Generating test specifications that meet Automotive SPICE SWE.6 requirements is increasingly challenging and time-consuming as projects scale to thousands of requirements. The manual process often consumes weeks of engineering effort, making automation a critical necessity. However, standard Large Language Model (LLM) approaches struggle at scale: processing requirements individually discards vital inter-requirement dependencies, while feeding entire corpora at once exceeds context-window limits, leading to incomplete integration coverage and redundant test cases.
This paper presents a novel "Cluster-then-Summarize" pipeline that addresses these limitations through three stages. Requirements are embedded using sentence transformers and grouped using UMAP dimensionality reduction followed by HDBSCAN density-based clustering. This grouping utilizes an automatic minimum cluster size selection driven by a quality criterion combining normalized Silhouette and Calinski-Harabasz scores. A multi-level map-reduce summarization algorithm then distills each cluster into concise, domain-conformant descriptions while preserving quantitative thresholds and safety integrity levels. The pipeline exploits the derived cluster topology to generate test specifications at two levels: individual requirement verification and cluster-level integration tests that verify cross-requirement feature behavior. A nearby-cluster context mechanism provides bounded cross-feature awareness during each LLM call, and Retrieval-Augmented Generation grounds all outputs in ISO 26262 and ASPICE standards. Evaluation on automotive requirement datasets of varying scale demonstrates that the cluster-aware approach improves integration test coverage and maintains summarization fidelity compared to baseline methods while scaling efficiently to thousands of requirements.
Blogger's Review: This paper introduces an innovative cluster-aware method for generating test specifications in automotive software requirements, successfully addressing the shortcomings of traditional methods in handling dependencies and context limitations. It enhances test coverage and shows high practical value and potential for broader application, providing a solid demonstration for automation in large projects.