运用不确定性方法估计高速公路基本路段交通状态

发布时间：2019-06-13 21:28

【摘要】：高速公路交通状态的准确估计是把握高速公路交通运行情况的关键。单一交通流参数只能间接、局部的反映实际的交通运行状况,依据多个不同交通流参数进行聚类分析是地点交通状态估计的典型方法,但聚类结果对样本数量非常敏感;另一方面,目前对路段交通状态的估计一般只考虑行程时间或行程车速,因数据采集手段的限制,估计的交通状态存在一定不确定性。针对这些问题展开研究,对改善高速公路交通状态估计系统应用效果具有实际意义。论文以高速公路基本路段为研究对象,运用不确定性方法来估计高速公路基本路段的地点和路段的交通状态。在地点交通状态估计中,以交通流参数样本点空间分布的不均衡性分析为突破口,重点解决样本数量不均衡性对交通状态聚类结果影响;在路段交通状态估计中,重点针对交通状态估计的不确定性问题,采用多源数据融合的方法加以解决。主要研究内容包括:①高速公路基本路段交通流参数特性的分析。首先,对地点交通流参数的时间相关性和样本点空间分布的不均衡性进行了分析;然后,对路段交通流参数估计交通状态时存在的不确定性进行了分析,为后面地点和路段交通状态估计模型的建立奠定了基础。②基于特征参数加权GEFCM算法的高速公路地点交通状态估计模型的建立。针对传统模糊聚类算法在交通状态估计时存在的不足,结合样本分布的不均衡性以及不同特征参数对于聚类影响权重的差异性,建立特征参数加权GEFCM算法的地点交通状态估计模型,并通过主成分分析法确定了不同特征参数在模型中的权重值,实验表明本文方法具有更好可靠性与适应性。③基于动态贝叶斯网络多源数据融合的高速公路路段交通状态估计模型的建立。针对采用路段相对密度和路段平均行程时间估计交通状态时存在的不确定性问题,引入动态贝叶斯网络,在研究选取相对密度、平均行程时间和交通状态为节点变量的基础上,确定了网络的拓扑结构,并最终研究建立了用于状态估计的动态贝叶斯网络模型,实验表明本文方法具有更好的可靠性。最后,对系统进行了设计与实现,并应用于渝武高速公路部分路段的地点和路段交通状态估计。结果表明:本文方法对地点和路段交通状态的估计结果相比较对比的方法均具有更高的拥挤判别率和更低的拥挤误判率。
[Abstract]:The accurate estimation of highway traffic condition is the key to grasp the traffic operation of expressway. The single traffic flow parameter can only indirectly reflect the actual traffic operation condition. Cluster analysis according to many different traffic flow parameters is a typical method of location traffic state estimation, but the clustering results are very sensitive to the number of samples. On the other hand, at present, the estimation of the traffic state of the road section generally only considers the travel time or the travel speed, because of the limitation of the data acquisition means, there is a certain uncertainty in the estimated traffic state. The research on these problems is of practical significance to improve the application effect of highway traffic state estimation system. In this paper, the basic section of expressway is taken as the research object, and the uncertainty method is used to estimate the location and traffic state of the basic section of expressway. In the traffic state estimation of the location, the uneven analysis of the spatial distribution of the sample points of the traffic flow parameters is taken as the breakthrough point, and the influence of the imbalance of the number of samples on the clustering results of the traffic state is solved. In the traffic state estimation of the road section, the uncertainty problem of the traffic state estimation is solved by the method of multi-source data fusion. The main research contents are as follows: 1. Analysis of traffic flow parameters in basic sections of expressway. Firstly, the temporal correlation of location traffic flow parameters and the imbalance of spatial distribution of sample points are analyzed. Then, the uncertainty of road section traffic flow parameters estimation traffic state is analyzed, which lays a foundation for the establishment of traffic state estimation model of rear location and road section. (2) the establishment of highway location traffic state estimation model based on characteristic parameter weighted GEFCM algorithm. In view of the shortcomings of traditional fuzzy clustering algorithm in traffic state estimation, combined with the imbalance of sample distribution and the difference of influence weight of different characteristic parameters on clustering, the location traffic state estimation model of feature parameter weighted GEFCM algorithm is established, and the weight values of different characteristic parameters in the model are determined by principal component analysis (PCA). The experimental results show that the proposed method has better reliability and adaptability. (3) the traffic state estimation model of expressway section based on dynamic Bayesian network multi-source data fusion is established. In order to solve the uncertainty problem of using relative density of road section and average travel time of road section to estimate traffic state, dynamic Bayesian network is introduced. On the basis of selecting relative density, average travel time and traffic state as node variables, the topological structure of the network is determined, and finally a dynamic Bayesian network model for state estimation is established. The experimental results show that the proposed method has better reliability. Finally, the system is designed and implemented and applied to the location and traffic state estimation of some sections of Yuwu Expressway. The results show that the proposed method has higher congestion discrimination rate and lower congestion misjudgment rate compared with the estimation results of location and road traffic state.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491

【相似文献】