Modelling and Simulation of Urban Road Network Evolution Using Generative Network Models

Empirical and modelling Road Network Evolution (RNE) research from a network science perspective has increased in recent years. Empirical RNE research has quantified real-world urban road network characteristics and dynamics. Modelling RNE research has proposed Generative Network Models (GNMs) to reproduce statistically empirical urban road network characteristics. This thesis proposes a novel framework to address the evolution of urban road networks through modelling and simulation of Node Addition and Link Connection, respectively. First, this thesis generalises the Link Connection generative mechanism of urban road networks, as a process of examining the proximity relationship between a new spatial location and the urban road network, using β-skeletons proximity relationships with β∈[1.0,2.0]. Proximity relationships come from a family of proximity graphs, which determine node connections by various geometric closeness definitions. Second, the proposed GNM of urban road network evolution and the generalised Link Connection are shown to be capable of giving rise to both static and dynamic network structures, raising in correspondence to empirical RNE findings. The simulation identifies originally parallels between the simulated network dynamics and empirical RNE characteristics, demonstrating the proposed model’s capacity in modelling the dynamic RNE in addition to network generation. By controlling the β parameter, the proposed GNM is shown to be capable of modelling a broader range of plausible urban road network structures than previous studies in this field. Third, this thesis proposes an original hybrid model of population and urban road network coevolution, which models population and road network dynamics on two inter-dependent layers and integrates GNM and RNE into the urban system through Node Addition. Various spatial decision combinations are explored, instead of assuming fixed population and road network spatial preferences. Fourth, the proposed coevolution model is shown to be capable of giving rise to diverse road network and population spatial structures, from centralised to decentralised on the global scale, clustered to dispersed on the local scale. The simulation suggests that related push and pull forces across urban system layers drive the coevolution, leading to a spatial structure spectrum, rather than fixed clear-cut types (Marshall, 2004; Huynh et al., 2017; Moosavi, 2017). The proposed model also simulates the emergence of population-driven dispersed spatial structure and road network-driven linear spatial structure. The simulation finds the variation of network spatial structure is one potential cause of differences in network characteristics, demonstrating a necessity to consider the spatial structure in urban road network structure analysis.