Pierre Gompertz from Magna explains in an interview why ADAS industry standards are still lacking today and what he recommends to students.
ATZ _ Springer Professional: Pierre Gompertz, the systems for Automated Driving (AD) are lagging functionality and in terms of timing. Did this not raise expectations and create a too high level of expectation among end customers, which is now falling on the toes of the OEMs?
Autonomous driving is a complicated technology, and it might have been misunderstood as well. As an automobile is only a technology like any other, it will follow the same innovation cycle as any other technology: the Gartner-Hype cycle. What is the Gartner-Hype cycle? It consists of several steps like the technological trigger with early adopters, the peak of inflated expectations, the valley of despair, the slope of enlightening and the mass production adoption. Currently, we are in the valley of despair building into the slope of enlightening which means that people were disappointed by the capability of the system and OEMs are now responding through a new wave of technology which will be accessible to all.
Can you give us an example, please?
On top of now following this cycle, the automotive industry is currently in the spotlight and all eyes are on it with mobility being on of the key pillars of our society. It shines lighter on the ups and downs of the field. Look at the history of the Apple iPad for example. No light was shining on this product, so only few are aware that it did not succeed originally in the 1990s and regain a full breath after 2010. To come back to the Gartner-Hype diagram: It explains exactly what automated driving is going through: no party is at fault, moreover it is about a technology following its course and finding its way to mass adoption.
What new milestones and annual figures can you give us today for 2027 and 2032 – now with a greater amount of accumulated experience? How are the markets developing – also with regard to the differences in the individual regions/countries of the world?
Even with the amount of knowledge that we gain today compared to ten years ago, the span from 2022 to 2030 will still be an era dominated by SAE level 2/2+ with level 3 building more and more momentum toward the end of the decade. As mentioned before with the Gartner-Hype curve, technology is existing, but we now need to build confidence from the market. This confidence and democratization of level-3 systems will come from the ease of use for the consumer through more mature and scalable vehicle architectures. When I say maturity, I mean that we manage and know well enough the technology to take the time to make it easily understandable by any user. As we can see with key mobility player, traditional as well as new companies, the human-machine interface, the HMI, is one of the answers to put automated driving at the reach of all. To answer to the question: I see this decade as being the years of maturity and ramp up of the technology for automated driving systems across the globe.
What does SAE level 4 look like? What is happening for commercial vehicles?
In terms of level 4, the USA will lead the pack and pave the way for the rest of the world. The flexibility of certain states there and the progress made by individual pioneers is unique across the world and is pushing the level 4 one step closer to reality. With Cruise becoming more aggressive on their timeline, the deployment of level 4 commercial vehicles is becoming imminent. A level-4 model matches definitely well with the commercial industry through trucks and delivery bots like Cartken. This concept is not limited to taxis anymore but on how to automatize daily repetitive tasks. With level 4 becoming a commodity in the commercial field, it should come around 2030 at the personal vehicle level.
Some say level 5 will never come because it is too complex. What do you think the opportunities are?
When we see robotaxis driving around certain city without driver, a common misconception is to associate this driverless capability with the level 5. If I want to use a robotaxi outside of its zone, the vehicle will not accept the request because the conditions to drive are not met, the vehicle will not take passengers. In my opinion, level 5 might never happen or at least not within the next ten years. In addition, all the talks about L3/L4/L5 take away from the main purpose of any AD system. What is its real purpose today? Being able to automate specific repetitive types of drives in order to make them safer. This is the real original goal, and it should continue to be the goal ahead of anything else.
Why are ADAS/AD industry standards still missing? Even for the small lidar sensor, there are no common rules on how to write the specification and how to test uniformly. What do VDA, DIN, SAE and UNECE have to do for this?
Issuing industry standards is complicated when the main actors do not agree on what should be the content. Tremendous progress was made by multiple organizations in charge of standards, but goals are still unclear from the perspective of the automotive market. The highway assistant can serve as an example. What are the minimum requirements at a speed of 130 km/h? Some experts call for detecting a pallet on the road ahead at a distance of 180 m, others want it for 250 m on passenger vehicle, some wish extremely specific use cases like a scenario realizing that a tire flies from opposite traffic onto your vehicle. What are the minimum requirements? What are the right requirements? Every OEMs and tier-1 supplier have a slightly different idea of how a highway assistant should function. Most of the standards are based on feedback from these same OEMs and suppliers. For lidar systems, for example, there is no agreed performance matrix to reach. Most of the OEMs and tier-1 suppliers have a different understanding of the performance requirements which leads to multiple lidar offerings all interesting for someone on the market. The start of production in 2026/2027 will be key in defining the next architecture generation. All these new approaches will help converging to a more robust agreement among automotive key players.
Central computer architectures versus zone computers: What are the advantages of both solutions for SAE level 3? And for level 5?
As cars evolve, they are becoming more and more tech oriented. We have hundreds of chips across a single vehicle today. ADAS industry is one of the hungriest users of these chips as well as relying on the most powerful ones. The combination of multiple ADAS systems for adaptive cruise control, surround view, and driver monitoring via radars and cameras into one more powerful central computer is already a reality. One next step is the centralization of other systems such as infotainment, interior control, or lighting. The main advantage is first cost, for sure. However, beyond cost, now we can leverage multiple systems to enhance performances: Interior monitoring cameras can be used for a conference call through video, to take selfie, or to do karaoke. The possibilities are endless.
Hardware was important then, and computing power is available at will – software is now the new gold in the automotive business, they say. What steps need to be taken to achieve a 100 % software-based vehicle?
We already entered the era of the 100-% software-based vehicle. Five years ago, drivers were talking about which hardware was on their passenger car: Look, I have a camera, a radar, my surround view system. Today, no driver is really interested in this sensor technology on the vehicle. He/she prefers to buy a function, a service that has value for him/her with no attention for hardware on the vehicle.
Volkswagen and Cariad are still looking for hundreds of software developers. What do you recommend to a fourth-semester computer science student today?
It depends on where students are living. In the USA, a bachelor’s degree is more than enough to start a career. In Europe, additional master would definitely help tremendously.
Pierre Gompertz, thank you for the interesting interview.
You can read more of the interview in ATZworldwide 1-2023, which will be published on December 30, 2022.