Spatial analysis of public transport network from a complex network perspective and its impact on vehicular network connectivity

Abstract

A graph, comprising a set of nodes connected by edges, is one of the simplest yet remarkably useful mathematical structures in the analysis of real-world complex systems. By representing the components of a complex system as nodes, and their interconnectivity as edges, the inherent nature encoded in the topology of a complex system can be studied. In this thesis, we emphasize on the topological analysis of two types of public transport networks (PTNs), namely, the bus and metro transport networks, using the concepts of graph theory. The topological analysis is accomplished for three cities: Hong Kong, London, and Bengaluru. Initially, we study the bus transport network topologies of the three cities as non-interacting mono-layers. Then, we study the bus and metro network topologies as multi-layer networks by considering and ignoring the interactions between the two layers. The topological analysis unveils intriguing network behaviors under both mono-and multi-layer analyses. Lastly, we integrate the concepts from graph theory and vehicular networks to demonstrate the dependency of vehicular network connectivity on the underlying transport network topology, from a macroscopic perspective. This thesis aims to analyze the topological properties of public transport networks and their impact on the vehicular network connectivity. First, graph theory concepts are employed to represent the PTN topology as a graph. The consideration of spatial embedding of PTNs results in a new network element called supernode, which yields in a novel approach in modeling the PTN topology called supernode graph structure representation. A static demand estimation approach is proposed to weigh a node's significance based on its real-world usage alongside its topological centrality. The crux of the demand estimation approach is that the accessibility factor of a node is greatly influenced by the presence of points-of-interests and the number of people accessing it. A zone classification approach is proposed which together with the static demand estimation approach aids in better identifying a node's significance. For the part of mono-layer analysis, a few topological metrics and structural behaviors of the three bus transport networks are studied in both the conventional and supernode representations, which reveals interesting topological information from both micro-and macroscopic points of view. At every stage of the analysis, we observe that the supernode representation offers better understanding of the inherent network behaviors as compared to the conventional representation. For the part of multi-layer analysis, initially, the bus and metro transport network topologies are considered as non-interacting mono-layers. Later, the approach of spatial amalgamation is used to integrate the two transport network topologies so that they can be studied as an interdependent multi-layer network. The static demand estimation approach is used to rank a node's significance with and without considering the interdependency of the multiple layers. The results demonstrate the fact that ignoring the interdependency between multiple transport modes will dramatically influence the understanding of true network behavior. Finally, we aim to integrate the two different fields of study: network theory and VANET (Vehicular Ad-hoc Networks), to study the influence of underlying transport network topology on the vehicular network connectivity from a macroscopic standpoint. Contact duration, a prime metric in VANET is used to understand the nature of V2I (Vehicle-to-Infrastructure) connectivity, which is primarily determined by the two topological metrics: weighted in-degree and node weight. Lastly, we assess the topological robustness of the vehicular network connectivity to determine the network's ability to sustain malfunctions while continuing to offer the best possible service.