动态数据驱动的交通仿真的研究与实现

发布时间：2018-06-23 12:41

本文选题：微观交通仿真 + 动态数据驱动　；参考：《南京航空航天大学》2014年硕士论文

【摘要】：随着智能交通系统(ITS)日益发展，微观交通仿真技术作为描述复杂交通行为的有效工具，在解决交通问题方面发挥越来越大的作用。围绕交通仿真模型建立和系统开发的研究已有许多成果，但现有交通仿真通常基于历史数据，忽略系统运行过程中各种突发事件，在实时动态条件下的仿真结果准确性较低。各种高新技术的发展使真实交通流数据较易获得，从而为基于动态实测数据的交通仿真提供了可能。动态数据驱动应用系统(DDDAS)的特点是将仿真和真实数据有效结合，使仿真能够动态接受实测数据的反馈，从而使仿真收敛更快、结果更可信。目前，DDDAS在危机管理、工程科学与灾难预报等具有充足实测数据的领域得到广泛的应用。基于上述背景，，本文旨在将DDDAS范式应用于微观仿真系统Movsim，提出动态数据驱动的交通仿真方法，该方法将车辆运行实测数据反馈到交通状态预测中，使预测结果更准确可靠。首先，对DDDAS范式和Movsim逻辑流程、车辆模型和路网模型进行分析，在此基础上提出动态数据驱动的交通仿真框架，剖析框架运行机制，探讨并行处理、数据同化等关键技术。其次，构建基于粒子滤波的交通仿真模型。粒子滤波算法给非线性、非高斯系统的状态估计提供了严谨解决方法，因此，引入粒子滤波算法实现框架的数据同化部分。本文设计了随机平移和分段车辆密度两种噪声模型、基于JSON中间件的数据映射模型、基于滑动窗口和传感器智能选择两种权重计算模型以及层次重采样模型，并给出算法图形化解释，介绍基于粒子滤波交通仿真模型实现流程。进而，设计并实现了基于粒子滤波的交通仿真系统。结合研究建立交通仿真模型对系统进行模块划分，对数据同化、动态数据注入、多线程管理、人机交互等主要模块进行设计与实现。最后，应用基于粒子滤波的交通仿真系统，构建直线和环形道路的仿真场景，对动态数据驱动的交通仿真框架的应用效果和仿真精度进行实验和分析。
[Abstract]:With the development of Intelligent Transportation system (its), micro-traffic simulation technology, as an effective tool to describe complex traffic behaviors, plays an increasingly important role in solving traffic problems. Many achievements have been made on the establishment of traffic simulation model and the development of traffic simulation system. However, the existing traffic simulation is usually based on historical data, neglecting all kinds of unexpected events in the course of system operation, and the accuracy of simulation results under real-time and dynamic conditions is low. With the development of high and new technology, the real traffic flow data can be easily obtained, which makes it possible for traffic simulation based on dynamic measured data. Dynamic data driven application system (DDDAS) is characterized by the effective combination of simulation and real data, so that the simulation can dynamically accept the feedback of measured data, so that the simulation converges faster and the results are more reliable. At present, DDDAS is widely used in the field of crisis management, engineering science and disaster prediction. Based on the above background, this paper aims to apply DDDAS paradigm to the microscopic simulation system Movsimand and propose a dynamic data-driven traffic simulation method. The method feedbacks the measured data of vehicle operation into the traffic state prediction to make the prediction results more accurate and reliable. Firstly, the DDDAS paradigm, Movsim logical flow, vehicle model and road network model are analyzed. Based on this, a dynamic data-driven traffic simulation framework is proposed. The running mechanism of the framework is analyzed, and the key technologies such as parallel processing and data assimilation are discussed. Secondly, the traffic simulation model based on particle filter is constructed. Particle filter algorithm provides a rigorous solution to the state estimation of nonlinear, non-Gao Si systems. Therefore, the particle filter algorithm is introduced to implement the data assimilation part of the framework. In this paper, two kinds of noise models, random translation model and segmented vehicle density model, data mapping model based on JSON middleware, two weight calculation models based on sliding window and sensor intelligent selection, and hierarchical resampling model are designed. A graphical explanation of the algorithm is given, and the flow chart of traffic simulation model based on particle filter is introduced. Furthermore, the traffic simulation system based on particle filter is designed and implemented. Combined with the research and establishment of traffic simulation model, the system is divided into modules, and the main modules such as data assimilation, dynamic data injection, multi-thread management and human-computer interaction are designed and implemented. Finally, the traffic simulation system based on particle filter is used to construct the simulation scene of straight line and ring road, and the application effect and simulation precision of the dynamic data driven traffic simulation framework are tested and analyzed.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495

【参考文献】