Virtually Amazing

Automotive software is becoming more complex as features and customer offerings evolve. The automotive industry must continue to find ways to not only meet those demands in terms of quality and feature development, but to do it with higher levels of innovation at the blazing speeds needed to stay market competitive.

Major automotive manufacturers are putting the pedal to the metal on the powers of virtualization and simulation to improve and expedite their software development capabilities. As Cedric Armand, Ford Motor Company’s Director of Virtualization, told us, “To deliver high-quality software at the pace our customers expect, we must embrace every tool that accelerates our development cycle. By centralizing and industrializing our virtualization and simulation practices at scale, we aren’t just moving faster-we’re building a more reliable, software-driven future.”

Ford’s Verification and Validation organization is a cornerstone of the Product Digital Integrity program. This initiative focuses on defect prevention, process compliance, data-driven quality, and customer obsession, ensuring rigorous digital assurance and warranty performance for Ford software across all embedded domains.

Armand is overseeing Ford’s strategy to develop and validate software using standardized, industrialized, and large-scale virtual processes. Half of his team develops virtual electronic control units (ECU) and virtual test environments. Virtual ECUs are virtual versions of the actual ECUs that run the operating software for all the systems in a vehicle. Virtual environments are models and simulation of said physical vehicle systems.

The other half of Armand’s team is developing a cloud-based virtual platform that will enable the company’s internal software developers and internal software testers, as well as external tier-one partners to leverage virtual ECUs, test environments, toolchains, and automated workflows through a self-service portal. The cloud-based platform will provide easily reproducible virtual test environments accessible to engineers from anywhere in the world, at any time.

Collaboration becomes less complicated. “(It) makes it very easy without worrying about if we are working apples to apples or is my environment matching your environment? Are we getting similar results?” he explained. It also zaps the limitations; testing can scale up leveraging cloud provider compute servers.

“If I have one physical hill bench to test on, I can run one test on that, and when it’s done, I can run the next one. But if I’m running everything on a cloud, I can spin up as many parallel instances. Of course, you get charged for it, but the cost of it is way more advantageous than if you were to do the same amount of testing using physical resources.”

Several key partnerships support these efforts, including an extensive partnership with Google Cloud Platformto provide cloud and server-based needs; Armand plans to leverage GCP to run large workloads and simulations.

Synopsys is a market leader in Electronic Design Automation solutions. The company is widely used for creating semiconductors, automotive electronics, and high-performance computing devices, providing both software tools and specialized silicon IP. “They have taken that expertise and are leveraging it to create high fidelity, virtual versions of ECUs, and are our primary partners in developing Level 4 virtual ECUs,” he reported. Level 4 vECUs execute unmodified, target-compiled production binary code. They enable “shift-left” development, allowing developers to test exact production software, including OS and drivers, on a PC-based virtual environment long before physical hardware is available.

dSPACE is a partner as a global technology leader for virtual simulations. “dSPACE’s expertise in software-in-the-loop (SIL) testing is helping enable virtual test environment creation and deployment. dSPACE’s toolchain gives us the foundation to co-simulate multiple virtual ECUs and models in a single virtual test environment.”

Other collaborators include, KPIT, a leading independent software development and integration partner. KPIT, has over a decade of experience in large-scale virtualization. “I have worked with them on previous virtualization initiatives. They have proven accelerators, understand the scale and journey we are embarking on, and match my excitement and energy for leading virtualization solutions.”

Speedrunning Invention

When it comes to managing the ever-growing software and system complexities that exist in automobiles, virtualization is a true game-changer. As these convolutions accelerate, so do the challenges of managing them. “It’s driven by the intricate interplay between software and physical components,” Armand pointed out. “With that, there’s going to be a large number of permutations of how these interactions can actually play out once you have vehicles in customers’ hands. And it becomes very difficult to test and probe the potential space of these interactions and scenarios and to do it in a repeatable fashion. But with virtualization, it’s much easier to do that.”
It may be surprising to learn that virtualization is more repeatable than physical hardware. With physical hardware, there are issues that cause loop-to-loop variation, such as thermal drift, component resistance and capacitance tolerances, parasitic losses, localized voltage drops, and physical wear. “In a virtual version, it’s just software running software, and that runs in a consistently repeatable fashion,” he noted. “So, if you do find a problem, it’s easy to reproduce it.”

“But the other thing is that now that you have a vast design space, you can use the extent of server power to run millions and millions of simulations, and get to corner cases (rare, complex scenarios that occur when multiple independent variables or conditions are simultaneously at their extreme limits) and situations that might be more difficult to actually execute on hardware, or might even have the risk of damaging the hardware. But you don’t have to be concerned about that with a virtual environment, and you can get to all those unmanageable design spaces in a structured way.”

For most major manufacturers, developing a new vehicle from the ground up takes roughly four years. As it relates to developing vehicle software, typically, the first two years are spent developing prototype hardware. The next two years are focused on developing the software on said prototype hardware.

“Now, since you have virtual versions of the hardware, you can actually start developing your software in parallel to the hardware development timeline,” Armand said. “Theoretically, the idea is to shrink your development cycle from the four years down to two years, and everything can meet at the same time.”

Virtualization is one of automotive technology’s most impressive force multipliers – and we can see the day that it transforms other industries as well.