I attended the Automobil Elektronik Kongress at Ludwigsburg outside Stuttgart. it was the 23rd year that it has run, and the third that I have attended. The atmosphere has been very different at all three. I summarized the changes affecting the automobile industry in a piece that I wrote a couple of years ago but is still valid, Triple Witching Hour for Automotive, the three big changes being the end of three things:

• Cars being powered by internal combustion...the move to electric traction
• People driving their own cars...the move to ADAS and eventually autonomous driving
• People owning their own car...the move to Uber and other forms of partial ownership

In 2017, the theme was that software was disrupting the automotive industry and, as a result, the industry was going to have to change its attitude and its structure. For example, see my post from back then Automotive Software Development Used to End with SOP and Bosch: Future of Connected Mobility. Software development is very different from hardware development, whether a chip or a car. So the theme, in one sentence, was that the automobile industry needs to get its act together, and cannot go on with business as usual. 

Last year, 2018, the theme was that the industry needed to restructure. The OEM, tier-1, tier-2, supplier layers would no longer suffice. In particular, the OEMs (car companies) needed to take control of their destiny and do some of their own software and system design. (See my post Automobil Elektronik Kongress 2018.) There was also a focus on China, since it had become the largest market for cars. One presentation was by VW's China manager which I covered in Trends, Technologies, and Regulations in China's Auto Market.

This year, the theme was getting practical. The investment required for autonomous driving is huge, but when there might be a payback is uncertain. In 2017, I would say everyone was stunned at how fast autonomous driving technology was advancing, and the industry was caught by surprise. By this year, it is clear that it is not coming that fast after all, but that changes what needs to be introduced when.

There are two high-level challenges about autonomous driving:

1. Technological challenge: With an infinite budget, can we build cars that can drive themselves safely?
2. Economic challenge: What technology can we ship when, at a price point that the customer will pay?

ZF Friedrichshafen

The first presentation on the first day was by Wolf-Henning Scheider of ZF Friedrichshafen on Realistic Ways to Automated Driving. In fact, the title of his presentation could have been the title of much of the conference. Some of what he presented was specific to the ZF product line, but much of it was higher level, about the factors driving the industry.

He divides the market into strategic segments with different characteristics:

• Passenger cars (traditional car companies)
• People and cargo movers (purpose-built vehicles)
• Ride-hailing (transport as a service, TaaS, providers)
• Commercial vehicles (commercial vehicle OEMs and fleet owners)
• Industry automation (automation as a cost reduction factor such as robotic forklift trucks)

Looking at these markets, he said that the challenge is:

We have to invest millions of capital expenditures, and we don’t know when we’ll see the return. But we know if we don’t invest now we’ll miss the train.

That sums up the challenge. A company like ZF has to go full-speed-ahead on investing in technology since everyone knows that driving will get increasingly automated. But in the meantime, they have to make money. The trends are speeding up too, with big changes in:

• Mobility behavior
• Diffusion of advanced technologies
• New competition, new cooperation
• Shifting markets and revenue pools
• Digitalization in products and processes

Wolf-Henning took a practical position on automated driving:

For the passenger car, we don't believe level 4 or 5 in the next few years. it will need good L2+ availability. Why is L2+ so important? It's affordable for private citizens, and we generate trust. Many consumers do not trust fully automated driving since they are not familiar with it. Plus, a L5 system is 100 times more costly than a L2 system today.

As an example, ZF has two solutions for passenger cars:

• Euro NCAP based on Mobileye, a scalable moderate cost solution for L2+
• ZF CoPilot based on NVIDIA, a more expensive solution with more capability

The problem is that the costs go up exponentially for higher levels of automation. He believes that fully automated systems will first be introduced in commercial vehicles.

Another big issue is that there are missing legal frameworks for higher automated driving levels on public roads. He believes that legislators "want us to prove it is safe" and these L2+ systems can be used to gather experience on the road and so prove the systems are safe. The new path is operation in highly restricted areas without the full complexity of a city, such as harbor areas, airports, logistics hubs, all the way up to designated areas like the self-driving shuttles on Singapore's Sentosa Island.

Next, Wolf-Henning took a look at the economics of autonomous vehicles in ride-sharing services such as Uber and Lyft. He sees the price per mile of $1.97 but cost per mile of $2.36, for a loss of 39¢. He breaks the cost down as $1.61 for the driver, 20¢ for customer acquisition, and 17¢ for OPEX. But if you remove the driver because the vehicle is autonomous you get to profitability, with an additional 23¢ per mile for the technology leaving a profit of 6¢ per mile. So autonomous robotaxis have a cost saving of 71% versus non-autonomous service.

One big debate in the autonomous world is whether lidar is required. He said that radar is getting high-precision results and are taking over functions that were inconceivable. Lidar is higher price but radar is now getting equivalent results.

Here is the range of ZF automation systems. I put this as an example of the range required to be both delivering entry-level automation at a low price, and fully autonomous systems for mobility-as-as-service applications. At the low end, it is 1 tera-operations-per-second (TOPS) with Mobileye, up to 600 TOPS with NVIDIA and Xilinx, which is about the power of 120 smartphones, enough processing capability for applications like the self-driving shuttles in Singapore.

In the Q&A, Wolf-Henning was asked, "You took over a mechanical company and are now moving it into the digital era. How do you make sure that workforce supports this?" His reply was interesting, about how ZF are managing the transition from both a business, expertise, and motivational point of view:

They read the newspapers and they know you cannot go on kicking the can down the road. The willingness to go for this transformation can be seen anywhere. We have developed a strategy that cushions the full effect through plug-in hybrids. That gives us some breathing space for the next 10-15 years so that the transformation will be stretched out over time a bit. It also gives us the earning base we need to support the development. Everyone is excited since everyone knows this is the future. Our transmission people, I’m telling them we’ll need them for many years to come since they are our earnings base.

The other question he was asked was that there was lots of hardware on his slides but not so much software. "Can I just purchase the software?"

He admitted that he didn't talk too much about software since they were a late starter.

We do have advanced s/w units we are building up. We just established an AI center. The vehicle in the initial video is NVIDIA drive but the software for restricted area trials like Sentosa we develop on our own. But there are possibilities for alliances.

As I said at the start, this was the first presentation. Other presentations hit on similar themes. In fact, that's something I've noticed at Ludwigsburg. The industry is at a certain stage, worrying about certain issues, and several presentations are very similar except for the company names.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.