CuraLight: Debate-Guided Data Curation for LLM-Centered Traffic Signal Control

2026-04-07

AI Summary

This study addresses key limitations in existing reinforcement learning (RL) and large language model (LLM)-based traffic signal control methods, which often suffer from poor interpretability, scarce interaction data, and weak generalization across heterogeneous intersections. To overcome these challenges, the authors propose a novel LLM-centric control framework that leverages RL agents to explore the environment and generate high-quality trajectories. A multi-LLM structured debate mechanism is then introduced to evaluate signal timing actions, producing preference-aware supervision signals for fine-tuning the model. This approach innovatively integrates RL-assisted exploration with multi-agent debate to automatically construct interpretable, high-quality control data and enable efficient training. Extensive experiments on real-world urban road networks in SUMO demonstrate that the proposed method outperforms state-of-the-art baselines, achieving average reductions of 5.34% in travel time, 5.14% in queue length, and 7.02% in waiting time.

Abstract

Traffic signal control (TSC) is a core component of intelligent transportation systems (ITS), aiming to reduce congestion, emissions, and travel time. Recent approaches based on reinforcement learning (RL) and large language models (LLMs) have improved adaptivity, but still suffer from limited interpretability, insufficient interaction data, and weak generalization to heterogeneous intersections. This paper proposes CuraLight, an LLM-centered framework where an RL agent assists the fine-tuning of an LLM-based traffic signal controller. The RL agent explores traffic environments and generates high-quality interaction trajectories, which are converted into prompt-response pairs for imitation fine-tuning. A multi-LLM ensemble deliberation system further evaluates candidate signal timing actions through structured debate, providing preference-aware supervision signals for training. Experiments conducted in SUMO across heterogeneous real-world networks from Jinan, Hangzhou, and Yizhuang demonstrate that CuraLight consistently outperforms state-of-the-art baselines, reducing average travel time by 5.34 percent, average queue length by 5.14 percent, and average waiting time by 7.02 percent. The results highlight the effectiveness of combining RL-assisted exploration with deliberation-based data curation for scalable and interpretable traffic signal control.

Problem

• traffic signal control
• large language models
• data curation
• generalization
• interpretability

Innovation

• LLM-centered control
• RL-assisted data curation
• multi-LLM debate
• imitation fine-tuning
• interpretable traffic signal control