多QoS约束条件下的多目标网络优化  

Multi-objective Network Optimization with Multi-constrains QoS Based on Genetic Algorithm

作　　者：明丽洪 吕光宏[1] 向虹佼 

机构地区：[1]四川大学计算机学院,四川成都610065

出　　处：《电子科技》2014年第3期18-21,共4页Electronic Science and Technology

基　　金：863国家高新技术研究发展计划基金资助项目(2008AA01Z105)

摘　　要：多约束、多业务、多目标的网络优化是一个复杂且涉及范围广泛的课题。文中在对该课题进行分析的基础上,提出了一种基于遗传算法的多目标网络优化算法(MOPGA)。该算法使用了多约束条件下的路径集预处理,使得每项业务能够获得所需的QoS服务质量,通过对所有业务的路由号进行编码,将问题的解空间转换到遗传算法的搜索空间,达到对全网业务的综合考虑。改进后的适应度函数刻划了网络的费用、链路利用率方差和最大链路利用率、爆破处理以及个体淘汰机制增加了种群多样性,挣脱了未成熟收敛。以求解精度作为算法终止条件,使得算法运行时间减少。仿真实验表明,所提出的算法能高效、快速解决实际多目标网络优化问题,同时在满足多QoS约束条件下可均衡各子目标函数。Multi-constrains multi-traffic and multi-objective network optimization is a complex issue. A multiobjective network optimization algorithm based on genetic algorithm （MOPGA） is proposed in this paper. Firstly, the algorithm meets the quality of service of each traffic in terms of multi-constrains path set preprocessing. Secondly, it transfers a solution space of the problem into a search space of the genetic algorithm. Thirdly, the improved fitness function depicts the network total cost, link utilization variance and the maximum link utilization. Fourthly, the blast processing and the individual selection mechanism increase the diversity of population and avoid falling into local optimum. Finally, according to the actual error requirement of the different traffic, the algorithm uses the solution error as an end condition. The simulation results show that it can efficiently achieve actual multi-objective network optimization in high speed and balance every subgoal functions while satisfying the multi-constrains QoS.

关 键 词：多业务 多目标 QOS 遗传算法 网络优化 

分 类 号：TP393.027[自动化与计算机技术—计算机应用技术]