There’s been an awful lot of noise made in recent years about the concept and development of the driverless car, and it’s not hard to see why.
It’s always seemed like a notion so quintessentially futuristic that, despite regular press reports on the gradual progress of eye-wateringly expensive initiatives backed by the likes of Google and Tesla, the very mention of fully autonomous automobiles still sounds like classic sci-fi fodder.
Along with Google, Tesla and Uber (the corporate triumvirate behind arguably the three most high-profile test programs launched to date), many of the bigger motoring companies have of course been monitoring developments closely.
Several of these – most notably BWM, Volvo, Nissan and Toyota – have already made significant direct contributions themselves, predominantly working with electronics giant Bosch to develop ever-more-responsive surround sensors for ongoing road tests.
Back in 2015, Bosch claimed at the Automotive News World Congress in Detroit they were effectively entering the early stages of a ten-year plan: by 2025, said (now former) board member Wolf-Henning Scheider, we’d be seeing the first cars hitting the road that were fully able to transport passengers door to door without any driver intervention at all en-route.
As anyone who’s been keeping tabs on subsequent progress will know, there have already been a fair few hitches along the way. Uber’s 2016 pilot program roll-out, for example, exposed a series a fairly major flaws in the nascent technology – not least the fact that, for all their impressive efficacy in smooth lane-switching, accurate pathfinding and automatic braking, the test cars showed rather less artificial intelligence when required to react appropriately to other human road users.
For the foreseeable future that’s arguably going to be the one problem still causing the biggest headache for the developers of this indisputably monumental technology. Theoretically speaking, our current best efforts can only really function properly when every car is driverless. This gives rise to a relatively seldom-discussed elephant in the room – namely all this heavy investment is banking on the probability that every vehicle on the road will eventually be robotic.
There’s another seldom-discussed area in the driverless car debate. Many people want a human driving the car they’re in. Cheif among this group are corporate and executive clients of UK chauffeur servies. Many people take great comfort in having a reliable, highly-trained driver meet them at a point such as an airport, and then drive them, in comfort and class, to their destination, such as their hotel. It seems that their will always be a demand for corporate chauffeured services such as Cars Exec, which works with clients throughout the United Kingdom.
Clearly that isn’t going to happen any time especially soon, and so logically, in order to get things moving forward, someone will have to be the first to develop an automated car that can integrate seamlessly with more organic driving behaviours.
In many ways there’s a bit of a catch-22 situation emerging which has the potential to hold things up enormously. Moreover, the whole ‘human error’ issue is also slowing progress dramatically for another major reason – and this time, it’s inside the car. Early in the prototype development stages of the industry, it was always assumed the future of autonomous vehicles would see integrated advancements rolled out gradually, cutting back on required driver input by increments until eventually we were all travelling as entirely hands-off passengers.
As it turns out that’s not really going to be much of an option in the long run. A partially autonomous hybrid model would require humans to function as the AI’s emergency backup, but since Google’s hybrid trials in 2012, tests have increasingly confirmed that we’re pretty terrible at performing that role.
We’re too easily distracted when idle, and too slow to react when suddenly called upon to take the wheel. Put simply, most companies have now accepted that automation effectively needs to be an all-or-nothing sort of deal.
The key underlying trade-off here is that of safety versus cost. Fairly successful trials have taken place for what is defined by SAE International as ‘level 2 autonomy’, which denotes cars that are able to accelerate, hold a lane position and decelerate automatically with the ebb and flow of traffic. However, the reaching level 3 autonomy – defined as the car having basic decision-making abilities, such that the driver can safely be ‘eyes-off’ so long as they’re still available to intervene at reasonably short notice – is a vast step up in both technological demand and associated expense.
And yet, while manufacturers have claimed level 2 autonomy could reduce fatal accidents by up to 80 per cent by drastically cutting down on human error (which, somewhat controversially, was found to be responsible for the first fatal crash of a partially automated Tesla Model S in 2016), level 3 autonomy doesn’t deliver an especially significant additional boost in this regard.
Ultimately, there is still progress being made, and that progress is steady if somewhat frustrating for the current cutting-edge manufacturers and program leaders. Companies like Uber will of course keep pushing forward as hard and as fast as they possibly can, because – and again, it comes down to cost – they stand to increase profits massively on the day they no longer need an HR department.
However, quite apart from the staggering scientific challenges of actually making this one-time sci-fi fantast into reality, there are still some major conceptual problems to be addressed that currently pose something of a paradox right at the very heart of the entire driverless car concept.
Quite when anybody will come up with a universally appropriate solution to those is even harder to predict than the next big technological advancement.
Until then, it remains very much a case of ‘watch this space’… just don’t necessarily start holding your breath quite yet.