Midterm colloquium Nasir Al Aeedanee

Reinforcement Learning with Model Predictive Control for Coordinated Highway Traffic Control

Supervisors: dr. A. Dabiri, F. Airaldi

Abstract: The integration of Reinforcement Learning (RL) and Model Predictive Control (MPC) presents a approach to enhance highway traffic management by addressing challenges such as congestion, inefficiency, and variability in traffic dynamics. This work focuses on leveraging the adaptability of RL and the constraint-handling capabilities of MPC to optimize control strategies for ramp metering (RM) and variable speed limits (VSLs). By combining the predictive modeling strengths of MPC with RL's ability to learn from real-time data, this research aims to develop a hybrid framework that improves traffic flow, reduces congestion, and enhances the resilience of highway systems to unexpected disturbances. While this study emphasizes coordinated traffic control using a case study of highway networks, the methodology is adaptable to broader traffic management applications, providing a foundation for scalable, data-driven, and robust control solutions in complex traffic environments.