Autonomous cars may never go global, says BMW

BMW's global sales and marketing boss Dr Ian Robertson is a visionary Brit who believes the automotive industry faces more change in the next decade than the last 50 years, thanks to the hyper acceleration of advanced connectivity and autonomous driving technologies.

But he also believes autonomous vehicles are decades away, if not longer, from dominating our roads because of legal, ethical and infrastructure issues, and that as long as drivers continue to drive they will never achieve the lofty safety advances predicted by their proponents.

Experts insist self-driving vehicles have the potential to significantly reduce the road toll because 90 per cent of collisions involve human error, but road fatalities won't be reduced by that sort of proportion while non-autonomous vehicles remain on the road.

They also point to the improved productivity of autonomous vehicle occupants, who will be able to work, communicate and even sleep while they commute, and even the death of car insurance because they'll be so safe.

But sceptics of driverless vehicles say autonomous cars won't solve traffic jams, will never eliminate the road toll, raise significant concerns about legal liability and the potential for cyber attacks and won't significantly improve productivity due to car-sickness.

Robertson agrees all of those remain issues with self-driving cars, but that won't stop BMW joining Audi, which will next year release a new A8 limousine that (local laws permitting) will be able to drive itself at speeds of up to 60km/h, in releasing its first Level 3 autonomous vehicle in 2021.

Speaking to motoring.com.au at last week's Geneva motor show, Robertson said it was the most exciting time the global auto industry had ever seen, but there was plenty of hard work ahead.

"It's not a glib statement to say we're going to see more changes in the next five to 10 years than we've seen in the last 50 – in everything [from] materials, layout, the look, to the way in which we sell them and buy them, the way in which they're financed, the amount of time people use them and share them, the technology of autonomous driving," he said.

"And by the way it's not easy, which is why we put this group together of Intel, Mobileye and ourselves with our mapping systems, because every car becomes a super-computer.

"If you went from the dawn of the 3 Series to last year's 7 Series, we've got 100 million lines of code in the 7 Series. If you go to autonomous driving this is a drop in the ocean.

"What you can't have is the back-end being the control room. You need to have a back-end but this world we live in has to have real-time management in the car, because if I want to drive to Frankfurt from Geneva tonight, there's a lot of stuff in the way and therefore the car must be capable of running that system itself."

Robertson said that while current mapping systems, which are dominated by three main companies in Google, Tom Tom and Here (now owned by Daimler, Audi and BMW), are accurate to six metres, "which is fine".

"But if I want to drive autonomously it needs to be 6cm, because there's a lot of things in the way and you'll run over a few things, so you have to have this real-time information that's extremely accurate being processed in the car."

Will they make our roads safer?

Robertson said the importance of autonomous cars in reducing the road toll is not in doubt, but stressed that their full benefits will not occur for decades or more, until people no longer actually drive. He said their implementation will be gradual, in predetermined zones in select regions, due to questions around legal liability and the ethics of artifical intelligence.

"We're going to see cars become a lot safer and there will be less accidents," he said, before cautioning: "Of course there's a huge transition period, because every year there's a 100 million cars sold and there's a billion cars on the road, so there's a long time before they're all technically capable.

"So there's this mis-match for decades.

"I think the other part of this is that the legislation is the unknown part of this. It's immature, people are just beginning to ask the question 'what if?'.

"Does the car have the ability to make the life or death decision? Yes it does. Are we ready for that? No we're not. So it will be limited not by the technology but by society.

"Now if I look at the US, I think 40,000 people died on the road last year. So if you halved it, on the one hand you'd say that's a really good step. On the other, if 20,000 people fell out of the sky in an aeroplane every year no one would be flying.

"So our perception of what this looks like is not easy, so how this develops I don't think will be limited by the technical steps. It will be limited by the other things around it: what we're prepared to accept and how we're prepared to accept it.

"We're really not ready for a machine to make that decision. So along the way there's going to be limits put on it. That's the thing we'll see in the next five to seven years.

"AI [artificial intelligence] is in its early dawn. We all make decisions very rapidly and we live with the consequences. The capability for machines to make those decisions is not the difficult part, but the consequences are.

"You can look at the airline industry. They've had autopilot for a long time, so it works. We could probably fly from Geneva to Sydney right now and fly on autopilot all the way. But there's a crew sitting there all the time because things go wrong and when they do you need to have a person take over and I doubt people would get on the plane if there wasn't a crew in there.

"So you could say that industry has some lessons but I'm not sure that's helpful, because of the circumstances of the vehicle and complexity with which it runs and the potential [for accidents]."

Who will be liable for accidents?

Asked if BMW would take responsibility for collisions caused by its autonomous vehicles, as Volvo and Audi say they will, Robertson was not definitive.

"That's another interesting question, because the built-in algorithms will make the decision from all the information. Who is ultimately responsible is another part of the jigsaw.

"I think the legislators are saying various things, because they don't know the answer either. How do you certify it, how do you regulate it, how do you put the boundaries around it?

"If you look at low-speed restricted areas, that's a lot easier to do than very high-speed unrestricted areas.

"So I think we're going to see part of the mobility world enabling earlier than others, but will it be global? No it won't.

"It's decades. The car park of non-autonomous vehicles is around for decades, if not even longer than decades, so you're going to have this mis-match. But I do think you're going to avoid a lot more accidents, but accidents will still happen."

Will autonomous cars make us sick?

Robertson said another less widely reported problem with autonomous cars is motion sickness when occupants attempt to do things other than drive.

"When we get to the next big steps in autonomous driving, which we believe will be 2021, then you can do more with what the passenger does whilst he's driving," he said.

"We're experimenting a lot with that at the moment and we have a very realistic test cell in Munich where I go regularly to test the experiment.

"[But] With some of it you get nauseous very quickly because as you start to move your seating position to watch something that's fine until you start seeing the road rushing past.

"Like you said to your kid when he said mum I don't feel well, keep your eyes looking forward. So how the interior of the future develops is still early days, but there's now this spring board where you'll start to see it develop."