Bus bunching occurs when two or more buses travel head to tail. It is an annoying problem in public transportation because it increases passengers' average waiting time and traveling time, wastes bus capacity, reduces the frequency of bus service and increases the pressure on bus drivers. So eliminating bus bunching is important in public transportation. Eliminating bus bunching is highly challenging due to the complexity and variability of the bus dynamics. Bus bunching results from a positive feedback mechanism of headway evolution, which is a flaw born with the bus system. In this thesis, we quantify the intensity of the tendency to bus bunching and propose a headway control modeling framework to reverse tendency. Our framework subsumes many headway control schemes to coordinate buses and so enables batch analysis. Given different headway information, our framework produces different control schemes under which headways self-equalize. The stability of the bus system under control is characterized by a single measure and it can be optimized. Besides, the bus system under control is robust against traffic conditions and the level of ridership. The framework is based on a snapshot model capturing the bus dynamics including the tendency to bunch by taking traffic conditions and the level of ridership into account. It is linear and time-invariant, which makes the bus dynamics tractable. This model considers a single control point and constant bus velocity in a deterministic manner, but it can be extended to handle many control points, inhomogeneous velocity along the route, and randomness. Using our framework, we further study two simple control schemes---Threshold control and ``Prefol". Threshold control drives headways to self-equalize the fastest but the corresponding bus system needs large slack time for robustness. "Prefol" needs small slack time but headways self-equalize slower. We hybridize them and find the hybrid control scheme balances robustness and fast headway equalization. We also show that it outperforms several state-of-the-art control schemes in tests on a simulated bus route in Chicago.

The context of this study is the increasing need for public transport as issues over high private vehicle usage are becoming increasingly obvious. Public transport services need to compete with private transport to improve patronage, and issues with reliability need to be addressed. Bus bunching affects reliability through disruptions to the scheduled headways. The purpose of this study was to collect and analyse data to compare how travel time and dwell time vary, to explore the variation of key variables, and to better understand the sources of these variations. The Orbiter bus service in Christchurch was used as a case study, as it is particularly vulnerable to bus bunching. The dwell time was found to be more variable than travel time. It appeared the Canterbury earthquake had significantly reduced the average speeds for the Orbiter service. In 1964, Newell and Potts described a basic bus bunching theory, which was used as the basis for an Excel bus bunching model. This model allows input variables to vary stochastically. Random values were generated from four specified distributions derived from manually collected data, allowing variance across all bus platforms and buses. However the complexity resulted in stability and difficulty in achieving convergence, so the model was run in single Monte Carlo simulations. The outputs were realistic and showed a higher degree of bunching behaviour than previous models. The model demonstrated bunching phenomena that had not been observed in previous models, including spontaneously un-pairing, overtaking of buses delayed at platforms, and odd-numbered bunches of three buses. Furthermore, the study identified areas of further research for data collection and model development.

京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第22653号 / 工博第4737号 / 新制||工||1740(附属図書館) / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 山田 忠史, 教授 藤井 聡, 准教授 SCHMOECKER Jan-Dirk / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM