Published on June 23rd, 2020 | by Alex Voigt0
Mercedes-Benz & Nvidia Partner On Autonomous Driving — Numerous Thoughts & Questions
June 23rd, 2020 by Alex Voigt
The CEOs of Mercedes-Benz and Nvidia announced today at a press event that they were launching a partnership for a, quoting Ola Källenius, “groundbreaking, software-defined, high-performance computer architecture for driving assistance and autonomous drive. That’s the next generation that will go into the Mercedes fleet targeting a launch sometime towards the end of 2024.”
In a nutshell, Ola Källenius and Jensen Huang announced that they were together building a central vehicle operating architecture and its operating system to allow over-the-air updates for the often-mentioned S-Class, to enable a subscription-based business model. They explained that it will allow the vehicles to improve over time and that they will build on and offer with that architecture different services, including driving assist features and fully autonomous driving in about 5 years from now.
Everything that Mercedes-Benz and Nvidia presented today could have been words from Elon Musk, CEO of Tesla*, 7 years ago in 2013. That is not bad, just late. If all works out as announced, and that is a big if, Mercedes-Benz would launch first functionality for truck hub-to-hub transportation 10 years after Tesla, in late 2024. Today, in 2020, my Tesla in front of my house already delivers most if not all of what Källenius and Huang described, and although I welcome the partnership and approach, it is not only late — but maybe too late.
What Ola Källenius did not explain at the event today is why BMW, Bosch, and Audi, all companies that previously announced similar partnerships with Mercedes-Benz, decided not to be a part of the announcement today. What was also not explained is how Daimler intends to develop software functionality alone that just a few months ago it tried to build as a partner with VW as well as with BMW. And neither of those two competitors apparently knew about the talks Källenius was having with the other.
Finally, it was also not explained how Mercedes-Benz intends to develop autonomous driving software that needs data to train self-learning algorithms without having any real-life driving data yet or even a fleet of vehicles with a required set of cameras and sensors to collect such data. To be honest, the partnership announced today reveals that Mercedes-Benz is more than 10 years behind, and while their intentions may be good, I believe their timeline is not realistic and is missing critical components to attract the necessary talent to build a working solution — including building, as confirmed by Källenius, missing hardware components like sensors.
Just 4 days ago, Mercedes-Benz and BMW announced that they were stopping their planned cooperation and partnership to develop driving-assist systems up to autonomous driving Level 4. An explanation we hear quite often these days was given by both companies — that cost factors and the economic crisis caused by Covid-19 were the root cause why the cooperation was ending. If high costs are the driving factor, is Mercedes-Benz now better off developing everything alone (or just with Nvidia)?
Costs, but also lack of relevant technology and experienced resources, are without a doubt some of the many issues German automakers have when trying to promise new autonomous technology.
“It takes more and more money, more and more software resources, and more and more hardware know-how,” Waymo CEO John Krafcik has said.
It does not make it more convincing, though, that Audi in 2019 was a part of the partnership discussions too but faded out of it late last year. Audi announced in April 2020 that the new Audi A8 flagship vehicle would come without the promised Level 3 autonomous driving capabilities, despite marketing efforts that had hyped it. Hans-Joachim Rothenpieler, Head of R&D and Audi Board Member, announced that the functionality would not to be available and would be left out because of regulatory issues.
Rothenspieler explained that the lack of uniform international legal regulations, in particular the question of liability in the event of an accident often being unclear, was the problem. That sounds again like something that has nothing to do with a lack of functionality for Level 3 autonomy for the Audi A8. Obviously, other automakers are willing to deal with those regulatory challenges, like Mercedes-Benz. Why can’t Audi? Peter Mertens, former Head of R&D at Audi, said on that topic in an interview in June 2020 that his successors may have promised “a little bit too much.”
The CEO of Audi since April 2020, Markus Duesmann, announced 4 days ago that he is taking over the R&D role, and Rothenspieler announced in return that he will leave Audi after working 34 years for Volkswagen Group. Herbert Diess, Volkswagen Group’s CEO, wrote a friendly goodbye PR letter, but it’s known that he brought Duesmann into VW from BMW and promoted him to the R&D role. Draw your own conclusions about that sequence of events.
The original plan of the 3 German automakers — Audi, BMW, and Mercedes-Benz — was to announce the autonomous driving partnership in September 2019 at the IAA conference. Now, 9 months later, just Mercedes-Benz is left. Also, Källenius mentioned that the Bosch partnership to develop robotaxis is also on hold now.
The objective of the partnership last year was, “to jointly develop the next generation of technology for driver assistance systems and automated driving on motorways as well as automated parking functions (each up to Level 4).” The 2019 plan was to deliver the functionality for customers in 2024. Does that somehow sound similar to you about what has been announced today?
Even more interesting is that the talks that started under former Daimler CEO Dieter Zetsche and former BMW CEO Harald Krüger may have been negatively affected after it leaked out that Daimler was in secret talks with VW about software cooperation too. BMW and VW have a relationship that can be best described with the word hostile.
The irritation from BMW, to use a polite word, was severe. Therefore, it is not a surprise that Mercedes-Benz talking with two competitors about the same cooperation secretly in parallel did not create credibility or trust.
Regardless of how we judge what German automakers have done so far, one part is very clear: without functionality that allows more than just basic driving assist support, consumers have a strong reason to turn away from a German vehicle and choose one that offers a reliable solution, like Tesla is offering today.
Since I’ve used the available Full Self Driving capabilities of my Tesla Model 3 about 98% of the time I’ve been driving on the German Autobahn for a year, including automatic lane changes at a top speed of 150 km/h, I may be an exception but can confirm from experience that it works very well. The Tesla Autopilot system even saved me in many situations in which, if I would have been driving myself, I probably would have reacted too late.
For safety, as well as relaxed driving, I herewith testify that never in my life do I intend to buy a vehicle that does not have similar functionality. In other words, if the German automakers won’t be able to deliver on their promises, then I will rule them out even if they offer a nice electric vehicle.
Since Daimler announced the new S-Class will have Level 3 capabilities and BMW said it will have them in the iNext, there are at least two German automakers with announcements that I am looking forward to. In any case, the self-learning algorithms used in those systems need to learn, and they can’t do that in a virtual or simulated environment, only with vehicles driving on real roads.
Also, the known energy consumption of the best Nvidia automotive chips is very high and this will have a negative impact on the vehicle range. To achieve the required critical balance between hardware and software mandates tailor-made solutions that are hard to achieve with a supplier like Nvidia that develops systems and chips for many automotive companies with different hardware, not tailor-made for Mercedes-Benz.
Nvidia may have the best available chips for processing and computing data but this is a part that needs to play perfectly well with other parts in a complex system to achieve the goal defined by many requirements. In the gaming industry that Nvidia is coming from, the energy usage never played an important role, but it does with electric vehicles. If Mercedes-Benz intends to develop autonomous driving functionality for fossil fuel vehicles, though, like the new S-Class, it makes another strategic mistake.
Vertical integration could change that supplier-related issue, but the German automakers didn’t even try up till now to become more vertically integrated and have lost critical time in their continuous ongoing negotiations in recent years that led nowhere.
Having driving data collected is critical as well, and although some test vehicles are seen from time to time in Germany, that does not scale up to the gradual approach from Tesla, which releases more functionality over time and step by step, like traffic light recognition and then going through a green light without confirmation. To have hundreds of thousands of vehicles driving and collecting data in all of the diverse situations of the world is another must-have component a Mercedes-Benz will need to build for many years, but as Källenius said today, they have not even decided on sensors yet. Without that decision, it is hard to imagine how they could line up a timeline to be feature complete by the end of 2024.
For me, as a German, what matters most is that I experienced Tesla Autopilot in the last 12 months and found it to be safer compared to me driving on my own, and I have done that for 35 years and never had an accident or incident — people call me a very safe driver.
In that respect, and it’s my subjective point of view but supported by hard facts and data that Tesla does release, Autopilot is so many years ahead that it may have achieved escape velocity already.
With regards to full autonomy, Källenius announced that truck hub-to-hub transport is their main use case they want to start with. One reason for that could be the latest announcement from Tesla that it is starting production of its Semi truck soon, which with its self-driving technology and lower TCO (total cost of ownership) per km/mile, may cut deep into the order book and demand for Mercedes-Benz commercial vehicles. Robotaxis in an urban environment he sees coming at a later point in time, which means a passenger car from Mercedes-Benz or Daimler with autonomous driving technology won’t be seen before 2024.
What I heard today is another one of those announcements that German automakers make to keep word of mouth going, with many promises to develop solutions that often never see daylight.
I hope to be wrong, but what has been presented to me today did not fit together.
*Disclaimer: I am a Tesla Shareholder
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.