The technology behind autonomous vehicles is already hitting the streets – but what more is needed for us to enter the age of driverless cars?
To date, driverless cars are usually seen as an emerging technology. But much of the underlying technology – from “autopilot” systems to parking assistance – is already being rolled out.
In recent months we have seen that this technology is not perfect, with the first reported fatality attributed to autonomous vehicle technology failure. Things can also go wrong with driverless systems due to human error. But the basic technology has been developed, and now a new, subtler set of challenges will take center stage.
As the quest for driverless vehicles rolls forward, stakeholders need to address mounting concerns, including: How can companies behind driverless tech make us feel safe? How can they get the balance right between real-life situations that require the car to be in control and those where the driver wants control? How can they make cars adapt to individuals’ driving styles?
In short, what – beyond the technology itself – will it take for driverless cars to become a reality?
The answers to that question are now becoming clear – and there are four key steps to get us there:
1) Learning from aviation
All parties that stand to benefit from making driverless vehicles a commercial success can learn a great deal from the aviation industry.
Planes are safe and the fear of mechanical failure is low. The fatality rate for planes is much, much lower than that of trains and cars: 0.003 for every billion kilometers traveled for planes compared to 0.27 fatalities per billion kilometers traveled by rail, and 2.57 fatalities per billion kilometers traveled by car.
The key point is that consumers trust air travel. We don’t draw a distinction between those parts of the flight when the plane is in autopilot mode and those where the pilot is a human. As passengers, we are equally comfortable with both.
Auto companies, dealers and marketers would do well to learn lessons from how flight has been marketed to make flyers feel so safe.
2) Educating and incentivizing customers
Humans are not particularly safe drivers. Ninety percent of accidents are caused by human error. Cars already on the roads that reduce reliance on human attentiveness with forward collision warning systems and automatic braking features have been shown to reduce collisions by up to 15%, according to a study by the Insurance Institute for Highway Safety.
The premise of driverless vehicles is that human driving is an inconvenience and inefficiency, and perhaps even a hazard, which should be cut out of the process of getting someone from A to B.
But this is only part of the story. Driving also offers the driver a sense of control over a potentially unsafe environment.
As a result, even if relinquishing control over a car to a machine is safer, it can feel unsafe, and this is a real barrier to driverless cars becoming a reality. Indeed, a recent EY survey found that 54% of drivers were worried about the prospect of traveling in a driverless car.
Like any new technology, this discomfort is likely to subside once people experience it for themselves. This means that both the dealers and the automakers must play a role in educating and incentivizing prospective customers on autonomous features and technology, and giving them a chance to try the vehicles for themselves.
Driverless taxis are likely to be a vital transitional step to driverless car uptake – proving the viability of the technology in the real world, and giving people a chance to try it for themselves without having to buy their own driverless vehicle. Driverless taxis have just rolled out in Singapore, and are planned for deployment in Japan in time for the 2020 Tokyo Olympics.
3) Incremental improvement
Another lesson regarding what it will take for driverless cars to become a reality is that if you want to get consumers and drivers to feel accustomed to something, introduce it incrementally. In the case of automation in the car, the process has started.
Cars now help us park. Parking sensors on the bumpers detect how far you are from an obstacle and beep to let you know. Some cars include graphical displays for the same purpose. Some cars include reversing cameras, some of which show a 360 degree image.
In April 2014, the National Highway Traffic Safety Administration proposed that all new light vehicles have ‘rear-view visibility systems’, in effect, requiring backup cameras. If the proposal is adopted, these cameras may end up being mandatory in some regions.
Then there is parking assist. These systems work out whether a parking space is big enough for the car or not, tell the driver, and if so, help them steer the car into the space. The driver just controls the speed.
Many can now do different types of parking too: parallel parking and parking in a bay, for example. Forward collision systems and automatic braking are all part of a trend of passing over control of the car from the driver to the car itself.
The lesson for those manufacturing and marketing driverless vehicles is to make driverless cars part of a spectrum of incremental improvements as much as possible.
4) How much human do driverless cars need?
Even after persuading people to trust driverless technology, there are additional challenges – not least of which is the mismatch in driving styles between human and self-driving cars. This will be a particular challenge as long as roads are shared by driverless and human-controlled vehicles.
Driverless cars drive in a different style from human drivers. They drive like computers that have been programmed to follow the letter of the law. Despite that, a 2015 study that investigated a small sample of driverless cars found that they had a higher crash rate per million miles traveled than conventional cars.
The twist was that this wasn’t because they were unsafe. Quite the opposite: almost all the accidents in the study were caused by human drivers unaccustomed to responding to a robotic driving style. In other words, human drivers are unaccustomed to responding to driverless cars’ ultra-safe driving style.
The solution is a convergence. Humans may have to adapt their driving style. But there is also a strong argument that driverless vehicles should be programmed to mimic human driving behavior to help them better adapt to the irrational behavior of human drivers.
Learning systems could also mean individual cars can tailor their driving styles to best accommodate styles with which the occupant feels most comfortable.
“The processing systems used for autonomous vehicles are expected to rely on advances in ‘machine learning’ to better mimic the human brain’s ability to deal with unique situations,” explains Randy Miller, EY Global Automotive & Transportation Leader.
“The software of a fully autonomous vehicle will need to be adaptive, intuitive and self-learning, like a chess supercomputer that learns from its opponents’ moves.”
The possibilities for refinement of the technology are vast, but also complex. Driverless cars need to be programmed to learn from humans’ driving styles – but not to repeat their bad habits.
The road to a driverless world
These challenges – of mimicking human driving, learning to copy individual driving styles and educating drivers to get comfortable with self-driving cars – are the tip of the iceberg.
As we get closer to seeing driverless cars become a reality, there remain many unknowns, and many potential applications and approaches to be explored, from autonomous vehicles’ impact on jobs to shifts in the insurance industry.
And it’s possible that the technology’s biggest impact won’t even be on the roads, but in fields and in factories, with autonomous vehicles turning industrial and agricultural supply chains and production into more efficient, fully-automated systems with minimal need for human intervention.
Only one thing is abundantly clear, says Miller, “As they move closer to reality, autonomous vehicles will not only play an integral role in the urban mobility ecosystem, but they will also support a number of new business models too.”