Study On Causes And Forecast Of Bus Bunching

Public transportation is a vital means to ease congestion and to improve the sustainable development of urban traffic. Urban bus transit is the main body of public transport, and the choice of means of transport mainly depends on the level of public transportation service. At the same time, the level of public transportation, or to be more specific, the bus service is severely influenced by bus bunching problem. As a result, the study of this thesis is to effectively forecast bus bunching problem based on real-time data from BDS(Bus Dispatch System). The study has both theoretical significance and practical value.The study focuses on three questions:What is bus bunching? Why bus bunching occurs? How to forecast and avoid bus bunching effectively? Firstly, a criterion of bus bunching is offered depending on the real-time data by adopting the concept of covariance of headway. Then, based on the real-time Bus Dispatching data of the 38 bus line in Zigong City, the study offers six kinds of typical bus bunching process and their logic relation illustration. Thirdly, a mathematical model is built considering the bus station node, vehicles, et al. This model describes four theoretical factors of bus bunching. Fourthly, through the analysis of the real-time data, various statistical analysis methods such as correlation analysis, variance analysis, and regression analysis are used to find the quantitative relationship among the data, while the logic analysis is used to determine the actual cause on car problems influencing factors. Moreover, regression analysis is used to order the importance of given 7 influence factors. Fifthly, after analyzing the difference among prediction models, RBF (Radial Basis Function) neural network is selected through analytic hierarchy process. By using RBF neural network, a bus bunching forecast model is built to provide data support for dispatching system. Later on a case study is carried out based on the real-time data of Zigong City. The feasibility of the theoretical model is verified Lastly, based on the analysis of the process of bus dispatching system and both historical data and real-time BDS data, by use of RBF neural network prediction results, the reasonable ideal interval travel time (or speed), stop time and headway is calculated. The real-time calculated guidance scheduling command is fed back to bus drivers to adjust the operating state as well as to the system as historical data.