Is it End of the Road for Hands-Free Cars?

Just a few years ago, the auto industry promised that autonomous cars were coming to a showroom near you. The problem, it seems, is that the human brain is smarter than anyone thought.

Article by Philip King

The Mercedes F 015 was conceived as an autonomous, interactive mobile living space. Photography courtesy of Mercedes Benz.

Whatever happened to the self-driving car? Go back four or five years and there was a chorus of executives confidently declaring that autonomous vehicles would be ready by now. Instead, the driverless dream has met a similar fate to the flying car: what was an appealing idea has stalled on the home straight, seemingly within sight of the finishing line.

The doubters emerged just before Covid-19. “The challenges are way bigger than anybody thought,” admitted Markus Schäfer, the research chief at Mercedes-Benz, in 2019. The Apple cofounder Steve Wozniak also changed his mind, saying the problems weren’t likely to be solved in his lifetime.

It turns out that even the worst driver reads the road in myriad subtle ways that artificial intelligence struggles to replicate. Overconfidence was a result of early wins. It’s relatively straightforward to program a car to drive itself on a clearly marked multilane highway in good weather, with the proviso the driver always be ready to take the wheel. Negotiating downtown Manhattan at rush hour in a blizzard is several orders of magnitude more difficult.

So for anyone wedded to the original vision of the autonomous vehicle — cars that can drive themselves anywhere at any time, without the need for human intervention — breath-holding is ill-advised. Optimists remain, but they have shifted the goalposts. The aim now is cars that can self-drive in prescribed geographical areas under specific conditions. The industry classifies full autonomy as “level 5” and at this stage most projects are struggling to achieve levels 3 or 4.

For years, the auto giants and tech titans behaved as if full autonomy was within their grasp, selling us a vision of the world in which cars lacked steering wheels and drivers would simply put their feet up. Some thought the underused asset in the driveway would no longer be needed at all, destined to be replaced by ride-sharing robo-taxis. Convenience was just part of the motivation. With self-driving cars, we were told, traffic pollution and congestion would become problems of the past. Most important of all, autonomous driving would radically reduce road accidents, the cause of about 1.35 million deaths globally each year, according to the World Health Organization (recent figures are likely to be closer to one million due to Covid-19). This is because the chief cause of accidents — the “nut behind the wheel” — would be replaced by computers that react more quickly and respond the same way time after time. Computers don’t speed, get distracted by Facebook or turn around to yell at the kids. Many forecasts suggested fatalities would drop by up to 90 per cent.

Over the past decade, autonomous driving has been as central to industry discussions as electric drivelines and connectivity, and just as ubiquitous in design studies. For much of that time, Mercedes- Benz was confident it would be first with the technology and the F 015 Luxury in Motion concept from six years ago embodies that thinking. A long, sleek silhouette houses a luxurious cabin that is accessed by saloon doors. All seats rotate so the four occupants can face one another, and they interact with the vehicle through gestures, eye-tracking and high-resolution touchscreens. Naturally, the vehicle has a zero-emission driveline. And while there is a steering wheel, the marketing blurb promises, “the future of driving means giving the customer the option to not have to drive at all”. It is an electronic tour de force, conceived at a time when the automotive industry could feel Silicon Valley breathing down its neck with rival self-driving projects and car software. This was Mercedes-Benz — the “inventor of the car” — asserting its position in the evolution of motoring. Volvo unveiled its offering, the 360c, in 2018.

Renault’s robo-fantasy, the EZ-Ultimo hotel limousine. Photography courtesy of Renault.

In common with most self-driving designs, the 360c has wheels positioned at its extremities to maximise cabin space. Volvo envisaged four possible interior configurations, including an office, lounge, entertainment space and full-length bed (Volvo says the sleeping cabin “allows you to enjoy premium comfort and peaceful travel through the night and wake up refreshed at your destination”).

Of course, these autonomous vehicles would have to share the road with traditional cars, which presented carmakers with a particular problem: how to interact in traffic without driver-to-driver eye contact. The solution involves sophisticated lighting built into the car’s exterior that signals the vehicle’s intentions to other road users (for example, the F 015 displays “Stop” or “Slow” on its rear panel). For its part, Volvo proposed a move towards a universal communication standard for automated vehicles.

At around the same time, Renault announced the EZ-Ultimo, the last of its trio of robo-vehicles “that explore tomorrow’s shared mobility”. It claims level 4 autonomy, which means the car has complete self-driving capability, as long as there are well-marked roads and highway infrastructure. The capacious cabin has room for three and is trimmed with wood, marble and leather. It is designed to fulfill the niche requirements of a hotel limousine, offering airport pick-ups, premium tours and shopping trips.

Features of these three concepts are echoed in virtually every one of the dozens of autonomous vehicle proposals released over the past few years. Granted, these concepts are at the sexy end of the scale; many are prosaic people movers, with shapes that owe more to the golf buggy than the conventional car. Much of the road-reading sensor technology has been developed by specialist suppliers and can be bought off the shelf, which has encouraged all sorts of startups and fringe players. And so robo-taxi proposals are ten a penny and they all look similar: glass cubes with sliding doors and small wheels at each corner. In May, the Russian lender SberBank revealed Flip, a six- seat “taxi of the future” that is typical of the genre.

The French company Navya has gone beyond the prototype, having sold about 200 self-steering minibuses in Europe, the United States and Japan since 2015. A couple of Navyas ply a route near Perth, though they are conceptually closer to trams than taxis.

The handful of robo-taxi schemes on the planet that actually operate on public roads look nothing like the concepts. In fact, they are standard cars disfigured by pods bristling with cameras, radar and lidar (Light Detection and Ranging, a key technology). There’s been a lot more talk than action, with the world’s first fee-charging, ride-hailing, open-to-the-public service beginning only in October last year. Operated by Waymo, a Google offshoot that has invested more than a decade and a small fortune in the technology, the service comprises a fleet of modified Chrysler Pacifica people movers that drives through the suburbs of Phoenix, Arizona. It is relatively unchallenging territory, but the cars still encounter conditions that leave them flummoxed, such as construction zones — and end up on YouTube.

The Volvo 360c, the cabin of which can be reconfigured as an office, lounge, media- or bedroom. Photography courtesy of Volvo.

Another major player is AutoX. Backed by the multinational Alibaba, AutoX launched a service, also using Chrysler Pacificas, in Shenzhen in January. As for the auto giants, General Motors is the closest to offering driverless ride-hailing, following a thumbs-up from regulators in June. Cruise, a company that is majority-owned by General Motors, will deploy adapted Chevrolet Bolts in San Francisco and has a dedicated robo-taxi, called Origin, due in 2023.

For private buyers, the building blocks of autonomy have been filtering onto car spec sheets for years, starting with anti-lock brakes and electronic stability control, and now extending to an alphabet soup of features that go under the general heading ADAS (Advanced Driver Assistance Systems). To avoid the dorky appendages of robo-taxis, sensors in mainstream cars are concealed by the front grille and are positioned in wing mirrors, bumpers and windscreen pods.These systems deliver great safety benefits, including the ability to warn a tired driver who strays from their lane, apply the brakes if they sense an imminent collision and carry out corrective or even avoidance steering. Many cars now have some self- steering ability that allows the driver to take their hands off the wheel for a short period — typically less than a minute.

But the technology is inching, rather than leaping, forward. For example, the new seventh- generation flagship Mercedes-Benz S-Class sedan, starting at $240,700, has an upgraded Drive Pilot system that offers conditional autonomy when traffic is backed up and in specific conditions on motorways (initially, this technology will only be available in Germany). Outside and in, it looks more like the previous model than the F 015, although the power of the silicon chip is felt to full effect in its comfort and convenience features, which gallop into the future.

Full autonomy is now a distant — and perhaps even unachievable — goal. In the meantime, today’s partially autonomous vehicles throw up challenges of their own. There’s a lot for the driver to learn about how these systems behave in the real world, and even for drivers who are familiar, there can be surprises. As high-profile accidents involving Teslas have shown, misunderstandings or overconfidence can have dire consequences. Partial autonomy requires as much alertness as none at all. So we’re in self-driving limbo and as promises drift into the future, drivers remain chained to the wheel.

A version of this article appears in print in our third edition, Page 60 of T Australia with the headline:
“End of the Road”
Order a copy | Subscribe