A Dual Large Language Models Architecture with Herald Guided Prompts for Parallel Fine Grained Traffic Signal Control

2025-10-31

AI Summary

Existing LLM-based traffic signal control (TSC) methods suffer from fixed phase durations and hallucination errors, while conventional reinforcement learning approaches exhibit poor robustness and limited generalization. This paper proposes HeraldLight, a dual-LLM collaborative architecture comprising a Guidance Module—real-time prediction of queue lengths per phase—and a Decision Module—generation and correction of signal actions—orchestrated via the Herald Guidance Prompting framework for temporal-aware, fine-grained parallel control. It innovatively integrates LLM reasoning with real-time contextual awareness and introduces a scoring mechanism for output refinement. Evaluated across 224 intersections in Jinan, Hangzhou, and New York, HeraldLight achieves an average 20.03% reduction in travel time and a 10.74% decrease in queue length. The method significantly enhances TSC accuracy, robustness, and cross-scenario generalization.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Search and Optimization: Learning to Search
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Leveraging large language models (LLMs) in traffic signal control (TSC) improves optimization efficiency and interpretability compared to traditional reinforcement learning (RL) methods. However, existing LLM-based approaches are limited by fixed time signal durations and are prone to hallucination errors, while RL methods lack robustness in signal timing decisions and suffer from poor generalization. To address these challenges, this paper proposes HeraldLight, a dual LLMs architecture enhanced by Herald guided prompts. The Herald Module extracts contextual information and forecasts queue lengths for each traffic phase based on real-time conditions. The first LLM, LLM-Agent, uses these forecasts to make fine grained traffic signal control, while the second LLM, LLM-Critic, refines LLM-Agent's outputs, correcting errors and hallucinations. These refined outputs are used for score-based fine-tuning to improve accuracy and robustness. Simulation experiments using CityFlow on real world datasets covering 224 intersections in Jinan (12), Hangzhou (16), and New York (196) demonstrate that HeraldLight outperforms state of the art baselines, achieving a 20.03% reduction in average travel time across all scenarios and a 10.74% reduction in average queue length on the Jinan and Hangzhou scenarios. The source code is available on GitHub: https://github.com/BUPT-ANTlab/HeraldLight.

Problem

• Addresses traffic signal control limitations in LLM-based and RL methods
• Proposes dual LLM architecture for fine-grained signal timing optimization
• Reduces traffic delays and queue lengths through error-correcting mechanisms

Innovation

• Dual LLMs architecture for traffic signal control
• Herald guided prompts extract contextual traffic information
• Score-based fine-tuning corrects hallucinations and errors