A three-layer hierarchical model-based approach for network-wide traffic signal control

This paper introduces a three-layer hierarchical model-based approach for network-wide traffic signal control, which consists of a region coordination layer, an intersection coordination layer and a local signal control layer. The hierarchical control problem is formulated by two model predictive control (MPC) schemes. The upper-level control deals with the region coordination problem, which considers the cycle length optimization for different regions based on the macroscopic fundamental diagram. Under the guidance of the upper level MPC, a distributed MPC is developed to address both the green split optimization at each local intersection and the intersection coordination. For computational efficiency, the network is decomposed into separated intersections while the Lagrangian algorithm is developed for intersection coordination. Experimental results on the Nguyen-Dupuis network prove that our proposed three-layer hierarchical control approach has great potential in reducing the computational complexity while maintaining the overall performance (such as total travel time) of the traffic network.