Examination and Analysis of Complex Public-Transport Networks

Abstract

The main function of public transport (PT) is to serve society. As such PT should provide users with a reliable, high-frequency, accessible, comfortable and well-coordinated service. Nowadays, due to factors such as the growth of modern cities leading to an increased level of vehicular traffic and congestion, PT service has been continuously experiencing problems harming its image and attractiveness: non-regular vehicle arrivals at PT stops and missed buses leading to increased waiting times. On the other hand, PT operators seek to make the service efficient and reduce operating costs through measures such as limited coverage comprising a small number of direct routes with a reduced fleet size and low frequency of service. Such measures directly influence passengers who need to perform more transfers between routes. The inconvenience of making transfers accompanied by extended waiting times coupled with the observed delays of arriving vehicles at transfer stations leads to missed transfers and increased waiting times, which impact the quality of PT service and makes PT less attractive.This doctoral research project is aimed at finding a solution to the above problem through providing PT users with a reliable, punctual, fast and convenient service which makes trips for current and prospective users within the network of PT routes more convenient and shorter in time – an intent implemented through designing a public-transport network (PTN) and performing multi-agent simulations of the passengers’ transportation process along the routes within the whole network, on a case study. The above was achieved in a sequence of steps including:(i) Development and application of a method combining computer programming, statistical data and large-scale network analyses, allowing examination of the structure of public transport networks(PTNs) and analysis of their topological properties. The method was applied on two real-life case studies comprising the public transport bus network in Auckland (New Zealand) as a first case study presented in Chapter 2 and as a second case study of the subway PTNs at Washington DC (USA) and Oslo (Norway) presented in Chapter 3. It was found in Chapter 2 that the examined bus-route network of Auckland is not scale free nor does it exhibit all the features of “small-world” networks. Instead, it can be considered as a mixture of exponential and scale-free networks, which means that the evolution of the bus route network in Auckland is a consequence of random rather than preferential attachment of newly opened stops. The network analysis performed in exploring the topological properties of the Washington DC and the Oslo metro networks documented in Chapter 3 showed that when represented in an L-space network topology, the examined networks do not exhibit small-world properties, and hence, they are not small-world networks. The examination of the Washington DC metro network and its analyses also showed that the network is neither a scale-free nor random network; this is based on the consideration of the network’s node degree distribution, the number of the metro lines servicing each station and representing the network as a bus station network. In contrast to the Oslo metro network, the metro network in Washington appears to be a complex network. The analysis considering the networks’ global efficiency, performed by using network science concepts and findings, showed that both the metro networks examined have low valued global network efficiencies and therefore appear not to be fault tolerant. (ii) Modelling the interaction between passengers, buses and cars along a route of PTN (on a case study in Auckland) as a small-scale multi-agent system (MAS) which facilitated establishing the influence of PT demand on passenger waiting time at bus stops when using vehicles in scenarios providing different passenger capacity and frequency of service. The simulation output results in Chapter 4 demonstrate that when PT demand is less than bus capacity, that capacity does not affect the average waiting time at bus stops. In cases where PT demand is high, high bus capacity resulting in an increased number of passengers boarding the bus, indirectly causes increased passenger waiting times, unless the frequency of service is increased. These results could serve PT operators well in the tradeoff situation when choosing between increased bus frequencies and larger size bus capacities, especially when the PT demand at bus stops is high. (iii) Developing a simulation framework and a model providing the options to model a high frequency,metro-like, autonomous PT service. The results obtained from the implementation of the developed simulation model, which was calibrated and validated on a numerical example under different simulation scenarios depicted in Chapter 5, showed that the model satisfactorily reproduces the parameters of the modelled system: (a) Among the scenarios, scenario 4 simulating a high frequency, metro-like PT service, due to small vehicle headways, provides passengers with the least average waiting time at bus stops; (b) Due to the small capacity of the high frequency vehicles used, which means less time for boarding/alighting (in contrast to the large-vehicle scenario), scenario 4 offers the lowest estimated average dwell time accumulated along the whole PT line; (c) The high frequency public-transit service provided in scenario 4 is the one that provides the least number of unserved passengers along the route of the modelled PT line; (d) A small decrease in the frequency of the vehicles (less vehicles) when a high frequency, metro-like service is provided, leads to an improved utilisation of vehicle capacity (operator’s perspective) at the cost of an acceptable increase of the average passenger waiting time at bus stops (user’s perspective) and insignificant increase in the number of unserved passengers.(iv) An application on a case study of the agent-based modelling concept in designing a MAS with interacting agents, such as PT users, self-driving vehicles and network sections, and performing multiagent simulations of the passengers’ transportation process along the routes of a newly-designed public transport network having a specific topology and features (considered from network science’s perspective), aimed to examine, analyse and improve the modelled system. The results of the simulations in Chapter 6 reveal the potential of the proposed monorail public transport network with driverless operating vehicles in achieving an attractive, reliable, punctual and fast public transport service.