To answer this question: How to make AI truly scalable and reliable for real-time traffic assignment? A research team from KTH Royal Institute of Technology, Monash University, Technical University of Munich, Southeast University, and the University of Electro-Communications has developed a new framework—MARL-OD-DA—that offers a promising answer. The approach redesigns learning agents at the origin–destination (OD) level and utilizes Dirichlet-based continuous actions to achieve stable and high-quality solutions under dynamic travel demand.
The team published their study in Communications in Transportation Research on November 21, 2025.
"Most multi-agent reinforcement learning (MARL) models assign one agent to each traveler, creating scalability and reliability problems when the demand grows," says Zhenliang Ma, Associate Professor at KTH Royal Institute of Technology. "Our framework redefines agents at the OD-pair level, reducing the number of agents by two orders of magnitude while still capturing essential routing behavior."
The method further introduces a Dirichlet-based continuous action space, enabling sparse and meaningful routing proportions while inherently satisfying simplex constraints. A reward function based on the relative gap, a key convergence metric in traffic assignment, improves learning robustness and stability.
To validate the framework, the researchers conducted extensive experiments on three benchmark networks: OW, SiouxFalls, and Anaheim. MARL-OD-DA consistently outperformed traditional optimization algorithms and existing MARL baselines in solution quality, robustness, and convergence speed. In the SiouxFalls network, trained agents achieved high-quality solutions within just 10 iterations, reducing the relative gap by 94.99% compared to conventional methods.
"These findings show that MARL-OD-DA is not only scalable but also deployable," says Leizhen Wang, one of the co-authors. "It can effectively support real-time route guidance, shared mobility services, and congestion management in future intelligent transportation systems."
The researchers believe that MARL-OD-DA could serve as a foundation for integrating high-fidelity simulations, dynamic traffic assignment tools, and real-world sensor data into next-generation AI-driven traffic management platforms.
About Communications in Transportation Research
Communications in Transportation Research was launched in 2021, with academic support provided by Tsinghua University and China Intelligent Transportation Systems Association. The Editors-in-Chief are Professor Xiaobo Qu, a member of the Academia Europaea from Tsinghua University and Professor Shuai'an Wang from Hong Kong Polytechnic University. The journal mainly publishes high-quality, original research and review articles that are of significant importance to emerging transportation systems, aiming to serve as an international platform for showcasing and exchanging innovative achievements in transportation and related fields, fostering academic exchange and development between China and the global community.
It has been indexed in SCIE, SSCI, Ei Compendex, Scopus, CSTPCD, CSCD, OAJ, DOAJ, TRID and other databases. It was selected as Q1 Top Journal in the Engineering and Technology category of the Chinese Academy of Sciences (CAS) Journal Ranking List. In 2022, it was selected as a High-Starting-Point new journal project of the "China Science and Technology Journal Excellence Action Plan". In 2024, it was selected as the Support the Development Project of "High-Level International Scientific and Technological Journals". The same year, it was also chosen as an English Journal Tier Project of the "China Science and Technology Journal Excellence Action Plan Phase Ⅱ". In 2024, it received the first impact factor (2023 IF) of 12.5, ranking Top1 (1/58, Q1) among all journals in "TRANSPORTATION" category. In 2025, its 2024 IF was announced as 14.5, maintaining the Top 1 position (1/61, Q1) in the same category. Tsinghua University Press will cover the open access fee for all published papers in 2025.
