基于交通流预测的控制子区交通状态识别技术研究

发布时间：2018-03-01 20:27

本文关键词： 关键路口关联度子区划分组合模型预测集成分类器状态识别　出处：《浙江大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着科学技术的不断发展以及人民生活水平的不断提高,交通拥堵问题越来越普遍。交通拥堵不仅会降低人民的生活质量,同时也会影响经济发展并带来资源浪费、环境污染等各种问题,因此,缓解交通拥堵刻不容缓。智能交通系统是提高交通控制水平和管理效率的重要手段,而区域交通状态识别是实现智能交通系统的前提,因而对交通状态识别技术的研究具有重要的理论意义和实用价值。本文针对我国城市交通特点,利用自动控制理论、人工智能原理和交通工程技术对交通控制子区划分、短时交通流预测以及子区交通状态识别等技术进行了深入研究。具体内容如下： (1)分别提出了静态关键路口和动态关键路口的选取方法。静态关键路口主要根据主干道原则、路口饱和度原则、流量原则、时间规律原则以及交通部门的经验进行选取。动态关键路口主要通过路口关键度进行选取,为此研究了基于层次分析法的路口关键度计算方法。 (2)提出了基于改进遗传算法的交通控制子区划分方法。首先设计了一种基于模糊逻辑的关联度计算方法,然后基于关键路口和路口关联度建立了交通控制子区划分模型,并利用改进遗传算法进行模型求解,实现了交通控制子区的划分。 (3)提出了基于变权重组合模型的短时交通流组合预测方法。首先分析了交通流特征以及影响交通流的主要因素,然后根据交通流特征设计了基于卡尔曼滤波模型和神经网络模型的变权重组合预测模型,其中两个单一模型的权值通过上一步的预测误差进行实时调整,为了增加模型的平稳性,还引入了惯性因子。 (4)提出了基于FCM集成分类器的区域交通状态识别方法。首先将当前时刻交通流数据与交通流预测数据的融合数据作为交通状态识别输入数据,然后分别提出了基于模糊C均值和基于FCM集成分类器的交通状态识别方法,交通状态识别的输出为“低饱和”、“中饱和”和“准饱和”三种状态。
[Abstract]:With the development of science and technology and the improvement of people's living standard, traffic congestion is becoming more and more common. Traffic congestion will not only reduce people's quality of life, but also affect economic development and bring waste of resources. Therefore, it is urgent to alleviate traffic congestion. Intelligent Transportation system (its) is an important means to improve traffic control level and management efficiency, and the recognition of regional traffic condition is the prerequisite to realize its. Therefore, the research of traffic state recognition technology has important theoretical significance and practical value. According to the characteristics of urban traffic in China, this paper uses the theory of automatic control, the principle of artificial intelligence and traffic engineering technology to divide the sub-areas of traffic control. Short-time traffic flow prediction and sub-area traffic state recognition are studied in depth. The specific contents are as follows:. 1) the selection methods of static and dynamic key junctions are put forward respectively. The static key junctions are mainly based on the principle of main road, the principle of saturation and the principle of flow. The principle of time law and the experience of traffic department are selected. The key degree of intersection is chosen by dynamic key intersection. Therefore, the calculation method of intersection criticality based on analytic hierarchy process (AHP) is studied. (2) A traffic control sub-area partition method based on improved genetic algorithm is proposed. Firstly, a method of calculating the correlation degree based on fuzzy logic is designed, and then the traffic control sub-area partition model is established based on the correlation degree of key intersection and intersection. The improved genetic algorithm is used to solve the model and the traffic control sub-area is divided. In this paper, a short-term traffic flow combination forecasting method based on variable weight combination model is proposed. Firstly, the characteristics of traffic flow and the main factors affecting traffic flow are analyzed. Then, according to the traffic flow characteristics, the variable weight combined forecasting model based on Kalman filter model and neural network model is designed, in which the weights of two single models are adjusted in real time through the prediction error of the previous step. In order to increase the stability of the model, the inertia factor is also introduced. Firstly, the fusion data of current traffic flow data and traffic flow prediction data are taken as the input data of traffic state recognition. Then a traffic state recognition method based on fuzzy C-means and FCM integrated classifier is proposed. The output of traffic state recognition is "low saturation", "medium saturation" and "quasi-saturation".
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495

【参考文献】