Traffic-R1: Reinforced LLMs Bring Human-Like Reasoning to Traffic Signal Control Systems

TL;DR

Traffic-R1 is a 3B-parameter foundation model that enables human-like reasoning in traffic signal control, offering zero-shot generalization, real-time edge deployment, and explainability, significantly improving traffic management efficiency.

Contribution

This paper introduces Traffic-R1, a novel foundation model for traffic signal control that combines reinforcement learning and large language model techniques for the first time.

Findings

Outperforms traditional RL and recent LLM-based methods in benchmarks.

Reduces average queue lengths by over 5% in real-world deployment.

Halves operator workload in traffic management.

Abstract

We introduce Traffic-R1, a 3B-parameter foundation model with human-like reasoning for Traffic signal control (TSC), developed via self-exploration and iterative reinforcement of LLM with expert guidance in a simulated traffic environment. Compared with traditional reinforcement learning and recent LLM-based methods, Traffic-R1 offers three main advantages: zero-shot generalization, transferring unchanged to new road networks and out-of-distribution incidents by leveraging internal traffic-control policies and reasoning; a compact 3B-parameter design that supports real-time inference on mobile-class chips for edge deployment; and an explainable TSC process that enables multi-intersection coordination through communication and an asynchronous communication network. Extensive benchmarks show Traffic-R1 outperforms strong baselines and training-intensive RL controllers. In production, the model now manages signals affecting over 55,000 drivers daily, reduces average queue lengths by more than 5%, and halves operator workload. Our model is available at https://huggingface.co/Season998/Traffic-R1.