Modern transportation systems—like urban traffic—face growing challenges of congestion, delay, and emissions. Managing demand-capacity balance at a network level, also referred as network-wide congestion management, is vital to improving operational efficiency, reducing delays, and lowering environmental impact. 

Despite advancements in real-time and adaptive planning and control algorithms, network-wide congestion management remains suboptimal due to heterogeneous uncertainty in traffic state estimation, dynamic of real-world environment (e.g., non-nominal events such as incidents, accidents, road closures, etc.), issues of scalability and coordination, and partially cooperative traffic controllers.

This research seeks to develop a novel context-aware control policies that adapt to fluctuating congestion estimation accuracy, ensuring the network traffic smoothness and efficiency by leveraging advanced adaptive algorithms to train policies that rapidly adapt to changing uncertainty profiles and hierarchical multi-agent reinforcement learning to reduce the multi-agent coordination complexity. It is expected to enable scalable, responsive, and robust control across complex urban traffic networks for reducing of traffic congestion, additional travel time and emissions.

The research supports Smart City objectives by enabling scalable, data-driven traffic optimization to reduce congestion and enhance mobility. In doing so, it builds public trust in AI-driven systems, contributes to environmental sustainability through emission reduction, and improves overall travel efficiency.