基于信息融合的城市道路行程时间预测研究

发布时间：2018-03-05 07:11

本文选题：城市道路　切入点：行程时间预测　出处：《西南交通大学》2016年硕士论文　论文类型：学位论文

【摘要】：随着城市中汽车保有量的快速增加和城市居民出行的频繁发生,城市交通在其经济发展中发挥的作用越来越明显,与之对应的是环境污染加剧,交通拥堵、交通事故频发。智能交通系统(ITS)被公认为处理交通问题的最有效方法,行程时间研究就是其中的关键问题,要想对城市交通进行有效的控制和诱导,行程时间是必不可少的因素,同时为了降低滞后性对交通管理和诱导的影响,行程时间预测研究一直被广大学者不断探索。研究城市道路行程时间预测的相关理论和方法,可以有效的缓解城市交通拥挤,加快城市现代化建设,对城市社会经济的发展具有不可替代的意义。为了开发一种简单实用的道路行程时间预测方法,论文首先系统的分析、总结了交通数据获取技术的相关理论,然后以包含交叉口的城市道路为研究对象,提出了行程时问估计模型、预测模型和预测时间融合模型。论文基本思路是：运用改进型的HCM2010建立了道路行程时间估计模型,并通过历史趋势模型和卡尔曼滤波分别建立了城市道路行程时间预测模型,由于各预测模型本身都具有其局限性,再加上单源交通数据会因为检测器本身精度原因影响预测值,论文最后根据神经网络模型构建了行程时间预测值的融合模型；其中,三个道路行程时间估计和预测模型的输出量——行程时间值都作为融合模型的输入变量,最终输出一个融合后的行程时间预测值。论文最后根据成都市某路段的道路实际参数和交通流数据,运用交通仿真软件VISSIM4.3对该路段进行仿真,得到相关交通数据,带入神经网络构建的融合模型中,得出城市道路行程时间预测值,然后与仿真行程时间进行比较。结果表明,融合模型得出的道路行程时间预测值误差绝大部分都在10%以内,比单预测模型精度高,融合模型有效性较好。
[Abstract]:With the rapid increase of vehicle ownership and frequent travel of urban residents, urban traffic plays a more and more important role in its economic development, corresponding to the intensification of environmental pollution and traffic congestion. Intelligent Transportation system (its) is recognized as the most effective way to deal with traffic problems. The study of travel time is one of the key problems. In order to effectively control and guide urban traffic, Travel time is an essential factor, and in order to reduce the impact of lag on traffic management and guidance, the study of travel time prediction has been continuously explored by the majority of scholars. It can effectively alleviate urban traffic congestion, speed up the construction of urban modernization, and has irreplaceable significance for the development of urban social economy. In order to develop a simple and practical method of road travel time prediction, this paper first systematically analyzes, This paper summarizes the relevant theories of traffic data acquisition technology, and then takes the urban roads with intersections as the research object, and puts forward the travel time estimation model. The basic idea of this paper is to use the improved HCM2010 to establish the road travel time estimation model, and to establish the urban road travel time prediction model by using the historical trend model and Kalman filter, respectively. Because each prediction model has its own limitations, and the single source traffic data will affect the prediction value because of the accuracy of the detector itself, the fusion model of the travel time prediction value is constructed according to the neural network model. The output of the three road travel time estimation and prediction models-stroke time values are all taken as the input variables of the fusion model. Finally, a fused travel time prediction value is outputted. Finally, according to the actual road parameters and traffic flow data of a section of Chengdu, the traffic simulation software VISSIM4.3 is used to simulate the road section and obtain the relevant traffic data. In the fusion model constructed by neural network, the prediction value of urban road travel time is obtained, and then compared with the simulation travel time. The results show that the error of road travel time prediction obtained by the fusion model is mostly within 10%. The accuracy of the fusion model is higher than that of the single prediction model, and the validity of the fusion model is better.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U491

【相似文献】