabled by huge developments in computing power. The virtual-test world replaces crash dummies with virtual models. These ‘full-body models’ test the impact of forces on the human body, bones, muscles and organs, and accurately model how a human will interact and behave in the event of a crash.

Virtual tools are already proving their value to vehicle manufacturers for the design of new vehicles. But what about regulators and consumer testing bodies?

In the consumer testing arena, trust and independence are everything. Virtual safety tests need to be independently verified and assessed. Otherwise, how can motorists trust manufacturer models and their safety claims? Crash tests where safety is assessed during physical crashes accompanied by video evidenceare generally understood and trusted by the motoring public. This is not the case with virtual simulation, where consumers are effectively asked to put their trust in a computer.

Cultivating the trust of motorists in AV testing is crucial, both for testers and automakers

Then comes the problem of interpreting the data submitted by automakers. Consumer testing requires independent assessments using virtual machine (VM) data where the virtual vehicle models and coding are proprietary. Most VMs use dedicated computer networks for their virtual testing, and all are rightly concerned with protecting their valuable intellectual property. Manufacturers do not currently share these data models and if they did, they would be unreadable to the outside world.

Type Approval regulations do not currently compel manufacturers to share this modelling data or to supply it in a standardised format. With the complexities of active safety testing and especially automation, virtual testing is a prerequisite for measuring the performance of vehicles in myriad scenarios. This is something the UN is exploring through its VMAD framework, which addresses the use of virtual modelling to validate automated driving.

Thatcham Research sees potential in some form of technology interface, whereby manufacturers can upload the data needed to validate safety claims without revealing secrets about their intellectual property (IP). An agreed framework for input parameters and access to IP-free vehicle data would allow for more rigorous stress testing, engendering greater consumer confidence.

Physical testing

The eternal challenge for operators of vehicle safety test tracks is to recreate real-road authenticity in an artificial environment. This challenge becomes more complicated with the introduction of autonomous technology. Yet the current climate is one of uncertainty. In the absence of greater clarity from lawmakersand more first-hand experience with a broader range of autonomous systems as they begin to trickle into the marketplacetesters do not necessarily have the all the information needed to define clear strategies.

For example, testing organisations will increasingly need to give thought to how they accurately recreate motorway driving conditions. Do they construct replica motorways at their facilities? For urban tests, do they make a substantial investment in the construction of a realistic city-centre track? Or in both cases, can virtual simulation be employed to recreate these environments authentically? And if they go the virtual route, what will be the agreed testing parameters?

There certainly needs to be greater guidance on testing parameters to ensure a degree of uniformity the public can trust. For example, if we disable certain functionality to focus testing on individual systems, is this test holistically fair? How well can we simulate the presence of other traffic and the random factor this introduces without endangering lives on real roads? Above all, without a framework in place, how can we be sure all manufacturers and independent testers are working towards the same goals?

Automaker testing facilities are impressive, yet they too are wrestling with the technical challenges posed by autonomous vehicles (AVs). Moreover, it would be unhealthy for the industry to find itself in a position where manufacturers mark their own homework, essentially underwriting the safety of their AV tech.

Running before walking

The Thatcham Research view is that we already enjoy many of the safety benefits of automation through current assisted driving technology and motorists are relatively familiar with these driving systems. In the meantime, more collective work is needed on the virtual and physical testing environments to prepare them for AVs.

Cultivating the trust of motorists in AV testing is crucial, both for testers and automakers. In the same way that drivers trust consumer safety ratings from EuroNCAP and Thatcham Research today, we must put in place the building blocks of a trusted programme that meets the demands of automation tomorrow.

It would be unhealthy for the industry to find itself in a position where manufacturers mark their own homework, essentially underwriting the safety of their AV tech

If motorists do not see the benefits of AV technology in terms of road safety or read reports of AV-related accidents in the media, consumer trust and confidence will inevitably be eroded. Motorists may then turn their backs on the technology altogether. This way nobody gets to benefit from the safety advantages of an AV. The stakes are high and so we must get this right.

There are multiple stakeholders in the mix. Governments are keen to encourage AV rollout and the technology, alongside electrification, sits at the heart of all carmaker strategies. Yet this must be balanced against the interests of testing organisations, regulators and ultimately motorists. We must ensure that all those with a role to play are pulling in the same direction.

About the author: Matthew Avery is Director of Insurance Research at Thatcham Research