航天嵌入式软件测试需求智能提取技术与实践OA

In mission-critical military domains such as aerospace embedded software,where stringent require-ments for safety,real-time performance,and reliability are paramount,testing serves as the core safeguard for software quality.The precision and efficiency of test requirement extraction directly determine the effectiveness of implementing full-process automation and intelligence in testing.However,current traditional methods rely heavily on manual operations,suffering from inefficiencies,incomplete coverage,and susceptibility to expert ex-perience.These limitations make it difficult to match the evolving landscape of aerospace embedded software,characterized by surging functional complexity,expanding scale,and doubled evaluation workloads.An intelli-gent extraction scheme driven by the dual cores of"domain knowledge completion and interactive prompting"is proposed.The scheme utilizes a multi-source information input layer to cover full-volume raw data.Through multi-dimensional parsing,cross-source information fusion,retrieval-augmented generation(RAG),and dynamic optimization of interactive prompts,the core processing layer achieves precise extraction of test requirements.Furthermore,the application layer supports the intelligence of the entire testing process,ultimately outputting standardized testing results.Experimental validation on real-world aerospace embedded software projects dem-onstrates that this scheme significantly improves the efficiency of test requirement extraction,achieving a level of standardization that meets or approaches manual performance.Moreover,through human-machine collabora-tion,the integrity and accuracy effectively satisfy practical testing needs.The proposed scheme provides core support for full-process automation in aerospace embedded software testing,facilitating the formation of a novel testing ecosystem.Future work will focus on further expanding capabilities for functional fusion and deep do-main adaptation.