LLM-Assisted Requirements Decomposition in Automotive Software Engineering - A Case Study at Volvo Cars

Sammanfattning

Requirements decomposition in automotive software engineering is a complex andcontext-dependent activity that requires architectural understanding, abstractionlevel judgement, and domain expertise. Although Large Language Models (LLMs)have shown potential in several software engineering tasks, their use for requirements decomposition in safety-critical automotive contexts remains insufficientlyunderstood.This thesis investigates how LLM-based assistants can support requirements decomposition in automotive software engineering through a case study at Volvo Cars. Thestudy follows a Design Science Research methodology combining semi-structuredinterviews, iterative artefact development, and industrial evaluation with domainexperts. The resulting human-in-the-loop artefact combines contextual grounding through standards, historical decompositions, and system-structure informationwith hierarchy guidance, structured prompt orchestration, and schema-constrainedgeneration.The artefact was evaluated through benchmark-based assessments, contextual evaluations using participant-selected requirements, and qualitative feedback sessionswith automotive domain experts. The results suggest that LLM-based assistantscan support requirements decomposition by reducing cognitive effort, improvingcontextual awareness, and providing alternative decomposition perspectives duringrefinement. The evaluation further indicates that contextual grounding, hierarchyguidance, and structured orchestration positively influenced expert-perceived output quality and reviewability.At the same time, important limitations remain related to abstraction-level consistency, incomplete contextual understanding, and the need for expert validation insafety-critical engineering contexts. The findings suggest that LLM-based decomposition assistants are most useful as decision-support tools within human-in-the-loopworkflows rather than as autonomous requirement generation systems.