Chat2Scenic: An Iterative RAG-Based Framework for Scenario Generation in Autonomous Driving

arXiv cs.AI · 6 hours ago

Chat2Scenic is a retrieval-augmented framework that uses a chatbot interface to generate autonomous driving test scenarios in domain-specific language from regulatory descriptions. The system achieves 76.42% compilation success rate and 58.17% framework accuracy, substantially outperforming existing retrieval-based methods that reach only 16-30% compilation rates. This enables more reliable and scalable automated generation of regulation-compliant test scenarios for validating autonomous driving systems.