Konzeption und Implementierung eines interaktiven Webtools zur Exploration und Kuration von sicherheitsrelevantem Wissen für hoch-automatisierte Systeme

Abstract

HazardDB contains 299 interconnected criticality phenomena for safety assurance of highly automated driving systems. However, its use requires time-consuming manual review or generic tools without domain-specific support. Curating new phenomena is error-prone, and lacking incentives threaten long-term maintenance. This thesis develops an interactive web tool for intuitive exploration and efficient maintenance of HazardDB. The solution is based on a three-tier architecture with RDF triple store, LLM-assisted curation, and automated quality checks. The system combines hierarchical navigation, interactive graph visualization, and intelligent suggestions for descriptions, tags, and relations. Scenario-based evaluation shows: exploration is intuitive, curation is significantly accelerated by LLM assistance, and automated health checks ensure data quality.