There is a lot of hype (or exciting development, depending on your perspective) around autonomous vehicles right now. Public acceptance of autonomous vehicles has increased quite a bit over the last two or three years. Still, a frequent cause of concern is: Are autonomous vehicles safe?
There are indeed reasons to believe so considering the prevalence of drink-driving, drowsiness, speeding and so on – behaviour not normally seen in autonomous vehicles.
Before going on I should point out that, in my view, the main benefit of automation is the potential for improved efficiency and productivity in the transport system, especially in commercial transports, where already partial automation can provide benefits.
Some are sceptic about a future where all vehicles are self-driving. But this is a common misconception about automation – the universal goal is not to make all vehicles self-driving. We will most likely have a transport system where vehicles of different levels of automation coexist (although not necessarily in the same space), each optimally tuned for its specific purpose. And that is the main point – automation is driven by smart solutions to mobility and transportation needs. So, the main driving force is efficiency – not road safety.
There are many challenges facing deployment of automated vehicles, some technology related, others related to business models and legislation. To me, safety is the primary challenge. If we disregard safety we could launch autonomous vehicles tomorrow. Obviously, we will not do that, so safety dictates the tempo of deployment. This is also a major reason why Volvo often cites confined areas as the first place where automation will be implemented – safety is easier to ensure here. The other major reasons are, pull from the market for high-productivity solutions and the fact that off-road legislation is more permitting of automation.
The highly-publicised death of a “driver” using an autopilot system in a car last May has really brought the issue of safety to the surface, with some people questioning whether we should be allowing autonomous vehicles on our roads at all. So let’s take a look at the question: will autonomous vehicles improve road safety?
We could just as easily ask: How safe are human drivers? How many would-be crashes do we avoid every day, just because we are humans? The quick answer: it is impossible to tell. Today, in the safest countries, a fatal crash occurs once in about every 300 million km driven (or 3 per 100 000 people per year)*. So, autonomous vehicles would probably need to be at least this safe. Although, I think people will find them acceptable only if they are in fact much safer.
Despite some fully autonomous vehicles having been tested successfully on road for several years, there is still not enough data to fully conclude scientifically how safe these (test-) vehicles are, so we have to do some informed speculation from here on.
In many ways, autonomous vehicles should be safer than humans. They will be designed to never speed, always keep a safe distance to other vehicles, and so on. Humans by contrast are fallible. This is why the majority of crashes involve the human factor, either the driver or another road user. We humans sometime overestimate our capabilities, we speed, we become tired or inattentive, and some even drink and drive. An autonomous vehicle will never do any of these things.
In addition, thanks to sophisticated sensors autonomous vehicles can keep track of many objects simultaneously and with great accuracy – which stands in stark contrast to humans who do things more or less in sequence: look left, look right, and so on, and who can only roughly estimate speed and distance.
So, in these regards they are indeed safer. But this is largely an ideal representation of an autonomous vehicle. Some of the necessary sensors are still under development or prohibitively expensive, and for the foreseeable future autonomous vehicles will need to share space with old-school vehicles in existing infrastructure.
Human beings, on the other hand, are fantastic at anticipating events in complex environments – far superior to computerized systems. So from a safety perspective we should seek the optimal mix between human and machine. Therefore, Volvo’s automation approach is human-centric and automation is designed to assist an operator or driver.
Ensuring safety in a partially unpredictable road environment with pedestrians, potholes and severe weather, is clearly a great challenge, and this will demand autonomous vehicles that are defensive and always sensibly cautious. Crashes due to freak events like a fallen tree across the road can still occur in the future, just as today, but thanks to active safety and connectivity the consequences can be mitigated to a much greater extent than today.
For similar reasons manufacturers are addressing cyber-security threats and developing failsafe solutions using redundancies in critical components so that – should a critical event occur – the consequences are not catastrophic.
As the autopilot-case proves, one of the biggest challenges in automation is to ensure the user fully understands what the system does and doesn’t do at any given time, so that unsafe behaviour is not encouraged by the design. Interestingly, removing the human from the driving task seems to actually demand an even greater attention to the human factors.
And oh – I know what you were thinking – the answer is yes, it seems reasonable to me that manufacturers should assume the main responsibility if an incident occurs as a result of a failure of the vehicle while in autonomous mode. It follows from having a product safety responsibility of the products, i.e. that they are safe when used as intended.
Active safety systems that assist the driver are already widely used today, and through this low level of automation we are enhancing human performance significantly. This to me represents one of the areas that are easily forgotten in the debate. Every new automation application will require the automotive industry to develop new and improved safety systems. And this development will not be exclusive to autonomous vehicles, it will benefit all vehicles. So, in terms of safety perhaps the journey is more important than the goal.
There are other ways as well in which automation can improve safety. Automation can reduce the need for human presence in potentially hazardous environments, significantly reducing health and safety risks. You may have seen this demonstrated by Volvo’s self-driving mining truck, wheel loader and articulated hauler, showcased a few months ago.
Automation becomes in a way a system approach to safety as it stimulates holistic changes to the transport system, from policy to communication technology and infrastructure. Today we try to do our individual best to avoid accidents, with automation we will benefit from a virtual co-pilot and air traffic controller who assist us at all times.
So, then, will autonomous vehicles improve safety? Add all the elements together and you have a strong case for automation improving safety when widely adopted – albeit not without its own set of risks and challenges. Both thanks to the development of the safety technology it stimulates, and through the benefits it may deliver when widely deployed. But to fulfil this potential we need to have regulation and traffic systems that cater to and are able to fully exploit the benefits of automated vehicle technology.
Done right we will be able to reap all the benefits that automation offers in terms of productivity and efficiency while enhancing safety. And every step of the way will push us to develop even better safety solutions, which will eventually benefit all vehicles. And that to me sounds like a really promising development for road safety.
*OECD ITF Road Safety Annual Report 2016