Iveco Group Accelerates Electric Vehicle Software Development with Virtual Vehicle Simulation
Fully Virtual Vehicles Enable Faster, More Scalable EV Software Testing
In the automotive industry, software development has traditionally moved slowly. Before a line of code can be validated, engineers need hardware, testbenches, and physical vehicles—as well as technicians to run them. It is a slow, resource-intensive process that scales poorly with the pace of modern software development. For Iveco Group, the Italian commercial vehicle manufacturer behind the eDaily electric van and a fleet of trucks and buses, closing that gap was an opportunity to develop software in the same way as the best technology companies.
“In the tech world, a 16-year-old teenager can change the world from their bedroom because they are empowered with technology and don’t have to ask anybody,” says Alessio Canepa, head of vehicle controls and dynamics software development at Iveco Group. “In automotive, we are often completely on the other side. You always need to wait for somebody else.”
Canepa leads the team behind eVECOP, Iveco Group’s in-house vehicle control software. As it grew in complexity, so did the testing burden. The more functions the team added, the more tests had to be rerun every time anything changed. Doing that manually on physical hardware was becoming impossible.
Iveco Group’s answer was Virtualis, a virtual vehicle simulator built in MATLAB® and Simulink®. It lets Iveco Group run its real production software against a fully virtual vehicle, including virtual battery, inverter, and charger systems, before any hardware ever enters the picture.
Scaling eVECOP for Virtualis
The eVECOP platform sits at the center of Iveco Group’s electric-vehicle control systems, coordinating dozens of interconnected functions across light vehicles, heavy trucks, and buses. These functions include everything from traction control and regenerative braking to thermal management and power distribution, all flowing through one in-house software platform built almost entirely in MATLAB and Simulink.
Previously, rebuilding the entire integrated model took roughly 20 minutes every time anything changed. Now, it takes 4 to 5 minutes to complete the model compilation. Part of that improvement came from switching to accelerator mode in Simulink.
Keeping eVECOP clean and scalable requires what Canepa calls a modular architecture, where each software function lives in its own self-contained module with clearly defined boundaries. A good architect, he says, is extremely important for a software platform team. When the architecture is done right, less experienced engineers can contribute to individual components without needing to understand the entire system. As a result, the platform can expand to cover new vehicle types without engineers needing to start from scratch.
Iveco Group built Virtualis on the same principle. Using referenced models in Simulink, the virtual vehicle simulator keeps each software module separate. This means that when something changes, only that module is recompiled.
Previously, rebuilding the entire integrated model took roughly 20 minutes every time anything changed. Now, it takes 4 to 5 minutes to complete the model compilation. Part of that improvement came from switching to accelerator mode in Simulink. Accelerator mode trades a small amount of debugging flexibility for significantly faster execution. The bigger gain came from the referenced models themselves. Because the code generation engine only recompiles what changed, most builds skip the majority of the work entirely. For a team making constant incremental updates across many different functions, the difference compounds quickly.
Speed matters even more at simulation time. Virtualis runs faster than real time, which means the team can create multiple parallel instances in the cloud and run tests simultaneously. The inspiration, Canepa says, came from watching how software companies operate.
“If you want to design your own Android® app, you use an emulator. You can test things end to end by yourself, autonomously,” he says. “A lot of what we are doing is driven by what we see happening in the cloud and app world.”
The key insight behind Virtualis is that this rapid testing works only when the environment is 100 percent virtual. If even one percent of the testing requires physical hardware, you lose the ability to run in the cloud, run in parallel, and run faster than in real time. That threshold is what makes Virtualis fundamentally different from what came before.
Testing Without Limits
The advantages of a fully virtual environment go beyond speed. Some tests can’t be performed on a physical testbench or a real vehicle; it is neither safe nor reasonable to test a car’s electrical components or wiring while it’s moving 100 kilometers per hour. In Virtualis, those tests are routine. “With a virtual platform, you have unlimited possibilities,” Canepa says.
“With a virtual platform, you have unlimited possibilities.”
Virtualis currently catches about 70 percent of software integration problems before the software ever reaches hardware. Canepa says it is hard to measure precisely, and the residual gap is not a fixed limitation so much as a reflection of what the team doesn’t yet know. When something fails on a physical vehicle that passed in Virtualis, Iveco Group treats it as a signal that the virtual environment needs to be updated to reflect reality more closely. Over time, that protocol is how the gap closes.
The gap that remains is structural rather than incidental. Virtualis performs level 1 virtualization, meaning it does not simulate the underlying firmware or electronics. That is a deliberate tradeoff. Going deeper would mean more coverage, but engineers would lose the ability to stop the simulation mid-execution, inspect every variable, and understand exactly what the software was doing at any given moment.
“When something goes wrong on a hardware system, you look at the logs after the fact,” Canepa says. “In Virtualis, you can stop the world.”
The shift has changed how Iveco Group thinks about its testing pipeline. Iveco Group now uses hardware-in-the-loop testing specifically for what Virtualis cannot reach: the operating system, low-level drivers, and physical hardware that comes from outside Iveco Group (for example, from the suppliers who build the ECUs that the software communicates with). Everything above that layer gets resolved earlier, faster, and cheaper.
Building at the Speed of Software
Getting Virtualis to this point was not straightforward. Scaling a MATLAB and Simulink environment to handle software of this complexity, while keeping build times short and debugging flexible, required working through architecture decisions that don’t have a single right answer. Iveco Group partnered closely with MathWorks engineers to find the limits of what was possible.
“What we have built is already a completely different way of working. The more time you save in development, the more time you have to make something that is actually great.”
“We want to achieve the highest quality in an exceptionally short time frame. That is the toughest challenge we had to face together,” Canepa says.
The collaboration focused on improving execution efficiency, structuring models so multiple engineers could commit updates simultaneously without conflicts, and building an end-to-end continuous integration and testing pipeline that runs automatically. As soon as a developer commits new code, the code triggers tests in Virtualis. Engineers could get results back within hours. Nobody needs to remember to run the tests because the system does not give them the option to forget.
Iveco Group puts the time saved back into the product. Rather than spending weeks on formal verification processes, Iveco Group’s engineers now have more cycles to iterate on the vehicle itself, which Canepa sees as the real measure of software quality.
Virtualis has also changed something less tangible about how the team works. In traditional automotive development, an engineer writing a single function might never see where that function goes or what it does to the vehicle. Virtualis puts every developer inside a complete virtual vehicle, so the effect of any change is immediately visible in context. That visibility changes how people feel about their work. An engineer can prototype a new feature, test its effect on the whole system, and know it will work when they demonstrate it.
“We are in a year where empowering people is tremendously important,” Canepa says.
For him, Virtualis is part of something bigger than faster build times or a more efficient test pipeline. Automakers are becoming software companies whether they are ready or not. The ones that figure out how to build and test software the way technology companies do—fast, iterative, virtual first—will be the ones that keep pace. Iveco Group intends to be one of them.
“What we have built is already a completely different way of working,” he says. “The more time you save in development, the more time you have to make something that is actually great.”
Read Other Stories
The Orion Spacecraft Is Headed to the Moon, with Help from the SLS Rocket
NASA’s Artemis Program Targets a Long-Term Lunar Presence
Applying Model-Based Design for Automotive Service-Oriented Architectures
Renault Advances ADAS Prototyping with ROS Toolbox and Simulink Integration
Powering Next-Generation Surgical Robotics
Model-Based Design Enables Development of Adaptive Surgical Robots