Antonio Avenoso, executive director at the European Transport Safety Council discusses the hot topic of autonomous vehicle safety with Government Europa.
According to a European Commission report, Autonomous Cars: a big opportunity for European industry, the auto industry has set an objective to see semi- to high-autonomous cars in Europe by 2020. As we are now only one year away from this objective, it is imperative that we understand the autonomous vehicle safety.
Government Europa Quarterly speaks to Antonio Avenoso, executive director at the European Transport Safety Council, about the safety implications which still need to be considered before autonomous vehicles can be rolled out across Europe.
There is a strong push for autonomous vehicles to be on the roads sooner rather than later. Do you believe that having autonomous cars on the streets of Europe will be a lot safer and provide many benefits?
There is a lot of hype around autonomous vehicles; and while the European Transport Safety Council believes the technology has potential, we also think there is a very long way to go before there is a strong enough autonomous vehicle safety case for putting them on the streets of Europe.
Of course, there are the obvious benefits such as ending drink driving, distraction, fatigue and speeding. However, not all crashes are avoidable through automation. A child that runs out in front of an autonomous car may still be killed, even with the potentially faster reaction times of an automated system.
Looking at future EU vehicle safety regulation from where we sit in Brussels, we see proven life-saving autonomous vehicle safety technologies already on the market, that have not yet been filtered down to the whole vehicle fleet. Policymakers should be focused on mandatory fitment of driver assistance technologies, such as Automated Emergency Braking and Intelligent Speed Assistance, as well as improved crash protection. If the most effective safety features were on all cars today, we would already be saving thousands of lives.
The European Commission’s 2016 study evaluated that 94% of traffic incidents are down to human error. Do you believe that rolling out autonomous vehicles will help to solve this issue and reduce traffic incidents?
Many crashes that involve human error also involve other factors that may have still led to a crash, even if the human had not committed an error in judgement or misperception. Errors linked to poor roadway design (e.g. roads designed for lower speeds than legally allowed, confusing junction design, etc.) or faulty vehicle and interface design (confusing display, interfaces or visual obstruction) are often attributed to human causes when they are, in fact, design-induced errors.
Human error can also be non-driver-related errors, by pedestrians, cyclists and motorcyclists. Since they won’t be automated, their errors will probably not be eliminated by automation.
Automation will not cut road deaths by 94% any time soon. The real world figure is likely to be much lower. And in the short and medium term, autonomous vehicle safety failures may cause collisions that would not have otherwise occurred. The recent fatal crash involving a prototype Uber vehicle being a case in point; the car failed to avoid killing a pedestrian, pushing a bicycle across the road directly in front. The safety driver responsible for monitoring the system was watching a television programme at the time.
In the short term, it would be better to install autonomous vehicle safety systems that actively assist the driver, but still require the full attention of a human, with the driver remaining in charge at all times.
Who will be liable for an accidental crash/traffic incident involving an autonomous vehicle?
One of the reasons we see Event Data Recorders – known by the nickname ‘black boxes’ in the aviation sector – as important for the future of road vehicles, is because they will become very important in the context of crash investigation when autonomous vehicle safety is involved. It will become very important to fully understand exactly what the vehicle was doing in the time before a collision – and it will be essential that investigators have easy access to this data. This will also help resolve liability questions, as the system should record what automated systems were active at the time, or if the car was driving autonomously.
However, this raises other questions – such as under what circumstances will it be possible to pass control from the driver to the vehicle, and vice-versa in a safe way. This is a very grey area, and we think a lot more research is needed to better understand the risks.
What are the main autonomous vehicle safety concerns that need to be discussed and addressed to make implementation possible?
It is worth looking at the aviation sector, which now operates with a high degree of automation in a very controlled environment, with highly trained operators. Commercial passenger planes still require two pilots, and systems are supervised at all times. Yet, there are documented cases of automated systems malfunctioning, and human error linked to misinterpretation of an automated system’s behaviour. We have to learn lessons from this and ensure that public safety is the number one consideration in the development of autonomous vehicle safety on public roads.
What other complications and misunderstandings can automated or autonomous vehicles bring to the roads of Europe?
Another issue we are raising is that of the communication between an autonomous vehicle and its surrounding (non-automated) road users, i.e. pedestrians, cyclists, motorcyclists and people driving regular vehicles. How can automated vehicles signal their intentions to other road users? There is a lot of work to be done in this area, known as external human machine interface. This will be absolutely critical to autonomous vehicle safety in a mixed environment.
Similarly, as many different manufacturers and technology providers develop systems independently, how can we ensure that drivers understand precisely how automated systems in different vehicles work? Even now there are risks when you rent a vehicle; you have no idea what safety systems are on board, and how they work. This may not be an issue for autonomous vehicle safety, as long as these are backup systems designed only to avoid a collision in an emergency. However, when it comes to handover of manual to automated driving – it must be intuitive to the driver what they need to do, as well as how and when they need to do it. There are a multitude of highly complex challenges such as this that need to be overcome before these vehicles are approved for use on our roads.
European Transport Safety Council (ETSC)