Dev48
Language
  • About
  • Services
  • Industries
  • Technologies
  • Articles
  • Contacts
Book a call
    Home/Articles/Reimagining automotive verification and validation for software defined vehicles
Dev48
info@dev48.org+375 29 320-40-96WhatsApp@isky_by

© 2026 · All rights reserved.

Reimagining automotive verification and validation for software-defined vehicles

Source: HCLTech•November 20, 2026 at 04:26 PM

Reimagining automotive verification and validation for software-defined vehicles

Reimagining automotive verification and validation for software-defined vehicles Content Group Article pallavi.parashar Fri, 11/20/2026 - 05:36 Trends Channel Engineering For more than a century, carmakers became very good at engineering machines that changed slowly. A new model took years to…

For more than a century, carmakers became very good at engineering machines that changed slowly. A new model took years to develop. Hardware choices were fixed early. Testing followed a largely predictable path. Once a car left the factory, its capabilities were mostly fixed.

Software-defined vehicles are changing that model.

Modern cars increasingly rely on centralized computing, connected services, advanced driver-assistance systems and software that can be updated throughout the vehicle’s life.

That changes where automotive advantage comes from.

Horsepower and handling still matter. But computing architecture, software quality and the ability to develop, validate and improve vehicle capabilities throughout their lifecycle matter far more than they once did.

Volvo Cars is expanding its work with NVIDIA around automated driving technologies. Mercedes-Benz is combining its MB.OS architecture with NVIDIA computing technologies for future vehicles. General Motors is also working with NVIDIA technologies across vehicle development, manufacturing and robotics.

These relationships point to a larger shift. Cars are becoming computing systems that happen to move.

That creates a difficult engineering problem. Software can change quickly. Automotive verification and validation must keep pace without weakening confidence in a vehicle’s behavior.

The testing and validation problem gets bigger

A conventional vehicle program could rely heavily on prototypes, test tracks and staged validation before production.

That becomes harder when thousands of software functions interact across a connected vehicle.

Change one part and the consequences may appear somewhere unexpected.

A change to one software function may affect other components that share data or computing resources. A driver-assistance feature needs to work properly not just on a dry highway in daylight, but across different roads, weather conditions, traffic situations and driver behavior.

Physical testing remains essential. It just cannot economically cover every useful combination.

Testing can no longer happen only at defined milestones. Validation must become part of development itself.

That shifts the task from testing a vehicle at defined milestones to continuously asking a harder question: What could this software do under conditions we have not yet tested?

Simulation becomes central

NVIDIA Omniverse provides an environment for simulation and digital twin development. Cosmos extends this into world modeling and synthetic data. Alpamayo is aimed at autonomous-driving reasoning and evaluating complex driving decisions. NVIDIA Halos provides a broader automotive safety system. It brings together simulation and physical testing to help developers generate measurable, traceable safety evidence for automated-driving systems.

These technologies solve different engineering problems. The bigger opportunity comes from connecting them.

Instead of engineers manually defining every test, running it and examining the results, software agents could handle more of that cycle.

They could interpret requirements, propose test scenarios, initiate simulations, identify failures and point engineers toward cases that need closer attention. NVIDIA Agent Toolkit and Nemotron open models can support reasoning and coordination within these workflows.

Engineers remain central. Engineering judgment and safety sign-off cannot be delegated.

Test the rare, not just the routine

This matters most at the edges.

Real roads create combinations that engineering teams cannot easily reproduce on demand.

Unusual lighting. Poor visibility. Unexpected road layouts. Conflicting movements by vehicles and pedestrians. Software changes that appear harmless until they interact with something unforeseen.

Simulation lets engineers examine these conditions earlier without waiting to reproduce them physically.

Physics-based models can reproduce aspects of the real world. Synthetic data can expand the range of situations available for development. For suitable workloads, surrogate models can speed up calculations that would otherwise consume significant computing time.

The goal is not to replace road testing.

It is to expand coverage earlier, find problems sooner and make better decisions about what still needs physical testing before release.

The computer matters as much as the code.

Platforms such as NVIDIA DRIVE AGX Thor support increasingly centralized vehicle computing. A scenario that works in simulation still needs testing on representative target hardware. Timing, memory use and thermal limits can change the outcome.

Virtual electronic control unit simulation can also move selected functional, regression, fault-injection and ECU software tests earlier in the development process, reducing dependence on physical hardware-in-the-loop (HiL) benches. HiL and vehicle-level testing remain essential for integration, timing and safety assurance. There is another part of the problem. Software-defined vehicles continue to change after they enter service.

Every meaningful software change raises a simple question: What needs to be tested again?

The same applies to an over-the-air update before it reaches vehicles already on the road. Running every available test after every change would be expensive and slow. Running too few creates risk.

This makes it essential for manufacturers to understand the dependencies between software components, vehicle functions, requirements and previous test results. That depends on something less glamorous than autonomous driving, but just as important: disciplined engineering data.

Requirements, simulation results, software changes and test evidence often sit in different systems. Connecting them through common data structures, lineage and traceability makes validation more targeted.

If engineers know exactly what changed and what depends on it, they can focus testing where the risk lies.

From test automation to validation intelligence

Our platform is intended to address this gap. The proposition combines our automotive engineering expertise with NVIDIA’s computing, simulation and automotive technologies. The aim is to connect requirements, scenario generation, test orchestration, simulation and traceable validation evidence, rather than treating them as separate engineering activities.

The same approach can support both new-vehicle development and risk-based regression testing before subsequent software releases.

Openness will matter too

Carmakers have good reason to be careful about where they place engineering knowledge, vehicle data and intellectual property.

NVIDIA’s open-model approach, including technologies such as Nemotron, Cosmos and Alpamayo, can give manufacturers more scope to adapt models and engineering workflows to their own requirements, subject to the relevant license terms.

That matters because automotive manufacturers need more than smarter software.

They need control over how it is developed, tested and governed.

The competitive question is changing with it.

The winners in the software-defined vehicle market will not simply be the manufacturers that release features fastest. Speed without confidence brings its own problems. The advantage will lie with companies that can change vehicle software quickly, understand the consequences and produce convincing evidence that it behaves as intended before release. For decades, automotive engineering focused on proving that a vehicle was ready to leave the factory.

The software-defined vehicle changes that standard.

Confidence must now be established throughout development and renewed whenever significant software changes are released.

The opportunity is straightforward: Validate earlier, test what matters and build the evidence behind every release.

← All articles

What this article says

More in Software Development

All →
Vulkan Rendering Is Now The Default on Linux Wayland
JetBrains

Vulkan Rendering Is Now The Default on Linux Wayland

JetBrains Air in GoLand: Build, Review, and Profile Go Code With Agents
JetBrains

JetBrains Air in GoLand: Build, Review, and Profile Go Code With Agents

How to build an ecommerce shipping strategy
Shippo

How to build an ecommerce shipping strategy

Open edX as the Operating Layer for AI Tools
Open edX

Open edX as the Operating Layer for AI Tools

Your coding agent leaked the key. It was only following orders.
Postman

Your coding agent leaked the key. It was only following orders.

Your test email went to spam. Congratulations?
Mailgun

Your test email went to spam. Congratulations?

More from HCLTech

Anthropic Ecosystem
HCLTech

Anthropic Ecosystem

Enhancing media supply chain operations for a global entertainment leader on AWS
HCLTech

Enhancing media supply chain operations for a global entertainment leader on AWS

The resilience gap in always-on media
HCLTech

The resilience gap in always-on media

Beyond AI adoption: Reimagining the workplace around intelligence and outcomes
HCLTech

Beyond AI adoption: Reimagining the workplace around intelligence and outcomes

AI Data Centers for What's Next
HCLTech

AI Data Centers for What's Next

The trust gap: Is your data ready for recovery?
HCLTech

The trust gap: Is your data ready for recovery?