Abstracts

The world of automotive software testing is becoming increasingly complex. As vehicles rely more and more on software, the volume and intricacy of code continue to grow. One of the biggest challenges lies in handling the growing number of software variants: different vehicle models, equipment lines, regional adaptations, and ECU configurations must all be tested individually. Each combination can result in different behaviors and interactions between systems. That's why comprehensive test coverage is essential to detect hidden issues and regressions early, especially in safety-critical systems. Automation testing tools are essential for verifying embedded systems, but writing test cases remains a time-consuming and manual task.

At tracetronic, we're tackling this challenge head-on with a new approach: AI-powered test case generation. Our solution leverages artificial intelligence to automatically create test cases and seamlessly integrate them into different test environments. By integrating the AI directly into ecu.test, we can access all the necessary information about the system under test - from signal definitions to interface configurations - enabling the generation of fully functional software tests for embedded systems right out of the box. This not only accelerates the test development process but also holds the promise of significantly improving test coverage and therefore the software quality. Unlike many existing approaches that rely on static templates or external model descriptions, our assistant generates executable and context-aware test cases.

To ensure that our AI-generated test cases meet quality standards, we have collaborated with the Fraunhofer Institute IIS/EAS to develop a robust evaluation framework for our AI model. Our motivation: trust and reliability are paramount in safety-critical domains like embedded software. Together, we assess key quality attributes such as correctness, completeness, stability, and syntactic and semantic validity. This partnership helps us validate the behavior of the AI and continuously monitor its performance.

In this session, we’ll take you behind the scenes of our development process: How did we build this AI-powered assistant? What hurdles did we face along the way? And where do we see its future potential? Together with our research partners at Fraunhofer, we’ll highlight the importance of establishing evaluation methods for developing new AI systems. Join us to explore how AI is redefining the way we test embedded software - faster, smarter, and more efficient than ever before.