Volvo Takes Responsibility for Its 2017 Drive Me Pilot Program
Motor City Blogman
Who will be responsible for the first road accident caused by a privately owned autonomous car? Volvo is arguably the leading manufacturer in advancing self-driving technology, and it will take responsibility for any accidents caused by its vehicles while operating in autonomous mode.
A test case could come as early as 2017, when Volvo releases 100 XC90 SUVs that can go into autonomous Drive Me mode on a ring road in Gothenburg, Sweden. Ordinary customers behind the wheel of the XC90 Drive Me models will be allowed to do other things - work, read, make phone calls - without being in control of the SUV. The Drive Me models will be able to change lanes and pass other vehicles autonomously on the ring road.
The 2017 Volvo XC90 Drive Me models will come with improved image tracking and sensors for the conventional model's three radars, front camera, and four surround-view cameras. It also adds lane-tracking, two more cameras in front, including a 360-degree camera, one Lidar and two more radars, including a 360-degree radar.
I recently sat down with Erik Coelingh, Volvo Car Group's senior technical leader for autonomy, in the Research & Development department, to learn details about Drive Me and find out why the Swedish automaker is so confident about its leading edge technology. This interview has been edited for length and clarity.
How did you decide to accept legal responsibility for accidents involving your autonomous cars?
Fairly early, we had come to the conclusion that it's unreasonable to expect that if you have a self-driving car and we tell our customers that you can do something else behind the steering wheel -- and something else goes wrong because of the failure of the car -- that to put the responsibility on the driver would be really strange. The self-driving car is something like riding in an elevator. If the elevator breaks, then you cannot blame the person in the elevator. It's the manufacturer of the elevator who has the responsibility. If independent data show that the cause is the car itself, then it is not unreasonable to think the OEM is responsible.
But you're the first to test in real-world traffic a system that allows the driver to give up control.
There are major societal benefits, at least on paper, with self-driving cars as far as traffic safety, the environment, traffic, efficiency. But there's also the big benefit for people to do other things behind the steering wheel. Just look around you. … It's obvious there's also a need for that. People do not want to sit being stuck in a traffic jam. They want to use their time in a better way.
Your initial test will involve just 50 kilometers (about 31 miles) in Sweden.
Yes, it's the ring road in Gothenburg. It's a motorway-like environment; multiple lanes in the same direction, there's a barrier in between. There are no traffic lights, no intersections. But we think that's attractive, because they're more organized roads … [and] that kind of daily commute is very boring driving. So from a customer's perspective, it makes sense. … We don't claim that we think this is solving the problem. But we will make a huge step forward. There are exceptional situations that the car cannot handle. And in those exceptional situations, the driver takes over immediately.
How would I engage Drive Me mode on one of these ring roads in Gothenburg?
You will see in the navigation system a note that you are approaching a road that is approved for self-driving. Once you approach the road, you will get an offer: 'Do you want to activate the auto-pilot system?' If you want that, you pull two paddles on a steering wheel [on conventional cars, they're used to shift gears in the transmission]. You keep them pressed for a short period of time. … Two buttons because we want to make sure you don't make a mistake. And then, the car will take over. The display will always tell you whether you are in manual mode or autonomous mode. And that difference is very, very important, because you have this issue of mode confusion. It goes back to the driver has to understand the role. … When we approach the point where the driver has to take over, we will have a countdown … then the driver pulls the two paddles again, and the display changes, drastically, again to something normal with speed information. One of the biggest safety risks we see when we do our analysis is that the driver switches off the system by mistake.
What if someone's simply not paying attention, the car counts down to zero, and the ring road autonomy ends?
Then the car will bring itself to a safe stop. And that is a mechanism that we have to implement with very high safety levels. Because a driver will fall asleep, even if we tell him, 'You should not fall asleep.'
Where is the safe stop at the end of this autonomous route?
That varies. Sometimes you need to have access to a safe stop very quickly. Worst case, the car has to stop in its own lane and put on hazard blinkers and put on the parking brake and warn other drivers. That will be extremely rare but better than the alternative.
Do you think there's not much likelihood of anything beyond minor accidents?
Once we release it for ordinary customers to use, we'll have to have high confidence in it. And that confidence is something we're building up right now. We feel that we're on the right track. But you cannot be sure until you have proven with data. We realize that there are things we do not know. This is really research. … It's a big leap. We're getting answers during the project. We're also getting questions.
How will you choose the 100 customers who get XC90s with Drive Me in 2017?
It's going to be leasing only. The technology is not designed to be operational 10 years. When you design a car, it has to have robustness requirements, lifetime requirements, and we cannot fulfill all those lifetime requirements in the first generation. So after a period we will take back the car.
A three-year lease?
We have not decided. I expect it to be one or two years.
And the 100 XC90s can use this ring road in Gothenburg for the entire period.
That's right. The research doesn't end with getting the cars out there. The real interesting period begins with these 100 vehicles. We think the customers will appreciate the self-driving car. But we are not sure. Once we have those 100 vehicles, we can start to do interviews, we can do analysis, and we can have loads of answers to these questions that we have. So then the non-technical research starts. … We do not want to have the enthusiasts who have already mailed us and said, 'I want to be part of the project.' We also want to have people who are maybe a little bit more skeptical about a self-driving car. We want to have people who are experienced drivers, inexperienced drivers, young people, old people, because then we will learn.
Do you see U.S. customers being too different, or could this be your next step after Gothenburg?
We think that this technology has the biggest benefits for customers who live in big cities. Those customers who have to sit a half an hour, hour in the morning, and the evenings in traffic. If we're successful in 2017, the U.S. is one of the first places where we do the next kind of program. But nothing is decided.
Do you have any goals - say 2020 or earlier -- you want something on the road in the U.S.?
If we're successful with the Gothenburg program, in '17 or '18, then we can move really quickly to produce more of the same cars, a little bit more mature, and do the same study somewhere else. And we would like to do it somewhere else, because Gothenburg is a small town.
What about city driving, with pedestrians and bicyclists and so on?
We're not focusing on that so much. Our main goal is not to replace the driver. What we're after is to automate driving where driving is boring. If driving is fun, go ahead, enjoy your car. It's a lot of fun to drive a car. But realize there are certain trips where it's pretty much boring and you'd rather be doing something else. It's the daily commute that's the trip most people would like to get rid of … or maybe it's the long-distance trip that you have to do quite often.
How do you view Google's effort, which has been concentrating more on city driving?
I think the product that they're making is a completely different product than what we are making. The autonomy technology is the same, and all the challenges are the same, but we're making a luxury car where we are giving the driver the choice, to be driven, and what they do is more like a transit system, replacing taxis, doing short trips downtown. Both [forms] can live side by side.
Mercedes-Benz is doing a lot of work. General Motors has plans with Cadillac, also in 2017. Do you think you're in the forefront?
I don't know how one would measure that. We have the benefit of having a strong heritage with safety -- and with active safety. Then I think our strength is, we have good focus. We don't try to do everything at the same time. We don't focus on fancy concept cars, self-driving racing cars, those kinds of things. We try to use the technology and focus on how to bring this to a product.
How much of your career has been devoted to autonomy?
I've been here at Volvo 15 years now. I started with the active safety systems, like city safety. We did automatic braking at the time when automatic braking was questioned. The department has a budget in total of half a million Swedish crowns, which is $70 million U.S., something like that, so it's a big project. But I have to make sure that it is not one-off, 100 cars. I have to make sure it is a technology that fits our technology strategy.
When will Drive Me models become commercially available?
You have to verify this on loads of different roads. I think that will be 2030, my kind-of guess. [The project] will increase in scope so that you can do more locations -- you go to Asia, to European cities -- and then you will learn what you have to do to roll this out in large volumes in more places. And it will take time. Start in Gothenburg in 2017, maybe in 2020 more, by 2030 maybe we will have learned how to roll this out on many roads.
Could you conceive of a world in which self-driving cars become so popular that that's how everybody drives?
Self-driving cars will be really, really safe. The technology will help drivers avoid collisions. In the U.S., 32,000 people are killed [each year in auto accidents]. We have to do something about that. That balance between self-driving and driving yourself, it wouldn't surprise me if that would change over time. But it will never completely disappear. People are still riding horses.