C-MAPPO-TSC: A Collaborative Multi-Agent Reinforcement Learning Framework for Urban Traffic Signal Control and Congestion Mitigation
Fundamental Scientific Reports in Multidisciplinary Areas · 25.06.2026 · 2 (02) · 170-181
DOI: 10.66238/fsrma103 ↗Kurz gefasst
Fachartikel in der Zeitschrift Fundamental Scientific Reports in Multidisciplinary Areas, online erschienen am 25.06.2026. Verfasst von Pinghao Luo und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Urban traffic congestion poses a persistent challenge to smart city development and public transportation system efficiency, largely due to fragmented traffic signal control strategies, complex congestion propagation mechanisms, and limited coordination among intersections. In large-scale urban road networks, isolated signal optimization often leads to local optimality, aggravating network-level congestion and reducing the reliability of public transport operations. To address these challenges, this paper proposes C-MAPPO-TSC, a Collaborative Multi-Agent Proximal Policy Optimization framework for urban traffic signal control and congestion mitigation. In the proposed framework, each signalized intersection is modeled as an autonomous AI agent that makes real-time control decisions based on local traffic observations and shared neighborhood information, enabling coordinated control across the network. A centralized training and decentralized execution paradigm is adopted to stabilize learning while preserving scalability, and a hybrid reward mechanism is designed to jointly optimize local intersection performance and global traffic efficiency. Furthermore, the proposed framework supports adaptive decision-making under disruptive scenarios, such as traffic accidents and demand surges, which are critical for public transportation operation and decision support platforms. Extensive experiments in a microscopic traffic simulation demonstrate that C-MAPPO-TSC substantially outperforms fixed-time control, actuated control, and single-agent reinforcement learning methods. Specifically, it reduces average vehicle waiting time by 54.1%, average queue length by 51.6%, and network-wide travel time by 33.3% under peak-hour traffic conditions. These results confirm that C-MAPPO-TSC is an effective, robust, and scalable framework for cooperative traffic signal control and congestion mitigation in urban networks.
Zitieren
APA
Luo, P., & Ding, X. (2026). C-MAPPO-TSC: A Collaborative Multi-Agent Reinforcement Learning Framework for Urban Traffic Signal Control and Congestion Mitigation. Fundamental Scientific Reports in Multidisciplinary Areas, 2(02), 170-181. https://doi.org/10.66238/fsrma103
Vancouver
Luo P, Ding X. C-MAPPO-TSC: A Collaborative Multi-Agent Reinforcement Learning Framework for Urban Traffic Signal Control and Congestion Mitigation. Fundamental Scientific Reports in Multidisciplinary Areas. 2026;2(02):170-181. doi:10.66238/fsrma103
BibTeX
@article{Luo2026,
author = {Luo, Pinghao and Ding, Xiaotong},
title = {C-MAPPO-TSC: A Collaborative Multi-Agent Reinforcement Learning Framework for Urban Traffic Signal Control and Congestion Mitigation},
journal = {Fundamental Scientific Reports in Multidisciplinary Areas},
year = {2026},
volume = {2},
number = {02},
pages = {170--181},
doi = {10.66238/fsrma103}
}
EUVIMED