Large-Scale Multi-Agent Reinforcement Learning for Traffic Signal Optimization
Keywords: Reinforcement Learning, Traffic Signal Control, Multi-Agent, Transformer
Abstract: We present a novel approach to Traffic Signal Control (TSC) in a multi-agent environment by modeling communication among agents as a sequence problem, enabling intersections within road networks to communicate with one another. Taking inspiration from point cloud processing and graph neural networks, we make our architecture capable of handling variable road network topologies, including differing numbers of intersections and intersection types, and demonstrate this by successfully training on real & randomly generated road networks and traffic demands. Furthermore, we demonstrate that even utilizing minimal state information can achieve competitive performance.
Primary Area: reinforcement learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 14250
Loading