【作者】 王茁；

【导师】 程绍武；

【作者基本信息】 哈尔滨工业大学 ， 交通信息工程及控制， 2012， 硕士

【摘要】 对选择公共交通方式的出行者来说，公交车到站时间可称为居民出行最为关心的公共交通信息，对公交到站时间预测的研究也受到了越来越多的学者的关注。本文立足于分析影响公交车到站时间的影响因素分析，设计到站时间预测模型，提高公交车辆到站时间预测工作的精度和可靠性，不仅能够提高公共交通对居民出行的分担，而且对城市的公共交通服务体系的向前发展也在一定程度上起了推动的作用。论文首先从公交车的路段行驶时间、停靠站延误时间和交通状态三个方面，分析了影响公交车辆到站时间的影响因素以及各影响因素的量化方法。其次，通过划分公交车辆的运行状态，确定公交车辆运行状态的转移函数和状态转移条件的识别方法，建立公交车辆运行状态的有限状态自动机模型。将公交车辆预测状态细分为三个时段，对哈尔滨市63路公交车的车载GPS终端获取的公交车辆到站时间的历史数据进行预处理，分别采用Kalman滤波、BP人工神经网络以及ARIMA时间序列，对高峰、平峰和低峰时段63路公交车从建工新区出发到哈尔滨铁路局站的行程时间进行预测，根据三种模型的预测误差，确定每个时段最适合的到站时间预测模型，最终，得到基于有限状态自动机的公交车到站时间预测模型。最后，利用哈尔滨市63路公交车的车载GPS终端获取的公交车辆到站时间的历史数据，对比基于有限状态自动机的预测模型、卡尔曼滤波预测模型和BP神经网络预测模型的绝对平均百分误差（MAPE），结果表明，基于有限状态自动机的公交车辆到站时间预测模型相比Kalman率波预测模型和BP神经网络预测模型的MAPE值分别提高了44.42%和27.89%。更多还原

【Abstract】 To select the public of the means of transportation for travelers, the bus arrival timecan be concluded as the trip of the concerns of the public traffic informations, this paperbased on the analysis on the influence factors of bus arrival time analysis, and designarrival station time prediction model, in order to improve the public transport vehiclearrival station time prediction accuracy and also play a role in making reliability ofwork to the city public traffic system service development. abstractFirstly the paper analyzes the impact of affecting factors of the public transportvehicles arrival time influence factors from the bus stop time, delay time and trafficcondition three aspects, and their quantitative methods.Second, by the division of the bus vehicles operation states, the paper determinedthe running state of public transport vehicle transfer function and status transfercondition of identification methods, and then establishes the public transport vehicle inthe running states of finite state automata model.Public transport vehicles will be predicted for three time-segments of HarbinNO.63buses car GPS terminal bus arrival vehicles for the pretreated historical datas ofthe running time. Kalman filter, BP artificial neural network and ARIMA time serieswere used to forecast travel time of the peak, flat peak and low time NO.63buses fromJiangong District to Railway Station of Harbin. According to three model predictionerrors, to make sure most suitable mode. Based on the finite state automata get the busstation time prediction model.Finally, using the Harbin NO.63buses car GPS terminal bus station vehicles forthe historical data, contrast the average percentage of absolute error (MAPE) betweenbased on finite state automata prediction model and Kalman, BP model seperately, andthe results show that our model compared Kalman model and the BP model MAPE ofvalue increased by47.08%and31.63%respectively.更多还原