Three steps towards practical application of public transport route optimisation in urban areas

The research on computer-based optimisation of public transport networks has seen huge advances in recent years. However, there is almost no application of its results in real-world planning processes. A review of the literature identifies several possible reasons for this: a) many developed algorithms do not consider important details for urban applications, b) methods to generate the required input datasets struggle when applied to irregular urban networks, or c) the absence of an interface to use the developed algorithms with transport modelling software. The three publications presented in this thesis address these three issues. The first publication introduces a method to scale down an available street network to a level where optimisation methods can be applied while preserving its characteristics. All travel times, demand data, and information regarding the permitted route endpoints are derived from openly available data. The methodology is applied to the urban area of Nottingham, UK, to generate new benchmark datasets for bus route optimisation. An optimisation procedure centred on a genetic algorithm and adapted for the use of restricted route endpoints is applied to the generated dataset. In the second publication, these methodologies are extended to use a more realistic zone-based representation of passenger journeys. Zone-based trip representations are rarely used in academic studies as the conventional node-based approach is considered simpler to implement and many input datasets are publicly available. The publication presents a new hybrid-approach to calculate zone-based journey times using established node-based concepts, and introduces a first publicly available, zone-based input dataset. The third publication introduces an interface between route optimisation algorithms and the professional transport modelling software PTV Visum. The interface manages the differences in data requirements between the two modelling approaches, allowing users to directly optimise the public transport network in a given Visum network model. It is successfully combined with Selection Hyper Heuristics to optimise passenger travel time and operator costs in network models based on real-world planning processes. An additional optimisation experiment exemplary reduces the number of private vehicles on a selected street, demonstrating the capabilities of the interface