Study On The Bus Dwell Time Prediction Based On GPS Data

As an important part of the public transport system, the bus station is an important aspect of public transit service. The reseach on bus dwell time can make the managements of the public transport system master the regular pattern of bus, which has a great support for bus station settings optimization and bus dynamic scheduling. Meanwhile, as an important part of the bus running time, the prediction of bus dwell time can improve the accuracy of the prediction of bus arrival time and provide technical support for bus priority signal control, which will improve the public transport system operation efficiency and service level assuredly. Therefor, the research on bus dwell time has the important theory significance and practical demand.The domestic and foreign research on bus dwell time is mostly focused on reducing dwell time delays and optimizing the bus station settings. The research on the pattern of bus dwell time and the predicting of the bus dwell time is not researched largly so far. In this context, this study calculates the bus dwell time based on the data preprocessing with the analysis of GPS data.Then according to the research on the temporal-spatial pattern and line characteristic of bus dwell time we discuss, we build the predicting model of the bus dwell time based on time series analysis method with Exponential Smoothing.Research work mainly includes the following three aspects:First, this paper analyses the factors that affect the bus dwell time.The bus station is an important node of city bus lines.The dwell time at the bus station is influenced by many factors. According to the characteristics of the GPS data at bus station, the bus dwell time is divided into three parts as slow down into station time,parking service time and accelerating out of station time. Then this paper analysises the influencing factors of the bus dwell time from two perspectives as approaching-departing time and parking service time.Secondly, this paper studies on the features of the bus dwell time based on the bus dwell time datas. At the first place, it proves that the datas of bus dwell time follow the Erlang Distribution at different order at a macroscopic level. Then,it studies the difference and correlation of the datas of the bus dwell time from both space and time,which supports for solving the key problem of predicting the bus dwell time later.Finally, this paper analyzes the characteristics of bus dwell time of different bus lines in order to building the predicting model of the bus dwell time.Thirdly, this paper builds the predicting model of the bus dwell time based on time series analysis. According to the characteristics of the GPS data and the features of bus dwell time,this paper analyzes historical average of the bus dwell time from the four dimensions of the line, station, hours and dates.Then it builds the predicting model of the bus dwell time using time series analysis method with Exponential Smoothing. The results show that the model has good estimation performance.Based on a large amount of datas, this paper undergoes a comprehensive analysis of the temporal-spatial and bus line bus. Then the predicting model of the bus dwell time it builds can apply to most of the ordinary bus lines. Experiments show that the prediction can predict the bus dwell time with high accuracy, which can serve a useful purpose on further research on bus dwell time.