X-Light: Cross-City Traffic Signal Control Using Transformer on Transformer as Meta Multi-Agent Reinforcement Learner

TL;DR

X-Light introduces a dual-transformer model for cross-city traffic signal control, significantly enhancing transferability and generalization across diverse urban environments, outperforming baseline methods in unseen scenarios.

Contribution

The paper proposes a novel Transformer on Transformer architecture for meta multi-agent traffic signal control, improving transferability across different cities.

Findings

Surpasses baseline methods by +7.91% on average in transfer scenarios.

Achieves up to +16.3% improvement in some cases.

Demonstrates robust generalization across diverse city environments.

Abstract

The effectiveness of traffic light control has been significantly improved by current reinforcement learning-based approaches via better cooperation among multiple traffic lights. However, a persisting issue remains: how to obtain a multi-agent traffic signal control algorithm with remarkable transferability across diverse cities? In this paper, we propose a Transformer on Transformer (TonT) model for cross-city meta multi-agent traffic signal control, named as X-Light: We input the full Markov Decision Process trajectories, and the Lower Transformer aggregates the states, actions, rewards among the target intersection and its neighbors within a city, and the Upper Transformer learns the general decision trajectories across different cities. This dual-level approach bolsters the model's robust generalization and transferability. Notably, when directly transferring to unseen scenarios, ours surpasses all baseline methods with +7.91% on average, and even +16.3% in some cases, yielding the best results.