Application and Optimization of Retrieval-Augmented Generation in Automotive Software Test Case Generation
Min Xia, Guang Tong · 2025
This study proposes a Retrieval-Augmented Generation (RAG) based Test Case Generation Framework integrated with Multi-Strategy Routing, enabling intelligent adaptation from rapid prototyping to rigorous verification through a Three-Tier Strategy Network. A progressive three-level test case generation strategy system is introduced, consisting of Basic One-Time Generation (Strategy A), Comprehensive Requirement Coverage Verification (Strategy B), and Compatibility Incremental Enhancement (Strategy C). By constructing a Dynamic Strategy Routing Network, intelligent matching between requirement complexity and generation strategies is achieved. Experimental results demonstrate that the three strategies exhibit effective complementarity in key metrics such as requirement coverage rate ($93.7 \%$ vs. $62.3 \%$) and generation efficiency (33.2s vs. 55.6s), providing optimal solutions across different testing stages.