微观交通仿真中的路径规划和车辆更新的并行计算

发布时间：2018-03-03 20:24

本文选题：并行计算　切入点：微观交通仿真　出处：《长安大学》2014年硕士论文　论文类型：学位论文

【摘要】：微观交通仿真对交通流的描述细化到单个车辆的行为模型，能真实的模拟实际交通流的运行状态，也方便进行交通管理、控制、规划和预测评估。随着城市规模拓展，车辆不断增多，以及交通信号控制与交通流量之间的耦合关系，交通仿真系统中路网规模和车辆数目都不断增加，对交通仿真软件的仿真效果、实时性以及分析功能的要求也不断提高。在大规模路网下，使用传统串行微观交通仿真难以满足实时仿真计算的要求。并行计算是提高计算性能的一种有效手段，并行微观交通仿真利用多个计算资源并行协同完成仿真计算，可以明显提高微观交通仿真的效率。本文在总结典型的并行微观交通仿真方法的基础上，分析了微观交通仿真的可并行性和基本的并行化方法，结合MPI(Message Passing Interface)特性设计了主从模式下各机之间的工作方式，将多台计算机通过网络互联组成分布式平台来实现并行计算，主要实现的工作包括：（1）在GPU协同CPU计算的异构系统中，使用CUDA(Compute Unified DeviceArchitecture)实现了路径规划的并行计算，在较大规模的路网数据下进行实验，验证了该方法可以有效提高路径规划的效率。（2）使用基于空间域划分的方法分解任务，，并结合路网的三层结构表示，将路网按照道路和交叉口的编号均分给多个仿真计算机，由多个仿真计算机来并行协同完成仿真计算。（3）使用MPI消息传递机制实现了主从模式下的分布式多机之间的通讯：仿真端使用多线程技术来完成消息发送和接收，主控端在每个仿真时间步收集仿真端所有车辆的信息。（4）设计了不同路网和不同车辆数目下的仿真实验，分析了影响分布式并行微观交通仿真加速效果的主要因素。
[Abstract]:The description of traffic flow in microscopic traffic simulation is detailed into the behavior model of a single vehicle, which can simulate the running state of actual traffic flow, and facilitate traffic management, control, planning and prediction and evaluation. With the increasing number of vehicles and the coupling relationship between traffic signal control and traffic flow, the scale of road network and the number of vehicles in the traffic simulation system are increasing, so the simulation effect of traffic simulation software is obtained. The requirement of real-time and analysis function is also increasing. Under the large-scale road network, it is difficult to meet the requirement of real-time simulation by using traditional serial microscopic traffic simulation. Parallel computing is an effective means to improve the computing performance. Parallel micro traffic simulation can improve the efficiency of micro traffic simulation. On the basis of summarizing the typical parallel microscopic traffic simulation methods, this paper analyzes the parallelism and the basic parallelization methods of the microscopic traffic simulation, and designs the working mode between the computers in the master-slave mode combined with the characteristics of the MPI(Message Passing interface. Several computers are connected to each other to form a distributed platform to realize parallel computing. The main tasks are as follows:. 1) in the heterogeneous system of GPU collaborative CPU computing, CUDA(Compute Unified device Architecture is used to realize the parallel computing of path planning. Experiments are carried out on a large scale road network data, and it is verified that this method can effectively improve the efficiency of path planning. 2) the task is decomposed by the method based on spatial domain division, and combined with the three-layer structure of the road network, the road network is divided into several simulation computers according to the numbers of the road and intersection, and the simulation calculation is accomplished by multiple simulation computers in parallel. The communication between distributed multi-computers in master-slave mode is realized by using MPI messaging mechanism. Multi-thread technology is used to send and receive messages. The main control terminal collects the information of all vehicles in each simulation time step. The simulation experiments of different road networks and different number of vehicles are designed, and the main factors influencing the acceleration effect of distributed parallel microscopic traffic simulation are analyzed.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U492.22

【参考文献】