Content Block

Science 17/07/2017

How long before we get true driverless cars?

Image: Daniel Dionne/Flickr

For the past few years it has felt as though we’re on the cusp of a revolution in transport technology. As a technology journalist, it feels as though I receive a press release on a near daily basis from a car company touting the latest and greatest advances in autonomous driving technology. In just the last few months I’ve test driven a Tesla on “autopilot” mode, I’ve ridden in the passenger seat of a fully autonomous Nissan Leaf, and I’ve witnessed an Ocado delivery van drop off some shopping without the need for a humanoid in the driving seat.

Every time it makes me ask the obvious question: How long will it be until this technology is rolled out for real? How long until we no longer need to worry about learning to drive, because the car will drive for us?

Autonomy isn’t simply a binary thing. Since 2014 the industry has standardised a five-tier system to describe the “level” of autonomy in any given vehicle. One being nothing but sensors and warning lights, which still require full manual control, and five being the dream, in which the “driver” can have a nap while their car takes them to their destination.

At the moment, the technology is perhaps somewhere between levels two (“hands off”) and three (“eyes off”) – the latter meaning the driver can safely turn their attention away from the road, but still intervene when necessary.

For example, Tesla’s “autopilot” mode only works properly on motorways, automatically steering the car so that it stays in lane, and it will also change lane by itself – once the human driving it has flipped an indicator to say they wish to change lanes.

While motorways are the fastest roads, they’re actually the easiest to make a computer system drive because movement is so structured. There are strict rules that motorists follow, and there are no pesky cyclists or pedestrians weaving between vehicles. Roads are designed to be simple (no strange junctions – just clearly designed slip-roads), and they are highly maintained, so that the solid white lines on either side that autonomous vehicles use for working out exactly where the road is are less likely to be hidden by trees or debris. Driving autonomously in town, by contrast, is much harder – but it’s the goal of Tesla CEO Elon Musk to have a full “level five” working in the next two years.

This timeframe is similar to the traditional car companies, which are (unsurprisingly) also investing heavily in autonomous R&D. Nissan’s recent test in East London, which I was invited to participate in, was genuinely exciting. As the car completed the test circuit, which covered residential roads, high streets and a dual carriageways, the executive in the driving seat never once felt the need to touch the steering wheel or the pedals – though he did say that human intervention was required on roughly one in three rides.

So could the technology be nearly there? The wider challenge for Nissan, and all non-motorway driving is in terms of the data collected to map out the built environment. Unlike motorways, streets are messy and complicated – so autonomous vehicles need detailed 3D models of where they are driving in order to safely navigate. Though the obvious intention is for all driverless cars to feed real time data back up to the cloud, so that these virtual maps will remain constantly updated, building the initial map is a challenge – which is why during the trial I experienced, the car was very much fixed on one short route. Similar to Musk though, Nissan is aiming at 2020 for it to achieve full autonomy. Google, Toyota and Volvo are also aiming for 2020 – and Ford and BMW are aiming at 2021.

In essence, if we believe these technological projections, level 5 autonomy could really be only a few years away.

But technology isn’t the full story. Inevitably, on the regulatory side governments are likely to lag behind the technology, and the process of making autonomous vehicles road legal will involve a slow transition.

The first cars – or “locomotives” as the Victorians called them – required every vehicle to be preceded by a man carrying a red flag in front, to warn people that a car was coming - and the shift to driverless will have similar rules that in retrospect will appear silly.

“It is inevitable that every car on the road will be autonomous”

For example, the first fully autonomous vehicles will look a lot like traditional cars, and laws will require the driver to be ready to jump in and take control at a moment’s notice. But once autonomy is fully accepted, it will completely overturn all of the existing assumptions about car design. Why should the seats face forward? Why does one occupant need access to a wheel that can manually modify the direction the car is going? And so on. This will appear as silly to future generations as the man with the flag. But for design to evolve, regulations will need to be updated – which is going to be a much slower process.

This said, some governments are proactive in updating legislation. In the most recent Queen’s Speech, the government announced the Automated and Electric Vehicles Bill, which will update the law to match new technologies. But any international standards are going to inevitably take longer to evolve.

“Uber is no-doubt willing along the day that it can put human drivers behind them”

There’s also the question of adoption. We know at some point it is inevitable that every car on the road will be autonomous (perhaps save for some enthusiasts who take their vintage Ford Focus out for a spin on weekends). But we don’t know how long that will take. The case for it taking a long time is simple: We don’t replace cars as often as we do, say, phones – the average age of a car on a British road is 7.7 years, and that has a very long tail. Perhaps there could also be a cultural resistance to autonomy – given how driving is romanticised and wrapped up in (particularly male) identity. So a day when autonomy is the norm could be decades away.

The alternative view – and one I’m sympathetic to – is that adoption of autonomous vehicles could go the other way: So immediately will the benefits and utility of not having to pay attention to the road become apparent, that it will drive a transformation of our roads much faster than we expect. Think about how quickly smartphones reached ubiquity, because the benefits of having the internet in your pocket are obvious.

Once companies discover that staff can be working instead of driving, business fleets will change quickly. Uber is no-doubt willing along the day that it can put human drivers behind them, as their new robot drivers will not require rest or working rights – so they will be able to offer even cheaper rides to customers. Manually driving a car could quickly become as weird as pulling out your Nokia 3310 for a game of Snake is today.

In fact, the first fully autonomous cars are probably already on the road – as the latest Teslas come equipped with all of the sensors required, they just needs a future software update to enable level 5 autonomy.

So is it time to stop worrying about passing your driving test? Will it soon be time to tear up in the pink card in your wallet?

Even if the technology arrives by 2020, as the industry expects – and that isn’t guaranteed – the regulatory regime won’t move as fast. I would be very surprised if we immediately switch to Uber taxis that are completely driverless, because government and public paranoia will insist on having a human in the driving seat ready to step in. Another clear indication that we’ll still be taking driving tests and learning “mirror, signal, manoeuvre” for some years to come is that the DVLA is busy working on making driving licenses digital – something that would be unnecessary if full autonomy really was imminent.

Level five autonomy is inevitable, and it will happen sooner than we think. We just might have to carry a red flag first.


James O’Malley is Interim Editor of Gizmodo UK