Continental has developed what it calls the zone control unit that operates between a high-performance computer and sensors, actuators, and electronic control units. (Image: Continental)

A few days after speaking with Danny Shapiro, Vice President, Automotive, at NVIDIA, a developer of systems-on-a-chip (SoCs) with performances measured in trillions of operations per second (TOPS) and AI capabilities, SAE Media attended the virtual annual briefing on the state of the global auto industry presented by international consultancy AlixPartners.

Mark Wakefield, Global Co-Leader of the firm’s automotive and industrial practices, said that OEMs are facing “technological and competitive forces that threaten to reshape the competitive playing field and the underlying business model that has been in place for decades.” Responding requires more than a pivot, he asserted – companies need to overhaul their core product development and manufacturing philosophies, including adopting a ruthless prioritization in the tradeoffs related to the features that will truly appeal to tomorrow’s customers. “This needs to be done with a newfound speed-to-market and with higher appetite for risk.”

Providing some context, his colleague Stephen Dyer, Co-Leader of the firm’s Greater China business and Head of the firm’s Asia Automotive and Industrials consulting practice, provided examples of how Chinese OEMs are gaining significant traction in their home market against Western brands, such that the Chinese auto companies will outsell the foreign brands in China this year for the first time ever and will likely have 65 percent of their home market by 2030.

Why is that notable? Because China is by far the largest light-vehicle market in the world, with 2022 sales of 23,240,500 light vehicles. And companies including General Motors and Ford, as well as most European marques, have interest in selling into that market.

Looking at the leading new-energy vehicle companies in China, Dyer said their models have an average of 1.3 years in the market – compared with 4.2 years for the top-selling models from non-China brands.

What’s more, Dyer noted, the Chinese brands provide higher ADAS content in their models compared with non-Chinese brands in the same segments (e.g., for vehicles in the $25,000 to $34,500 range, 91 percent of the Chinese vehicles have ADAS compared with 83 percent of the non-domestic vehicles).

NVIDIA’s Drive Orin says the system-on-a-chip has the ability to perform 254 trillion operations per second. Drive AGX Orin, shown here, is slanted toward automated-driving and robotics applications. (Image: NVIDIA)

‘New-Tech’ Focus to Develop Fresh E/E Architectures

Dyer said that the newer Chinese OEMs such Xpeng, BYD, and Zeeker generally are taking an approach to product development that is largely predicated on the extensive use of digital modeling, including prototyping and testing, and are, compared to Western companies, “new-tech driven,” which he contrasts with the traditional approach that he labels “engineering-driven.”

This is not a knock on engineering in any way, but a matter of focus. Dyer said that while the traditional OEMs spend significant time, energy, and resources on attributes like superb ride-and-handling, power and NVH, the Chinese approach is to make those attributes “good enough” and focus more on providing new levels of technology in their vehicles. This is why there is that higher level of ADAS in the Chinese vehicles.

Which comes back to Nvidia’s Shapiro. While talking about companies that are concentrating on making their vehicles increasingly “software-defined” by developing new E/E architectures, he cited Tesla (of course), but then rattled off the names of several Chinese OEMs. Nio, he said, is using four NVIDIA Drive Orin SoCs in a supercomputer it has developed for its vehicles. The Nio Adam computer has the capability of performing over 1,000 TOPS. This unit is in production vehicles (Nio’s ES7, ET7, ET5), not concepts.

ZF has developed cubiX software, which controls functions including the brakes, front and rear axle steering of the Lotus Eletre EV. Traditionally there are separate ECUs for each of those chassis functions but here there is a domain controller used. The software can be updated or upgraded over-the-air. (Image: ZF)

Two of the SoCs in the Nio Adam are used to process the data produced by an array of sensors - on the order of 8 gigabytes every second. The third is a backup. The remaining Orin is for training (remember that NVIDIA provides AI capability). The point: There is a technology transformation going on globally in control architectures. AlixPartners sees it in the market. NVIDIA sees it in current implementations.

And this goes to something that is sometimes overlooked in discussions about changing from existing architectures to ones based on high-performance computers (HPCs): what Shapiro called ‘headroom.’ That is, rather than just having enough capability to perform a particular function, the new architectures have the ability to be, as Shapiro described it, “truly software updatable. Some companies say they are making software updates, but they are just tweaking an algorithm.”

Some companies are performing over-the-air (OTA) updates that modify the infotainment system. Others can change braking capabilities and battery performance. In the not-too-distant future (or in the present, for some of the new vehicles being built in China or in Tesla plants elsewhere) the architectures will result in vehicles being able to do things that they weren’t capable of when they left the factory.

Today and Tomorrow

“Nothing. Absolutely nothing.”

That’s what’s wrong with the current architectures in vehicles today, said Uwe Class, Vice President, Advanced System Development, ZF. But that notwithstanding, they are, he continued, insufficient for meeting evolving requirements. They lack headroom.

“Today’s architectures work,” Class said, “but if you want to answer future requests and if you want to improve integration capabilities, that requires the new architecture.” A new architecture that is based on zone controls and HPCs.

David Muscat, Segment Chief Engineer of architecture and networking business area, Continental North America, explained that in most of today’s vehicles there are roughly 100 electronic control units (ECUs). Each controls two or three features that can be performed in the vehicle. There are modules in the doors (one for each door), tailgate, steering wheel, telematics, and so on. This means that all of these ECUs need power and communications, which results in an array of harnesses and wiring. These not only add mass but potential points of failure. And making functional updates likely requires that three or four ECUs need to be accessed and synchronized.

Nio’s ET5 is one of the brand’s models to gang several Nvidia Drive Orin processors in its Adam controller to perform in excess of 1000 TOPS. Chinese automakers are considered to be at the cutting edge of advanced electrical architectures for passenger vehicles. (Image: Nio)

“The trend today is that the OEMs want to reduce the number of ECUs and, at the same time, to make it less complicated to update features on the fly;’ Muscat said. Accomplishing this requires HPCs that are connected to zone controllers, which are located close to the functions they control (e.g., the doors).

Using a window as an example, Muscat said that with the traditional arrangement, when the button is pushed to have the window go down, the command is read by the door module and it instructs the motor. In the developing architecture, the zone controller obtains the information that the button has been pushed, which it sends to the HPC, which makes a determination of what to do and then sends the command back to the zone controller to activate the motor.

But the thing is – and this gets to the point of the software-defined vehicle – is that the HPC may do something different. It is handling an array of information beyond the status of what’s going on in the door. For example, it is obtaining sensor information. So, it might determine that it is raining and send a message to the driver asking whether the window needs to be opened.

In other words, the advanced system is leveraging some of the vast amounts of information that can be obtained from a vehicle. This explains why there is a need not only for capable processors, but also for high-speed networks like Ethernet and PCI Express.

What Really Changes

The analogy of the vehicle as smartphone is often cited. Someone’s phone on day one has less capability than it has on day 100 because of the downloaded apps and the operating system updates. Things that are, in Shapiro’s term, more than tweaks. But the physical phone is the same on day one as it is on day 100.

General Motors in 2020 announced it developed an all-new electrical architecture called Global B that was claimed to have five times the processing capability of then-contemporary electrical systems. The company did not at the time detail any microprocessor partners. (Image: GM)

Similarly, the components in a vehicle are the same throughout its life: The motor driving the window up and down will always be the same motor.

The way that the software-defined vehicle can be different over time is based on the developers and engineers taking what exists in the vehicle and then adding something to it. To go back to the window example, perhaps the HPC gets connected to the GPS system and can obtain information from weather services so that it “knows” where the vehicle is and can determine whether it is supposed to be raining in that location based on the weather information and so queries the driver who wants the window to go down.

There are apps that can be added to the vehicle for infotainment purposes, but there also are operational changes that can be achieved, such as when EV companies adjust the parameters of the battery and propulsion performance to achieve greater range. But the change to a new architecture, one where there is a separation of the hardware and the software, is not simply one of technology.

Class explained that today a given actuator is developed, built, and tested by a single entity. However, when there is a separation of the hardware (i.e., the actuator) and the software (which controls what the actuator does), then there can be one organization that develops and builds the actuator and another responsible for the software - there needs to be validation that demonstrates that the two operate together as required.

Organizationally, this separation is not the way things generally work today. “The change to more-centralized, zonal control requires structural changes in the way we do development. This is something that is not readily seen from the outside and why an evolutionary, or step-wise approach is taken,” he said.

So, the architectural changes that are being made have as much to do with how engineers work within their organizations and with engineers from other organizations as it does hardware and software development.

Consider the AlixPartners observations about the need for companies to “overhaul their core product development and manufacturing philosophies.” The change in redefining what a vehicle can do when it is running a high-speed processor and is connected to the cloud (suddenly the vehicle is a potential revenue generator for the OEM throughout its life) and the change in consumer expectations (if a smart watch can suggest whether someone has atrial fibrillation, shouldn’t a vehicle?), will require a change in how people in the industry work with one another.

Gary Vasilash is a Detroit-based automotive technical writer and editor. For more information, visit here .