摘要
公交车辆到站时间预测是公交信息服务、公交动态调度的关键参数。基于实时和历史的公交车辆自动定位数据(AVL)需求分析,将公交车辆到站时间划分为站点停靠时间、区段全程运行时间和区段部分运行时间,分别采用点估计法、BP神经网络法和自适应指数平滑法对其进行动态预测。最后结合实验线路公交车辆的AVL运行数据,对预测模型进行了验证和评价分析。研究结果表明:本预测模型由于将历史数据规律和实时交通状况进行了有效融合,从而提高了公交到站时间预测的鲁棒性和预测精度。
Prediction of bus arrival time plays an important role in transit information service and dynamic scheduling.Based on needs analysis of automatic vehicle location(AVL) data,bus arrival time is divided into three components,including dwell time on stations dynamically predicted by point estimate algorithm,all travel time in intervals by back-propagation network algorithm and part travel time in intervals by self adaptive thrice exponential smoothing algorithm.In addition,these models are validated and evaluated based on the AVL data provided by the experimental bus line.It's found that the model improves predicting index performances of robustness and precision because of the effective integration of regular patterns from historical data and real traffic conditions from dynamical data.
出处
《重庆交通大学学报（自然科学版）》
CAS
北大核心
2012年第5期1014-1017,1041,共5页
Journal of Chongqing Jiaotong University(Natural Science)
基金
国家自然科学青年基金项目(51008231)
中国博士后科学基金特别资助项目(201104287)
上海市教育委员会重点学科建设项目(J51401)
上海高校青年教师培养资助计划(shgcjs022)
关键词
公交到站时间
实时预测
自动车辆定位数据
BP神经网络算法
自适应指数平滑法
bus arrival time
real-time prediction
automatic vehicle location data(AVL)
back-propagation network algorithm
self-adaptive thrice exponential smoothing algorithm
