Marelli and AWS automate SDV test case generation

The modern motor vehicle is, by almost any metric, more software than machine. A contemporary premium car may contain upwards of 100 million lines of code, distributed across dozens of electronic control units and increasingly orchestrated by centralised computing platforms designed to be updated long after the vehicle leaves the production line. For the engineering teams responsible for ensuring that all of this software does precisely what it is supposed to do, the validation burden has become one of the defining technical challenges of the decade.

The response from Marelli, the global automotive technology supplier, and Amazon Web Services (AWS) is a tool that targets this problem at its most labour-intensive point. The two companies have developed an AI-driven System Test Generation (STG) Agent that automates the generation of system test cases directly from engineering system requirements. It is not a peripheral efficiency improvement. It addresses the structural core of how vehicle software is validated, and in doing so, points to a broader transformation in how automotive engineering is practised.

The weight of software complexity in modern vehicles

The shift towards software-defined vehicles has altered the engineering approach for automotive suppliers in ways that are only beginning to be fully understood. Where vehicle behaviour was once largely fixed at the point of manufacture, modern platforms are reconfigurable, updatable and dependent on software for core functionality that earlier generations handled mechanically. The commercial advantages are substantial, while the engineering implications are far-reaching.

Managing the resulting volumes of requirements, data and system specifications demands a level of rigour and traceability that manual processes struggle to sustain at scale. A single customer platform may generate tens of thousands of individual system requirements, each of which must ultimately be reflected in test cases that verify the corresponding product behaviour meets its specification. Doing this by hand is slow, susceptible to inconsistency and resistant to the acceleration that vehicle makers are now demanding of their supply base.

Read the article

Amazon's AWS outage exposes automotive production's digital vulnerability

Amazon's 15-hour AWS outage affected billions in transactions globally. With major OEMs running production systems on the same infrastructure, the incident raises urgent questions about automotive manufacturing's digital resilience.

The pressure, however, is structural rather than cyclical. Development timelines for software-defined vehicle programmes are compressing; engineering headcount has not grown proportionally with the complexity of the software those engineers are asked to validate, and tools that address this imbalance by amplifying the productive capacity of existing teams, without sacrificing traceability or consistency, are no longer merely desirable, but are an operational necessity.

How the STG Agent automates validation at source

Within Marelli's established development process, customer requirements are first translated by R&D engineers into system requirements, a human-driven step that defines precisely what a product must do. It is from this point that the STG Agent takes over.

The tool analyses each system requirement, identifies the expected behaviours implied by it, and then generates corresponding system test cases that are clear, structured and fully traceable. The outputs are designed to support Marelli engineers in verifying that each product feature behaves exactly as the requirement specifies. This is not an approximation. It is a direct and auditable mapping from requirement to test, produced at a speed and volume that manual methods cannot match. 

“The STG Agent represents an important step forward in how we validate solutions for software-defined vehicles,” said Daniele Russo, Head of System Performance Optimization in Marelli’s Electronics Engineering team. “By combining our engineering expertise with advanced AI capabilities from AWS, we significantly accelerate validation cycles and ensure consistent quality across global programs. This solution enables us to support our customers faster and more efficiently, strengthening the foundation for the next generation of software-defined vehicles.”

The agent's design reflects a considered view of where automation adds the most value in a complex engineering workflow. The human judgement that shapes system requirements from customer specifications is preserved entirely. What changes is the mechanical and repetitive work of converting those requirements into test cases, a task that is both voluminous and demanding of precision, and one that AI, given the right architecture, can perform more reliably and at far greater scale than human effort alone.

Equally significant is what the solution does not disrupt. It is designed for easy integration with existing requirement management tools and built for compatibility with established automotive engineering workflows. Adoption does not require wholesale change to validated processes, a consideration that carries particular weight in an industry where process certification and audit trails are regulatory requirements rather than optional niceties.

The Amazon generative AI stack behind the solution

The STG Agent was developed with the expertise of the AWS Generative AI Innovation Center, drawing on a set of capabilities that reflects the current frontier of applied large language model deployment in industrial settings. At its core are Amazon Nova foundation models, which provide the language understanding and generation capability that makes automated test case synthesis possible. These are supported by Amazon Bedrock Knowledge Bases, which allow the agent to draw on curated, domain-specific engineering knowledge rather than relying on generalised model training alone, a distinction that matters considerably in a field where precision is not optional.

Read the article

From BMW to Stellantis & NXP: Software-defined vehicles are transforming the automotive industry

It’s now time to take a more holistic view of software-defined vehicles (SDVs). After all, they involve completely new business models. Who is pursuing which strategy?

The orchestration layer is provided by the Strands Agents framework, an open-source tool for building, deploying and managing AI agents in a modular and extensible fashion. The architecture suggests a system designed to evolve as both the underlying models and engineering requirements change, rather than a fixed solution tied to a specific configuration. As foundation model capability continues to improve, the agent’s performance can improve with it.

“Marelli’s approach to automating system validation demonstrates the transformative potential of generative AI in automotive engineering,” said Giulia Gasparini, Country Leader of AWS Italia. “By leveraging Amazon Nova foundation models and Amazon Bedrock, companies are setting new standards for how software-defined vehicles are developed and validated. This solution shows how advanced AI can accelerate innovation while maintaining the rigorous quality and safety requirements that define the automotive industry.”

A structural shift in how suppliers manage quality at scale

The immediate gains from the STG Agent are measurable in time saved and in the tighter alignment between system requirements and validated product behaviours. But the more significant implication is what this kind of tool represents for the operating model of automotive suppliers at large.

The industry's quality frameworks have long placed documentation and traceability at their centre. Requirement management, test planning and validation reporting generate administrative workloads that consume engineering capacity that would otherwise be directed at substantive technical work. A tool that produces structured, traceable test cases as a natural output of its operation, rather than as a retrospective documentation exercise, changes the economics of that administrative burden in ways that compound across large, multi-market programmes.

Read the article

Cybersecurity now tops production concerns as smart factories expand vulnerabilities

Cybersecurity has vaulted to manufacturers' top strategic priority as 84% rate digital threats as critical - yet the very smart factory systems delivering productivity gains are creating unprecedented attack surfaces across connected production environments

There is also a consistency argument that merits careful attention. Manual test case generation is not merely slow; it is inherently variable. Different engineers, working under different pressures at different stages of a programme, interpret the same requirement with different levels of rigour. That variability introduces risk that is difficult to detect until late in the development cycle, when the cost of resolution is highest. Automation imposes a structural consistency that human processes cannot sustain at the volumes software-defined vehicle development now demands.

For vehicle manufacturers evaluating their supplier relationships, the ability to demonstrate AI-generated, traceable and structured validation processes may increasingly carry weight in both commercial and technical assessments. The automotive industry is a procurement-driven environment, and suppliers that can credibly demonstrate both development speed and validation quality are better positioned in an intensifying competitive landscape.

Marelli's collaboration with AWS reflects a broader pattern in which tier-one suppliers are turning to technology partnerships to deliver engineering capabilities that were not accessible even three years ago. Generative AI applied not to the vehicle itself but to the engineering processes that underpin its development is beginning to reshape the industry's operational architecture. The STG Agent is one concrete instance of that shift, and it is unlikely to be the last.