Self-driving car concepts are a dramatic departure from today’s cars, envisioning living rooms on wheels with rear-facing seats and massive infotainment screens. Even if parts of these futuristic design experiments get lost on the cutting room floor as autonomous driving develops, it’s clear that vehicles will need to undergo a fundamental change that’s much less visible.
As automakers have added new functions to vehicles, the parts responsible for controlling them have multiplied, creating a complex architecture that has become increasingly difficult to upgrade. To facilitate major technology advances such as autonomous driving, manufacturers will have to make a 180-degree turn from industry practices and simplify the number of parts while increasing the power to run software.
The traditional vehicle architecture is highly distributed with fragmented software, it’s not scalable and, finally, it’s hardware-centric,” Glen De Vos, chief technology officer at Aptiv, told reporters late last year. “If you think about our industry, where most of us are hardware engineers, it makes sense. It tends to be inertia. You don’t change, because that’s how you think.”
De Vos pointed to Tesla Inc. as a company that started with a simplified structure from scratch, consolidating most software functions into a centralized computing platform. This architecture allowed the automaker to implement over-the-air updates ahead of the rest of the industry.
This capability became more than just a convenience feature for Tesla customers in September, when the automaker remotely increased the battery range of its vehicles for owners attempting to evacuate during Hurricane Irma.
Tesla had the advantage of building its first vehicles in the era of connected technologies, allowing it to design with those capabilities in mind. For its more established competitors, such architecture will require a much deeper effort to change outlooks and operations.
“For existing vehicles, it will probably take two generations to complete this transition,” said Walter Sullivan, who heads the Silicon Valley lab at Elektrobit, an autonomous driving software supplier. “Whereas these new EVs being designed from scratch, those have the benefit of not having the legacy architecture, so they’ll make the transition faster within a single product generation. They’re probably already designed for this eventual transition.”
As the age of autonomous and connected vehicles dawns, manufacturers have begun to tout the “millions of lines of code” these cars will need to function. However, the current fractured vehicle structure won’t be able to efficiently handle that level of complexity, experts say.
“With each additional function that came into the car, an electronic control unit came in with it,” De Vos said. “Eventually, you’re basically connecting all of these individual controllers. You could imagine the complexity of the wiring harness and power and signal distribution that results from this; it becomes just a huge nightmare.”
To enable autonomous driving, powerful processors from chipmakers such as Nvidia and Intel are making their way into the vehicle, as is an increase in battery power as electrification spreads. These two trends allow automakers to bundle multiple functions on just a few ECUs, rather than having one for each function.
“One of the strongest trends is moving to a central computer,” said Dennis Nobelius, CEO of Zenuity, an automated driving software company formed by Volvo Cars and Autoliv. “Now, in the [Volvo] XC90, you have, like, 110 different ECUs or processing units. That is being moved to one central base, where it could be Nvidia or someone else that is handling the processing capability.”
This consolidation already is happening, according to Patrick Little, senior vice president of automotive at Qualcomm. Most new infotainment systems and digital clusters, including Qualcomm’s, are now powered by only one ECU.
With a centralized control structure, automakers will be able to meet the needs of a technology-laden car much faster, similar to Tesla’s remote battery upgrade during Hurricane Irma.
Manufacturers can use over-the-air updates to upgrade the functions of the entire car or specific parts, only needing to target the primary controller. These updates also can add capabilities or strengthen the security of the car’s software, Little said.
“A lot of times we learn new things about security, or learn new things about functions, after the car is off the manufacturing line,” he said. “The cars have to be updated for that.”
The ability to remotely enable functions also allows manufacturers to deploy varying levels of autonomous driving — once the software is validated and the necessary hardware is in place — without having to build a new vehicle to incorporate each advancement, Nobelius said.
“You can scale up or downwards depending on what segment you’re targeting, or what kind of need you have as a company,” he said.
Sherif Marakby, vice president of autonomous vehicles and electrification at Ford Motor Co., said ECU consolidation is one of the major architecture changes the automaker is making toward its ultimate goal of designing longer-lasting vehicles for use in mobility and delivery services.
The car “is going to go through hundreds of thousands of miles in much less time,” Marakby said. “Mechanically, or architecturally, we need to change it.”
Instituting fundamental architecture changes will take more than just coming up with a new vehicle design. Automakers will have to refocus and reorganize themselves away from the hardware-centric approach they’ve taken for decades.
“We have to let the software be developed in a Silicon Valley way,” Marakby said. “That’s a mindset change.”
He added that Ford has tapped outside talent — partly through its investment in self-driving software company Argo AI and acquisition of Palo Alto, Calif., startup Autonomic Technologies — to aid in this paradigm shift, in addition to relying on veterans within the automaker.
But for an industry that has operated largely in silos, this transition may not be so simple. Aptiv’s De Vos said current organizational structures are based around specific vehicle functions, such as powertrain, transmission or electrification, that are set in their own development cycles. These divisions will blur as software takes precedent over hardware.
“Those departments have a tendency toward self-preservation,” said De Vos, whose company was spun off by Delphi Corp. as an autonomous-driving technology business. “All those groups are actually software groups now. It’s a reprofiling.”
He added that to be successful, automakers will have to invest significantly in software engineers while also reducing the cost of integrating new technologies in the vehicle.
They’ll also have to move much faster to keep up with the encroaching competition.
“That’s a challenge for every single one of them because things are moving more,” said Qualcomm’s Little. “They’re all struggling with how to gear themselves for this rapid innovation dynamic.”