RISE OF THE ROBOTS
MEET AUDI’S HEAD OF AUTONOMY WHO’S OUT TO MAKE DRIVING AN OPT-IN ACTIVITY
Audi’s autonomy guru on why Ingolstadt’s not asleep at the wheel
THE CRYSTAL BALL that provides an insight into the future of autonomous driving is a weird sphere of deeply divisive properties. In fact, it would seem that clarity comes not from the ball itself, but from the eyes of whoever is gazing into it.
Some experts see nothing but a hazy point far into the future where humans are always going to be required behind the wheel. For others, the vision sparkles with complete clarity, where driverless taxis shuttle you seamlessly to your destination, or where you nap in the driver’s seat of your car on the boring, traffic-snarled commute home.
No surprise that Audi’s Head of Advance Development Automated Driving, Miklos Kiss, is very firmly in the latter camp. He’s been a senior figure in this area for nearly a decade, but as he and I settle into an Audi A8 for the three-hour drive from Sydney to Canberra, my first line of questioning gets right back to basics: does the automotive world really need full autonomy? Are customers clamouring for this technology, or is it simply a case of having to pursue it for fear of being left behind?
“No question customers want it,” he says unequivocally. “We see evidence of that each time we have introduced a feature that takes some of the chore away from the driver, like radar cruise or lane-keep assist. Customers keep optioning these features, and our research says they want more. The fact is, a lot of driving – long motorway journeys – is pretty mundane, and customers want to be able to use that time more productively.”
Okay, but how feasible is it really? Plenty of well-placed
“We can’t rely on the driver when the car is at Level 3 autonomy. His brain is off after a couple minutes”
industry senior heads, like Waymo’s John Krafcik, have publicly said that full autonomy may never happen; that it’s far harder to achieve than everyone originally thought.
Kiss smiles, and says, “I love John, and we agree on plenty of things, including the fact that reaching Level 5 is probably science fiction.
“But let’s be clear on what the levels really mean: they refer to the driver’s task, and that’s it. Level 2 is where we are at currently. The move to Level 3 is the largest step: the driver is able to hand over full control to the car, but has to be ready and able to take back control within eight to 10 seconds. Level 4 means that under certain circumstances, the driver isn’t needed at all – for example, as you arrive at a multi-story carpark, the driver can get out and the car goes off by itself and parks. And returns to pick you up. No driver is needed in the car.
“So Level 4 on a freeway, for example, you should be able to nap, or do things completely unrelated to driving. Off the freeway, yes, the driver will be needed in some situations.”
Before we discuss time frames, Kiss explains that in development terms, there are two quite distinct projects in play: bringing Level 3 to market for passenger cars, and the target of Level 4 for driverless commercial vehicles, so-called robo-taxis operating in a geo-fenced urban environment.
“These cars can’t leave the city confines, but they would not require wheel and pedals; I see that as realistic,” he says.
“There are two very different games at play here. The initial part is for the private car, raising automation bit by bit, then one day the robo-taxi thing will step in. We can’t say for sure what day that is, but perhaps within the decade.”
So when will we see Level 3 offered on production cars? “Definitely within five years,” he replies.
Okay, but if the driver still has to retain a level of attentiveness, is it really going to be worth the cost it will add to a premium vehicle? Isn’t it a bit like having a personal assistant in your office, but you have to watch every email that’s typed, and listen to every phone call?
Kiss tries hard not to scoff. “If you experience Level 3 on a long trip, you will never ask this question,” he advises firmly. “Monitoring speed and distance, lane-keeping… the concentration required for this over a long journey is substantial. You remove this, you arrive fresh.”
One of Kiss’s primary areas of expertise is in the way humans interact with machines, and so-called machine learning. He explains that a huge amount of effort is going into making drivers completely comfortable with handing control over to the car, and retaking control when required.
“We do testing with people in the driver’s seat where they don’t know the expert in the passenger seat can take back control of the car. So we know that people trust the technology very quickly; they can relax and rely completely on the vehicle.”
But that raises a converse issue. “So we can’t rely on the driver in Level 3. His brain is off after a couple minutes. We must make sure there is no sudden handover; it has to be a comfortable handover of control when the car requires it. The average time needed for a driver to fully re-engage with his surroundings and the situation is around five to eight seconds. So we are aiming for worst-case handover of control of no less than 10 seconds.”
What about the question of emergency? Who does the car choose to protect in the case of an inevitable accident? “That’s a discussion for the next two hours,” he says with a smile.
“First up, an algorithm will never be able to predict the exact outcome of an accident.” I push further, posing the scenario of a conventional car suddenly coming head-on to a car under Level 3 autonomy, but where swerving would cause its own catastrophic accident. I don’t get a clear answer.
Instead, Kiss shifts to the net safety benefit he sees in the future when autonomy is mainstream. “The machine never gets in a hurry, never gets tired or distracted. Plus the leap in vehicle-to-vehicle technology [where the car ahead can alert others behind of an emergency or unforeseen conditions] will help this a lot. I see a huge net benefit in safety.”
So which car companies would he suggest are currently leading the race to autonomy? Does he have deep knowledge as to how Audi is progressing compared to the other premium German brands, or Ford, GM, or the Japanese? “No question it will be a head-to-head race. Each of the premium brands has certain strengths. Let’s see who will be first,” he says.
So is he under immense pressure to not be among the late arrivals to deliver the technology?
“Previously there was pressure, yes, but over the last year or so the pressure has been reduced. The board realises the full depth of responsibility it is taking with a sign-off of this technology for production.
“From the vehicle development side, we can implement a Level 3 system that can follow a car in front very safely, stay mainly in the slow lane, drive quite conservatively. But the problem we have is from the marketing department. ‘Oh, no-one wants that,’ they say. We may explain that currently, our sensors can’t see a motorcyclist approaching from behind at 300km/h and be able to supply information fast enough for the car to avoid a lane change. Marketing says, ‘But no-one should be travelling that fast!’ Then the legal department steps in; they are crucial to avoiding anything that could go wrong and damage the brand. These are the sort of conundrums we face.”
Ultimately, though, victory in the race to autonomy may turn out to be won not just by the battle for technical superiority, but by navigating a global legal minefield.
“For sure, the legal side presents perhaps the biggest hurdle,” Kiss concedes. “Most important for us is to have a homogenous legal framework; if we have state-by-state or country-specific laws, we’re not going to be able to roll out these systems that will work on roads worldwide. In Europe, for example, we have Germany and Belgium moving quickly, they are the only two countries where Level 3 driving is allowed. The rest are taking more time than expected.”
So a unified global position is required, yes?
“We’re working on that very hard,” he says.
But exactly how that stands to play out, well, not even Kiss’s crystal ball can provide that answer.