Special Feature • Master's Thesis
Knowledge Graph-Augmented RAG for Automotive Safety Case Generation
Expected start Aug 2026 (Registration Pending)
Institution: Technical University of Munich (TUM)
Chair: Chair of Robotics, AI and Real-Time Systems (Prof. Alois Knoll)
Advisor: André Schamschurko
Overview
This thesis investigates a novel retrieval layer for Retrieval-Augmented Generation (RAG) in safety-critical automated software generation. By replacing flat vector-based RAG with a structured Knowledge Graph retrieval layer, the project aims to retain regulatory semantics and structural dependencies required for safety case compliance.
The domain focus is CeCaS, an engineering pipeline designed to generate Ecore and OCL safety case code directly from automotive regulatory standards, specifically UN Regulation 152 (Advanced Emergency Braking Systems for M1 and N1 vehicles).
Research Objectives and Methodology
- Knowledge Graph Retrieval Layer: Design and implement a structured Knowledge Graph layer that maps regulatory constraints, vehicle operational design domains, and safety assertions.
- Hallucination Reduction: Quantify the reduction of hallucinated or syntactically invalid safety assertions during automated code generation compared to standard vector databases.
- Benchmarking Framework: Benchmark performance against a baseline RAG system (RegulaRAG) and HippoRAG.
- Evaluation Metrics: Evaluate retrieval and generation accuracy using Precision, Recall, F1 score, alongside syntax success and failure rates in generated Ecore/OCL code.
Living Project Updates
This page serves as the central living hub for updates regarding the thesis. Architecture diagrams, benchmark results, and publication materials will be added here as milestones are completed.