Multi-Intersection Traffic Optimisation: A Benchmark Dataset and a Strong Baseline

Abstract

The control of traffic signals is fundamental and critical to alleviate traffic congestion in urban areas. However, it is challenging since traffic dynamics are complicated in real-world scenarios. Because of the high complexity of the optimisation problem for modeling the traffic, experimental settings of existing works are often inconsistent. Moreover, it is not trivial to control multiple intersections properly in real complex traffic scenarios due to its vast state and action space. Failing to take intersection topology relations into account also results in inferior solutions. To address these issues, in this work we carefully design our settings and propose a new dataset including both synthetic and real traffic data in more complex scenarios. Additionally, we propose a novel baseline model with strong performance. It is based on deep reinforcement learning with an encoder-decoder structure: an edge-weighted graph convolutional encoder to excavate multi-intersection relations; and an unified structure decoder to jointly model multiple junctions in a comprehensive manner, which significantly reduces the number of the model parameters. By doing so, the proposed model is able to effectively deal with the multi-intersection traffic control optimisation problem. Models are trained/tested on both synthetic and real maps and traffic data with the Simulation of Urban Mobility (SUMO) simulator. Experimental results show that the proposed model surpasses multiple competitive methods. The traffic data and the code can be found at https://git.io/JPdU1.

Model-Based Graph Reinforcement Learning for Inductive Traffic Signal Control

“…However, an increase in parameters leads to an explosion of costs (computational power, memory, and data requirements) with the number of intersections, which hinders further scalability. Parameter sharing in MARL methods (Devailly et al, 2021;Wei et al, 2019b;Zang et al, 2020;Chen et al, 2020;Wang et al, , 2021 enables far greater scalability (up to a thousand controllers). Such methods usually rely on GCNs (Devailly et al, 2021;Wei et al, 2019b;Chen et al, 2020;Shi et al, 2024).…”

Reinforcement Learning-Based Traffic Control: Mitigating the Adverse Impacts of Control Transitions

“…Other areas of research and application where deep learning and RL have proven useful are traffic signal control [36], [37], [38] connected automated vehicles in mixed autonomy traffic [39], variable speed limit control at bottlenecks and ramps [40], or at round-abouts [41], to name but a few.…”

Bandwidth-Based Traffic Signal Coordination Models for Split or Mixed Phasing Schemes in Various Types of Networks

“…In the second category, TRANSYT [7] and TRANSYT-7F [8] are the frequently used simulation-based traffic signal timing optimization programs. Artificial intelligence methods represented by machine learning have been widely used in the field of transportation, such as traffic safety-related prediction [9], [10], traffic state prediction [11], [12], automated driving [13], [14], and traffic signal control [15], [16], [17]. References [15], [16], and [17] report the use of deep reinforcement learning to estimate the optimal traffic signal timing plans, which can be classified into the third category.…”

“…Artificial intelligence methods represented by machine learning have been widely used in the field of transportation, such as traffic safety-related prediction [9], [10], traffic state prediction [11], [12], automated driving [13], [14], and traffic signal control [15], [16], [17]. References [15], [16], and [17] report the use of deep reinforcement learning to estimate the optimal traffic signal timing plans, which can be classified into the third category. These three types of methods can greatly improve the operational efficiency of the traffic streams.…”