Generative artifcial intelligence-powered multi-agent paradigm for smart urban mobility: Opportunities and challenges for integrating large language models and retrieval-augmented generation with intelligent transportation systems

This chapter explores the integration of generative artificial intelligence (GenAI) technologies, particularly large language models (LLMs) and retrieval-augmented generation (RAG), into multi-agent systems (MASs) for smart urban mobility. The proposed framework leverages intelligent transportation systems (ITS) data, advanced analytics, and simulation models, enabling GenAI agents to provide tailored, context-aware, and human-centric solutions. By pairing LLM agents with retrieval agents and task-specific agents, the MAS can efficiently process traffic data, interpret user queries, and interact with simulations or optimization services. This approach aims to improve scalability, accessibility, and responsiveness in managing congestion, enhancing road safety, and reducing emissions. We discuss how GenAI-powered MASs can personalize route guidance, support traffic operators and planners with strategic insights, and improve public engagement through intuitive conversational interfaces. The chapter also identifies key challenges, including task orchestration, data sovereignty, and AI accountability. Addressing these barriers is essential to ensure trust and reliability in future smart mobility solutions. Overall, the chapter highlights that integrating LLMs, RAG, and MASs can transform ITS into a more adaptable, user-friendly, and sustainable ecosystem, paving the way for next-generation urban transportation services.