分布式交通网络仿真平台的动态负载均衡算法研究及应用
发布时间：2018-08-31 08:59
【摘要】：采用分布式交通仿真系统来对现实交通情况进行模拟,是提高城市交通效率和机动性的有利途径。而分布式仿真系统中,任务是否均衡分配关系到整个仿真系统的仿真速度。因此,如何使分布式交通仿真节点的负载均衡成了亟待解决的问题。本文以分布式交通仿真平台为基础,借鉴博弈论的思想,提出了基于博弈论的动态负载均衡算法,并通过仿真实验和理论分析来验证了算法的优越性。本文的主要工作包括以下几点:(1)提出了基于分布式交通仿真的网络模型。首先,在自主研发的分布式微观交通网络仿真平台的基础之上,设计了混合模式的动态负载均衡模型。该模型采用CS和P2P的结合来完成,总控端与仿真终端间以CS的模式存在,而仿真终端和仿真终端之间以P2P的模式存在。同时为了提高仿真系统的仿真速度、降低数据采集成本,本文将路网中道路简化为带权重的点,而路口简化为带权重的边。采用该模型对路网进行仿真,能够在不失仿真的真实性的同时,简化仿真路网模型,提高仿真系统的仿真速度和仿真效率。(2)提出了基于博弈论的动态负载均衡算法。本文借鉴博弈论中纳什均衡的思想,将道路的迁移变为主动行为。算法首先对道路的效用函数进行定义,并通过理论推导证明该函数存在纳什均衡;然后,采用迭代生长的方式来对地图网络进行初始分割;最后,道路将根据效用函数计算自身的费用,并根据自身的费用来决定是否进行迁移以及迁移的目的终端,直到完成整个路网的均衡;并且在仿真系统运行中再次出现不均衡状况时,同样也可以依据道路的费用来对道路进行迁移,并再次达到均衡状态。(3)进行了仿真对比实验。本文采用自主研发的微观交通仿真系统为实验平台,通过设定实验参数等,与经典的递归对分算法进行了两大组对比实验。实验的主要内容包括:终端数变化及车辆数变化对仿真时间、仿真时间方差、通信时间总和及通信时间方差的影响。实验结果证明,本文提出的基于博弈论的动态负载均衡算法的优越性。通过理论及对比仿真实验的结果验证,本文提出的基于博弈论的动态负载均衡算法能够均衡的划分路网,并且能够解决大规模分布式交通仿真系统的动态负载均衡问题。
[Abstract]:Using distributed traffic simulation system to simulate the real traffic situation is a good way to improve urban traffic efficiency and mobility. In the distributed simulation system, whether the task is balanced or not is related to the simulation speed of the whole simulation system. Therefore, how to balance the load of distributed traffic simulation nodes becomes an urgent problem. Based on the distributed traffic simulation platform and using the idea of game theory for reference, a dynamic load balancing algorithm based on game theory is proposed in this paper, and the superiority of the algorithm is verified by simulation experiments and theoretical analysis. The main work of this paper is as follows: (1) A network model based on distributed traffic simulation is proposed. Firstly, based on the distributed micro traffic network simulation platform, a hybrid dynamic load balancing model is designed. The model is implemented by the combination of CS and P2P. The CS mode exists between the master control terminal and the simulation terminal, while the P2P mode exists between the simulation terminal and the simulation terminal. At the same time, in order to improve the simulation speed and reduce the cost of data acquisition, the road in the road network is simplified as the point with weight, and the intersection is simplified as the edge with weight. Using this model to simulate the road network can simplify the simulation road network model and improve the simulation speed and efficiency without losing the reality of simulation. (2) A dynamic load balancing algorithm based on game theory is proposed. This paper draws lessons from Nash equilibrium in game theory and turns road transfer into active behavior. Firstly, the utility function of the road is defined, and the Nash equilibrium is proved by theoretical derivation. Then, the map network is initially segmented by iterative growth. The road will calculate its own cost according to the utility function and decide whether to carry out the migration and whether to migrate the destination terminal according to its own cost until the balance of the whole road network is completed; and when the simulation system runs again when the imbalance occurs again. It is also possible to transfer the road according to the cost of the road and reach the equilibrium state again. (3) A comparative simulation experiment is carried out. In this paper, the micro traffic simulation system developed by ourselves is used as the experimental platform. By setting the experimental parameters and so on, two groups of comparative experiments are carried out with the classical recursive algorithm. The main contents of the experiment are as follows: the influence of terminal number change and vehicle number change on simulation time, simulation time variance, communication time sum and communication time variance. The experimental results show the superiority of the proposed dynamic load balancing algorithm based on game theory. The results of theoretical and simulation experiments show that the proposed dynamic load balancing algorithm based on game theory can evenly divide the road network and solve the dynamic load balancing problem of large-scale distributed traffic simulation system.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491;TP391.9
[Abstract]:Using distributed traffic simulation system to simulate the real traffic situation is a good way to improve urban traffic efficiency and mobility. In the distributed simulation system, whether the task is balanced or not is related to the simulation speed of the whole simulation system. Therefore, how to balance the load of distributed traffic simulation nodes becomes an urgent problem. Based on the distributed traffic simulation platform and using the idea of game theory for reference, a dynamic load balancing algorithm based on game theory is proposed in this paper, and the superiority of the algorithm is verified by simulation experiments and theoretical analysis. The main work of this paper is as follows: (1) A network model based on distributed traffic simulation is proposed. Firstly, based on the distributed micro traffic network simulation platform, a hybrid dynamic load balancing model is designed. The model is implemented by the combination of CS and P2P. The CS mode exists between the master control terminal and the simulation terminal, while the P2P mode exists between the simulation terminal and the simulation terminal. At the same time, in order to improve the simulation speed and reduce the cost of data acquisition, the road in the road network is simplified as the point with weight, and the intersection is simplified as the edge with weight. Using this model to simulate the road network can simplify the simulation road network model and improve the simulation speed and efficiency without losing the reality of simulation. (2) A dynamic load balancing algorithm based on game theory is proposed. This paper draws lessons from Nash equilibrium in game theory and turns road transfer into active behavior. Firstly, the utility function of the road is defined, and the Nash equilibrium is proved by theoretical derivation. Then, the map network is initially segmented by iterative growth. The road will calculate its own cost according to the utility function and decide whether to carry out the migration and whether to migrate the destination terminal according to its own cost until the balance of the whole road network is completed; and when the simulation system runs again when the imbalance occurs again. It is also possible to transfer the road according to the cost of the road and reach the equilibrium state again. (3) A comparative simulation experiment is carried out. In this paper, the micro traffic simulation system developed by ourselves is used as the experimental platform. By setting the experimental parameters and so on, two groups of comparative experiments are carried out with the classical recursive algorithm. The main contents of the experiment are as follows: the influence of terminal number change and vehicle number change on simulation time, simulation time variance, communication time sum and communication time variance. The experimental results show the superiority of the proposed dynamic load balancing algorithm based on game theory. The results of theoretical and simulation experiments show that the proposed dynamic load balancing algorithm based on game theory can evenly divide the road network and solve the dynamic load balancing problem of large-scale distributed traffic simulation system.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491;TP391.9
【共引文献】
相关期刊论文 前10条
1 胡红;刘小明;杨孝宽;;基于最小安全间距的应急交通疏散车辆跟驰模型[J];北京工业大学学报;2007年10期
2 许有俊;郭楠;;城市快速路出口辅路附加车道长度的仿真研究[J];北京工业大学学报;2009年11期
3 许有俊;张s，
本文编号：2214517
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2214517.html
