RAGTraffic: Utilizing Retrieval-Augmented Generation for Intelligent Traffic Signal Control

Traffic Signal Control (TSC) plays a critical role in optimizing urban traffic flow and reducing congestion. Traditional TSC methods, including transportation engineering and reinforcement learning (RL), often struggle to generalize across different traffic scenarios and lack interpretability. This paper presents the RAG Traffic framework, which uses Retrieval-Augmented Generation (RAG) to increase decision accuracy by integrating historical traffic data with real-time information through a retrieval mechanism. The framework employs agent-based simulations to mimic the decision-making processes of traffic management experts, incorporating an iterative chain-of-thought approach to improve performance in multi-step reasoning tasks. Extensive experiments on five real-world datasets demonstrate that RAG Traffic significantly outperforms RL-based baseline methods in terms of interpretability and generalization capabilities, highlighting its superior adaptability to the complexities and dynamics of the traffic domain.