Improving the efficiency of software development is a critical challenge in the automotive industry, particularly as system complexity increases. This study addresses the integration test process by developing a method to automatically generate test scripts from natural language test case specifications using a Large Language Model (LLM). To overcome the lack of domain-specific knowledge in LLMs regarding the Application Programming Interfaces (APIs) of automotive test tools, we employ Retrieval Augmented Generation (RAG) with a carefully constructed vector store that incorporates both API manuals and supplemental workflow information. Evaluation on sample test cases from an Electronic Control Unit (ECU) development project demonstrates that the proposed approach successfully generates the required scripts and reduces test execution man-hours by 43% compared to manual execution. These results highlight the practical benefits of context-enriched LLM utilization for automating specialized software engineering tasks in the automotive domain.
Pedro Luís Fonseca Critical TechWorks and Faculty of Engineering, University of Porto, Bruno Lima LIACC, Faculty of Engineering, University of Porto, João Pascoal Faria Faculty of Engineering, University of Porto and INESC TEC
