### Introduction
The automotive industry is grappling with an ever-increasing complexity in software development, driven by demands for electrification and autonomous driving, alongside pressing concerns around performance and security. The Virtual Engineering Workbench (VEW) framework, powered by AWS services and generative artificial intelligence, has emerged as a transformative solution, enhancing automotive software development and testing through advanced virtualization, improved developer efficiency, collaborative opportunities, and streamlined test case creation.
In this exploration, we focus on Schaeffler, a global leader in motion technology and automotive innovation. Schaeffler’s collaboration with AWS presents compelling insights into cutting-edge advancements in virtual testing and how they’re being applied in real-world automotive scenarios. Schaeffler’s deep commitment to digitalization and AI aligns seamlessly with AWS’s mission of empowering organizations with cloud technologies. Together, they are shaping the future of automotive software development and testing.
### The VEW in Practice
The VEW framework is not merely theoretical but is actively utilized by Schaeffler to support innovation projects. Recent updates reveal enhancements that include key support for architectures based on the AUTOSAR Classic Platform and AI-driven workflows tailored for various software engineering use cases. In collaboration, AWS and Schaeffler have initiated a joint study to identify the most promising generative AI workflows. This examination revealed several high-impact use cases, particularly in the realm of requirements translation, coding assistance, test case generation, and intelligent process guidance.
### Addressing Real-World Challenges
Today’s automobiles are bounded by an extensive landscape of functional and non-functional requirements, often necessitating multiple test cases for rigorous verification. Specifically, Schaeffler faces the daunting task of navigating hundreds of thousands of requirements, many of which are poorly optimized for human understanding. A Test Manager at Schaeffler noted that this results in testers consuming hours each week sifting through complex requirements.
To understand the real-world implications of this challenge, it is essential to recognize the industry standards guiding automotive software testing, particularly ISO 29148 and ISO 29119. These standards dictate a structured four-step process for effective test case generation, which includes reviewing requirements, structuring feature sets, identifying test conditions, and finally, deriving test cases.
### Decoding the Testing Process
Let’s illuminate these steps with a practical example: an automotive adaptive cruise control (ACC) system. The first step involves reviewing the requirements to ensure clarity and testability. In this context, a requirement might state that “the ACC system shall maintain a minimum following distance of 2 seconds,” and it falls upon the test engineer to verify the requirement’s comprehensibility and compliance with industry standards.
Next, structuring features and feature sets is essential for creating a logical testing schema. This could encompass various feature sets like speed control, distance management, and emergency handling.
The third stage necessitates identifying specific test conditions—situations under which each feature should be tested. Scenarios might include normal operation, such as maintaining a following distance, edge cases, and failure events (e.g., sudden vehicle deceleration).
Finally, deriving test cases consolidates all information explored in the previous steps into actionable, detailed documentations, specifying inputs, expected results, and execution sequences.
Unfortunately, this process can be labor-intensive. For instance, experienced testers typically dedicate 820 hours preparing test cases for a mere 837 requirements—an observation underscored by Schaeffler’s data.
### Harnessing Generative AI
To streamline and enhance this arduous process, Schaeffler and AWS forged a dedicated generative AI platform. By leveraging state-of-the-art Large Language Models (LLMs) housed within a secure, organization-wide generative AI gateway, Schaeffler employees can tap into a myriad of AI-driven capabilities. Tools such as Amazon Bedrock ensure responsive access while keeping costs manageable.
The generative AI platform supports improvements in the test-generation workflow, significantly reducing the time required to prepare test cases. In practical application, it facilitates natural language requests to derive insights, remove redundancies, and enhance usability, all while maintaining operational security.
### Streamlining User Experience
However, as efficiency improves, new challenges emerge. Saying goodbye to cumbersome manual conversions from spreadsheets to machine-readable formats, the VEW web application has been upgraded to streamline the transition from AI-generated data outputs. By preserving session history, allowing collaborative work, and guiding users through test case generation workflows, the framework redefines the user experience for Schaeffler’s engineers.
### Architectural Insights
To visualize this foundational structure, the solution architecture is not merely complex but is meticulously designed to represent each step in the test case generation process. Whether it’s through REST API endpoints enabling seamless data transfer, AWS Lambda functions managing data storage in Amazon S3, or leveraging DynamoDB for metadata management—each component plays a crucial role in enhancing operational efficacy.
### Projecting Business Outcomes
The results of these advancements are nothing short of impressive. The implementation of these generative AI capabilities has accelerated the speed of test case generation by up to 60%. For example, the time required to prepare test cases for 837 requirements has plummeted from 820 hours to just 265 hours. Moreover, minimizing the risk of human error brings with it increased confidence in the validity of system requirements, thereby reinforcing the critical nature of this transformation.
—
These developments lay the groundwork for the future, with further integration of VEW into Schaeffler’s existing testing software on the horizon. Expect improvements across the board, from automating the export of requirements to facilitating seamless imports of test cases, all guided by an overarching vision of holistic efficiency powered by generative AI.