基于视频联网和多智能体的区域交通联动控制关键技术研究

发布时间：2018-04-26 08:26

本文选题：视频联网 + 多智能体　；参考：《北京交通大学》2017年博士论文

【摘要】：随着车流检测技术的不断发展,传统检测方式难以满足实际应用需求。近年来,将智能视频监控技术应用于交通信息获取与处理,解决交通拥堵问题,已成为智能交通系统中的一项关键技术。目前虽已建立起了较为完善的智能交通系统,但各子控制系统之间难以进行有效的数据融合和信息共享,只能进行简单的信息采集和处理,并没有通过视频联网共享和交通控制策略结合起来,致使系统之间的联动控制机制没有形成。本文主要以国家自然基金项目(F030209) “基于Agent和演化博弈的城市交通信号区域联动控制协调运行研究”为依托,开展对城市交通智能化管理和控制的关键技术进行研究。利用视频识别相关的技术手段,实现对交通流参数的实时提取,实时判别交通状态,结合agent技术,针对相邻交叉口交通流的关联性,通过相邻交叉口间的信息交互,对控制小区进行调整,并了解其协调任务的简易度和紧急度,及时采取联动控制措施。首先,根椐交通的实时性需求,构建了基于虚拟线圈的交通参数提取模型,描述了车辆检测和车辆跟踪的实现过程,分析了参数提取的重点要素,指出了车辆检测和跟踪是交通参数提取的重点。其次,根椐交通参数的时空变化规律分析,提出基于模糊认知图的交通状态快速识R%与跃迁转变模型,构建了交通参数关联关系图,指出了跃迁转变的内部和外部影响,分析了跃迁转变的过程。然后,根据基于系统动力学的拥挤传播模型对交通拥堵的产生及传播规律进行分析,应用节点收缩法确定关键节点,得到一个关键节点重要度的排序。根据交通拥挤扩散规律,以关键节点为控制小区中心,以最短路径阻抗作为约束条件,确定控制小区的范围。根据交通关联度和相似度对节点和关联路径进行划分,提出了基于节点收缩法的控制小区动态调整与优化模型。接着,根据多Agent协调联动控制不仅要考虑自身的运行效果,而且考虑自已排放车流对下游交叉口的影响,将每个路口看作一个Agent,建立基于遗传算法的信号优化模型;根据交通任务的简易度和紧急度,建立基于模糊理论的Agent协调控制选择模型;根据直接信任协调和间接信任协调确定协调团队,构建虚拟控制校区,提出了基于多Agent的协调联动控制模型。最后,根据对原有系统的更新和改进,并要求Agent不同功能之间的互联方式、信息传递和数据共享,以集成的思想,设计了一个基于多Agent的城市交通协调联动控制体系。通过对视频识别和多Agent的协调联动控制关键技术的研究,可以有效提高提取交通参数的速度和准确度,并增强交通信号控制能力和覆盖范围,实现城市交通实时、多点以及协同控制的目的,对缓解城市交通拥堵具有指导意义和参考价值。
[Abstract]:With the continuous development of vehicle flow detection technology, the traditional detection method is difficult to meet the needs of practical applications. In recent years, the application of intelligent video surveillance technology to traffic information acquisition and processing, to solve the traffic congestion problem, has become a key technology in the intelligent transportation system. At present, although the intelligent transportation system has been established, it is difficult to carry out effective data fusion and information sharing among the sub-control systems, so it can only carry out simple information collection and processing. There is no combination of video sharing and traffic control strategy, so the linkage control mechanism between the systems has not been formed. Based on the National Natural Fund project F030209) "study on coordinated Operation of Urban Traffic signal Regional coordinated Control based on Agent and Evolutionary Game", this paper studies the key technologies of intelligent management and control of urban traffic. By using the related technology of video recognition, the real-time extraction of traffic flow parameters and the real-time identification of traffic status are realized. Combined with agent technology, according to the correlation of traffic flow at adjacent intersections, the information exchange between adjacent intersections is achieved. Adjust the control area and understand the simplicity and urgency of the coordination task, and take the linkage control measures in time. Firstly, according to the real-time demand of traffic, a traffic parameter extraction model based on virtual coil is constructed, the realization process of vehicle detection and vehicle tracking is described, and the key elements of parameter extraction are analyzed. It is pointed out that vehicle detection and tracking are the key points of traffic parameter extraction. Secondly, based on the analysis of the temporal and spatial variation of traffic parameters, a model of fast recognition of traffic state based on fuzzy cognitive map is put forward, and the correlation diagram of traffic parameters is constructed, and the internal and external effects of transition transition are pointed out. The transition process is analyzed. Then, according to the congestion propagation model based on system dynamics, the generation and propagation rules of traffic congestion are analyzed, and the key nodes are determined by node contraction method, and a ranking of the importance of key nodes is obtained. According to the law of traffic congestion diffusion, the key nodes are taken as the control center and the shortest path impedance is taken as the constraint condition to determine the range of the control cell. According to the traffic correlation degree and the similarity degree, the dynamic adjustment and optimization model of the control cell based on the node contraction method is proposed. Then, according to the multi- coordinated linkage control, we should not only consider the effect of their own operation, but also consider the effect of their own traffic flow on the downstream intersection, and consider each intersection as an agent, and establish a signal optimization model based on genetic algorithm. According to the simplicity and urgency of traffic task, the Agent coordination control selection model based on fuzzy theory is established, the coordination team is determined according to direct trust coordination and indirect trust coordination, and the virtual control school district is constructed. A coordinated linkage control model based on multiple Agent is proposed. Finally, according to the update and improvement of the original system and the requirement of interconnecting mode, information transmission and data sharing among different functions of Agent, an urban traffic coordination and linkage control system based on multiple Agent is designed with the idea of integration. Through the research on the key technology of video recognition and coordinated linkage control of multiple Agent, the speed and accuracy of extracting traffic parameters can be improved effectively, and the control ability and coverage of traffic signal can be enhanced to realize real-time urban traffic. The purpose of multi-point and cooperative control is of guiding significance and reference value to alleviate urban traffic congestion.
【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：U495;TP391.41

【参考文献】