Exploring PageRank Algorithm and Voronoi Diagrams for Dynamic Network Partitions Facilitating Feedback Linearization-Based Control

This paper explores a novel approach to dividing a traffic region (network) into sub-regions for efficient traffic control among the areas. The macroscopic flow diagram (MFD) in each of these sub-regions, referred to as sub-MFD, can then be used to determine the macro-state of that sub-region and subsequently design controllers. The region division is based on the theory of complex networks. We exploit the inherent network characteristics through the PageRank centrality algorithm to identify the most significant nodes in the traffic network. We use these significant nodes as the seeds for a Voronoi diagram-based partitioning mechanism of the network. A feedback linearization-based controller is then presented, which controls the traffic flow between the sub-regions. A case study is performed for the Manhattan area in New York City to demonstrate the network partitioning approach; the control approach is demonstrated through a toy example containing two sub-regions.