基于数据的交通拥堵评价与预测方法

发布时间：2018-03-10 21:08

本文选题：城市道路　切入点：多源数据融合　出处：《浙江工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着城市机动车保有量的迅猛增加,国内大部分城市尤其是特大城市的交通拥堵状况日益严峻。城市道路的拥堵情况严重影响着居民的日常工作和生活。基础道路建设受到诸多条件限制发展缓慢,智慧交通作为一种治理交通拥堵的新方法,已成为交通管理部门的工作重点。城市道路上安装的各种传感器每天能够采集大量交通数据。如何利用这些数据制定有针对性的交通策略从而缓解城市交通拥堵,成为了研究者们关注的重点。交通拥堵评价不仅是道路交通服务水平的重要依据,也是交通管理与控制的前提。准确地评价交通拥堵状态,对道路拥堵预测、交通诱导以及最佳路径规划均有非常重要的意义。但由于实际道路交通数据获取难度较大,信息共享程度较低,大多数研究工作还停留在理论建模与仿真阶段。因此对实际道路的交通数据分析还有较大的提升空间。本文基于实际道路的大量交通数据(卡口数据、微波数据、GPS数据)对交通拥堵状态的评价与预测方法进行了创新性研究,主要取得了以下三方面的研究成果：(1)对道路上常见的多种传感器采集到的数据进行详细分析,设计了一种车牌Hash算法用于去除冗余数据。针对各种传感器的优缺点,同时结合实际道路上采集的数据质量,提出了一种多源交通数据融合的方法,有效的修正了原始数据中的奇异数据。(2)对实际交通数据进行拥堵状态评价。首先使用数据挖掘中常用的K-means聚类方法,依据聚类结果得出交通拥堵状态评价方法。拥堵状态是交通状态分析的重点,但由于拥堵状态在交通状态中所占比重较低,K-means聚类方法不能有效的将拥堵状态划分出来,针对这一问题提出了一种基于密度的交通拥堵评价方法,该方法可以较好的划分出拥堵状态。(3)对短时交通拥堵状态进行预测,针对一阶马尔可夫模型预测交通拥堵状态,存在预测拥堵状态滞后的问题,提出了一种高阶马尔可夫模型的短时交通拥堵状态预测方法。该方法对交通拥堵状态的预测准确度有了一定提升达到92.7%,并且有效的消除了预测滞后问题。
[Abstract]:With the rapid increase in the number of motor vehicles in cities, The traffic jams in most cities in China, especially in mega-cities, are becoming increasingly serious. The congestion of urban roads seriously affects the daily work and daily life of residents, and the construction of basic roads is restricted by many conditions and develops slowly. Intelligent traffic as a new way to deal with traffic congestion, The sensors installed on urban roads can collect a large amount of traffic data every day. How to use these data to develop targeted traffic strategies to ease urban traffic congestion, Traffic congestion evaluation is not only the important basis of road traffic service level, but also the premise of traffic management and control. Traffic guidance and optimal path planning are of great significance. However, because of the difficulty of obtaining the actual road traffic data, the degree of information sharing is low. Most of the research work is still in the stage of theoretical modeling and simulation. Therefore, there is still much room for improving the traffic data analysis of the actual road. This paper is based on a large number of traffic data (bayonet data) of the actual road. Microwave data and GPS data) has carried on the innovative research to the traffic jam condition appraisal and the forecast method, has obtained the following three research achievements mainly: 1) to carry on the detailed analysis to the road common many kinds of sensors to collect the data, A license plate Hash algorithm is designed to remove redundant data. According to the advantages and disadvantages of various sensors and the quality of the data collected on the actual road, a multi-source traffic data fusion method is proposed. It effectively corrects the singular data in the original data and evaluates the traffic congestion. Firstly, K-means clustering method is used in data mining. According to the result of clustering, the evaluation method of traffic congestion state is obtained. Congestion state is the key point of traffic state analysis, but the K-means clustering method can not effectively divide congestion state because of the low proportion of congestion state in traffic state. In order to solve this problem, a density-based traffic congestion evaluation method is proposed, which can be used to predict the short-term traffic congestion, and the first-order Markov model can be used to predict the traffic congestion. There is a problem of predicting the lag of congestion, This paper presents a high order Markov model for short time traffic congestion prediction, which improves the accuracy of traffic congestion prediction to 92.7%, and effectively eliminates the problem of prediction lag.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495

【参考文献】