Junction-based Model Predictive Control for urban traffic light control

This article studies Junction-based Model Predictive Control (JMPC) in urban traffic light control. The entire traffic system is decomposed into junction-based subsystems with their own MPC controllers based on a frequently and regularly initialized difference equation (FRIDE). These MPC controllers consider the control effects on adjacent subsystems and inputs effects from adjacent subsystems, which work cooperatively towards system-wide control objectives. The simulation results show that JMPC works much better than some bench-marking strategies. Besides, JMPC is scalable to larger network because of its distributed nature and makes it possible to deal with the complex urban networks.