基于云模型的城市快速路交通状态识别方法研究

发布时间：2018-08-25 15:10

【摘要】：伴随着交通需求的急剧增加,城市道路交通日趋饱和,严重制约了社会经济的发展。为了满足城市组团间的快速出行需求,各大中心城市相继开始规划建设高快速路网。作为城市交通系统的基本骨架,快速路承担着较大比重的机动车交通出行量。由于快速路良好的通达性和密集的匝道口设计,其交通运行态势呈现出逐年恶化的趋势,交通运输效率大幅下降,与其功能定位严重不相吻合,主要表现为:拥挤强度加剧、拥堵范围蔓延和拥堵时间持续增长。因此,提升城市快速路的交通运行效率成为城市交通管理的首要任务。准确提取实时、可靠的交通状态信息是智能交通管理的前提,而交通状态具有很强的随机性和非线性特征,造成快速路的交通状态识别十分复杂。云模型是一种很好的定性定量相互转换的工具,为快速路的交通状态识别提供了一种新的研究方法。论文基于改进云模型对快速路交通状态的识别方法进行研究,主要包括以下几个方面的内容:(1)系统回顾当前针对交通状态识别的研究进展,确定选用云模型进行快速路交通状态识别;对交通状态评价指标进行筛选,为避免流量的二流特性对识别结果产生干扰,选择速度和时间占有率作为评价指标。(2)研究数据预处理的方法,为识别模型作数据准备。对错误数据的识别和处理,设计基于时序权重的时间序列方法补齐丢失数据,通过算例分析表明:基于时序权重的数据补齐方法更能反映交通流参数的时变特性,补齐结果更加接近真实值。(3)建立基于云模型的快速路交通状态识别方法。采用K-means聚类分析对历史数据进行聚类,通过逆向云发生器算法得到初始模版云,并利用梯形云对模版进行改进,获取实际快速路交通状态的模版云;进一步通过信息熵函数对评价指标进行动态赋权,得到快速路的交通状态和交通拥堵指数。(4)通过具体案例对模型识别效果进行评价。基于成都市二环高架的微波数据,将本文所建模型和速度阈值法、V/C比阈值法的识别结果进行比较分析,验证基于云模型的交通状态识别模型的有效性。研究结论表明:基于云模型的交通状态识别方法能改善单一指标阈值划分方法的不足,解决了传统正态云模型在极端情况下识别精度不高的问题,更能反映真实的交通运行状态,具有较强的实用性和可移植性。
[Abstract]:With the rapid increase of traffic demand, urban road traffic becomes increasingly saturated, which seriously restricts the development of social economy. In order to meet the needs of rapid travel among urban groups, the major central cities have begun to plan and build high speed road network. As the basic framework of urban traffic system, expressway bears a large proportion of motor vehicle travel volume. Due to the good accessibility and dense ramp design of expressway, the traffic running situation is worsening year by year, and the efficiency of transportation is greatly decreased, which is not in accord with its function orientation seriously, and the main performance is: the intensity of congestion intensifies. Congestion spread and congestion time continues to grow. Therefore, improving the efficiency of urban expressway traffic becomes the primary task of urban traffic management. Accurate extraction of real-time and reliable traffic state information is the premise of intelligent traffic management, and traffic state has strong randomness and nonlinear characteristics, resulting in the expressway traffic state recognition is very complex. Cloud model is a good tool for qualitative and quantitative interconversion, which provides a new research method for expressway traffic state recognition. In this paper, based on the improved cloud model, the identification method of expressway traffic state is studied, which includes the following aspects: (1) the current research progress of traffic state recognition is reviewed. The cloud model is chosen to identify the traffic state of expressway, and the evaluation index of traffic state is screened to avoid the interference of the two-flow characteristic of the flow to the identification result. The speed and time share are selected as evaluation indexes. (2) the method of data preprocessing is studied to prepare the data for identifying the model. To identify and deal with the error data, the time series method based on the time series weight is designed to correct the lost data. Through the analysis of an example, it is shown that the time varying characteristics of the traffic flow parameters can be better reflected by the data alignment method based on the time series weight. The complement results are closer to the real value. (3) the method of expressway traffic state recognition based on cloud model is established. K-means clustering analysis is used to cluster the historical data, and the initial template cloud is obtained by the reverse cloud generator algorithm, and the trapezoid cloud is used to improve the template to obtain the template cloud of the actual expressway traffic state. Furthermore, the evaluation index is dynamically weighted by the information entropy function, and the traffic state and traffic congestion index of the expressway are obtained. (4) the effectiveness of the model identification is evaluated through specific cases. Based on the microwave data of Chengdu's second ring viaduct, the recognition results of the proposed model and the velocity threshold method / V / C ratio threshold method are compared and analyzed to verify the validity of the traffic state recognition model based on the cloud model. The results show that the method of traffic state recognition based on cloud model can solve the problem that the traditional normal cloud model has low recognition accuracy in extreme cases. It can reflect the real state of traffic and has strong practicability and portability.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491

【参考文献】