Multi-Agent Deep Reinforcement Learning for Large-Scale Traffic Signal Control

Reinforcement learning (RL) is a promising data-driven approach for adaptive traffic signal control (ATSC) in complex urban traffic networks, and deep neural networks further enhance its learning power. However, the centralized RL is infeasible for large-scale ATSC due to the extremely high dimension of the joint action space. The multi-agent RL (MARL) overcomes the scalability issue by distributing the global control to each local RL agent, but it introduces new challenges: now, the environment becomes partially observable from the viewpoint of each local agent due to limited communication among agents. Most existing studies in MARL focus on designing efficient communication and coordination among traditional Q-learning agents. This paper presents, for the first time, a fully scalable and decentralized MARL algorithm for the state-of-the-art deep RL agent, advantage actor critic (A2C), within the context of ATSC. In particular, two methods are proposed to stabilize the learning procedure, by improving the observability and reducing the learning difficulty of each local agent. The proposed multi-agent A2C is compared against independent A2C and independent Q-learning algorithms, in both a large synthetic traffic grid and a large real-world traffic network of Monaco city, under simulated peak-hour traffic dynamics. The results demonstrate its optimality, robustness, and sample efficiency over the other state-of-the-art decentralized MARL algorithms.

I. Introduction

As a consequence of population growth and urbanization, the transportation demand is steadily rising in the metropolises worldwide. The extensive routine traffic volumes bring pressures to existing urban traffic infrastructure, resulting in everyday traffic congestions. Adaptive traffic signal control (ATSC) aims for reducing potential congestions in saturated road networks, by adjusting the signal timing according to real-time traffic dynamics. Early-stage ATSC methods solve optimization problems to find efficient coordination and control policies. Successful products, such as SCOOT [1] and SCATS [2], have been installed in hundreds of cities across the world. OPAC [3] and PRODYN [4] are similar methods, but their relatively complex computation makes the implementation less popular. Since the 90s, various interdisciplinary techniques have been applied to ATSC, such as fuzzy logic [5], genetic algorithm [6], and immune network algorithm [7].