Elektroniknet Logo

Autonomous driving

»We are on the verge of a breakthrough«

Fortsetzung des Artikels von Teil 1

Computing power is a problem

But that would mean that no real progress can be expected in the next ten years either, because if autonomous driving were based on sheer computing power, it would be tight...

That’s true to a certain extent. That’s why I believe that other approaches will lead to the goal. You have to select exactly those use cases that make economic sense, i.e., that will also be paid for. For example, an autopilot for the motorway journey from Munich to Nuremberg, and this use case must be made safe. That is relatively simple. If the goal is to have a vehicle drive autonomously in city traffic in unknown cities, then that is much more complicated. If the application or use case is clear, then optimized solutions can be developed precisely for it - and even then we are talking about Level 4.

A generalized variant that works everywhere is much more difficult and will take at least 20 years.

Would training a car for Chinese highways be easy?

Yes, this is especially true of any highway - Chinese, German, Canadian, etc. - because the entire highway network has been designed to be very safe. If you disregard speed in this case, the highway application is the easiest. Because theoretically you can drive hundreds of kilometers without getting into a difficult situation. Highways are by definition-controlled environments with only on and off ramps, so complicated scenarios are much more limited – no sidewalks, no pedestrians, much less things to consider. Systems like autopilot work pretty well here. Traffic in a city (urban center) seems to be the real sticking point for these systems. The problem is that this is exactly where most accidents happen.

I think a lot of people would be interested in a vehicle that is able to drive autonomously on the highway, but these cars don’t exist...

I also think that people would be willing to pay a reasonable amount to not have to drive themselves on the highway. And even if corresponding vehicles are not yet available for purchase, I would say: we are already very close to this goal.

It took a long time...

Yes, because everyone is trying to solve this extremely complex problem alone: tens of companies want to achieve the same thing instead of working together and using standardized training data, for example, but that would be much easier. A look at history shows: mankind has always solved the biggest problems together; in the Apollo moon program, the biggest companies worked together to achieve this goal. With automated driving, everyone does their own thing, that doesn’t work, the problem is far too complex for that.

But the automotive industry is certainly one of the industries that puts competition at the top...

True, but the automotive industry in particular shows: where standards have been created, success can be achieved. Today, over 90 percent of vehicles are equipped with injection systems from Bosch. Why shouldn’t that also work with regard to autonomous driving?

No question, but that sounds more like long-term success, because at the moment ADAS applications in particular are something with which OEMs differentiate themselves from the competition...

I think there are definitely signs that a rethink is taking place here. One example is Volkswagen, which invested billions in the Ford subsidiary Argo AI. I can’t think of any bigger competitors than Ford and Volkswagen. When these companies work together, it shows that it is feasible. On top of that, so far there is no differentiation at all in autonomous driving because autonomous driving doesn’t work. They all deliver the same non-functionality.

BMW wants to hire tens of thousands of software engineers, and everyone wants that....

Yes, because every OEM believes that software is the key to success in the future, engine development certainly isn’t anymore...  Yes, it’s true, the automotive industry is facing enormous challenges, the OEMs have to make sure they can differentiate themselves, but I don’t think they will succeed in automated driving by going it alone.

When you look at the different OEMs, have some understood where differentiation is possible?

I would say yes. GM, for example, wants to introduce its Super Cruise System in 20 models, a clear sign that GM is convinced that the approach works. That’s a good first step, even if the functionality is still limited. But GM is offering the functionality in mid-size vehicles, and that’s a strong commitment. Honda has also started to deliver Level 3 vehicles in Japan, which I also see as a strong commitment.   From my point of view, it is particularly remarkable that companies like GM or Ford, which are typically seen as more conservative, are making very good progress in terms of autonomous driving. Typically, you would expect newcomers like Tesla, for example, to be successful on this topic, but you would expect companies like GM to be far behind the innovation drivers. But that’s not true, some of the most established companies seem to be leading the way here. For me, this is a good sign, because it shows that even traditional OEMs can definitely succeed with new approaches.

It used to be said that with Level 3 - or whatever an automated driving in certain environments is called - LiDAR will become a must, can you confirm that?

I am convinced that Level 3 will come. And I am absolutely convinced that LiDAR will facilitate Level 3 or automated driving in certain environments. Absolutely necessary? I wouldn’t say that, but LiDAR clearly helps to implement the functionality really safely.

LeddarTech is already delivering its systems to the market, are there any other applications besides robotaxis where LeddarTech’s LiDAR technology is already on the road?

LeddarTech’s LiDAR technology is not currently on the road, but there are ongoing programs with OEMs for production vehicles. We started with OEMs that are active in the premium segment. It was no different with radar, but we hope that LiDAR won’t take 30 years to establish itself in the mass market.

What do you think is necessary to advance the idea of automated driving?

Commercialized and reliable solutions on the road. They are necessary to create trust in the technology, tens of thousands of Tesla drivers are not enough for that. Drivers need to get used to the idea of the vehicle driving a part of the journey autonomously. So far, the experience is that the systems don’t work.  When the systems work, attitudes will change. When airbags came along, there was also a lot of press saying that airbags kill people rather than protect them. In the end, the technology proved that it makes sense, and it makes sense for the driver. Today, no one questions the sense of an airbag. And that must also happen with autonomous driving.

You are convinced that only a small step is still necessary for autonomous driving, but in the meantime the skepticism towards the technology is anything but low, not least because of the negative reports in the press...

True, but if you look at how many accidents are caused because the driver makes a mistake, all the accidents caused by ADAS systems are almost negligible. What I’m saying is that automated driving increases road safety, the systems just have to work reliably.  There are now commercial solutions on the road that show what the approaches bring. And that creates a decisive momentum. Today, people tend to be critical when a vehicle drives autonomously; negative reports have so far only confirmed that such approaches do not work. But now everyone can try it out for themselves and find out whether it is true or not. And when it is clear that the cars are reliable and safe, then the market will pick up, just like with airbags.

  1. »We are on the verge of a breakthrough«
  2. Computing power is a problem

Verwandte Artikel

LeddarTech Inc.