【作者】 金劲；

【导师】 洪毅； 李战明；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2011， 博士

【摘要】 在Internet中,由于多媒体通信和网络视频的增长,网络资源和网络需求之间的矛盾越来越突出,网络路由和网络流量的平衡成为制约网络发展的瓶颈。抑制网络拥塞,提高网络的质量,使网络达到动态均衡,已成为Internet相关技术的研究热点。网络单播、组播路由是检验网络性能的重要指标,在网络研究中的作用越来越重要。特别是在多约束网络中,采用QoS指标讨论网络的质量性能,使网络达到动态均衡;并对网络的延时、延时抖动、带宽、丢包率、业务代价等多个参数同时提出性能要求,这些参数相互独立时,选择满足多个参数限制的路由就成为NP完全问题。在研究网络单播、组播路由的同时,国内外许多学者对网络流量展开研究,主要集中在流量工程(Traffic Engineering,即TE),而流量工程的热点问题又集中在多约束路由。在网络满足多约束的条件下,根据网络流量与资源的状态,通过实施合理的控制,使流量均衡地分布在现有的网络中,从而优化网络的运行性能。本文主要包含网络单播路由、组播路由、网络流量均衡等几个方面的内容,利用群集智能算法的特点,对单播路由优化、组播路由优化、网络流量控制等方面进行系统的研究,同时对算法的收敛性和网络安全等方面进行较为系统分析。本文的主要研究成果包括:1.对QoS网络的路由优化改进技术的深入研究,第一,提出了基于粒子融合的QoS蚁群优化路由算法,使用粒子群算法生成初始解,避免了蚁群算法在局部最优化,拓展了蚁群算法的求解范围,提高了算法的自适应能力和优化精度。第二,提出了基于混沌算子的蚁群优化改进多约束QoS路由算法,利用混沌“随机性”、“遍历性”及“规律性”,能够快速获得全局最优解的优点,采用混沌因子改进蚁群算法,提高了搜索性能,算法搜索到的结果总体要明显好于基本ACO算法。仿真结果表明,两种改进算法具有较高的优化性能。2.结合组播路由网络的特点,对多约束条件在组播网络展开分析,在此基础上提出了基于克隆粒子群融合的约束QoS组播树算法。在提出新的播路由算法中,通过粒子的速度和位置变化查找组播树,并且利用免疫克隆算法混合搜索,从而减少了局部搜索和全局搜索的时间。通过克隆算子的引入,增加了克隆复制、克隆变异、克隆选择等3个环节,在克隆变异阶段,利用变化后的个体以一定的概率具有更高的适应性能,然后利用克隆选择环节,避免了种群易经常出现的退化,而且提高了算法收敛速度和全局搜索能力。仿真结果表明,该算法具有更好的优化性能。3.通过对网络流量和网络路由之间的关系进行了深入分析,提出了基于带宽受限模糊权重的蚁群优化控制算法(Fuzzy-ACO)。在基于带宽受限的蚁群优化控制算法中,利用模糊控制网络流量权重,建立网络流量的数学模型,降低了大量的探测分组带来了网络开销;并通过采用时间顺序输入不同流量,可以动态及时的反映网络性能,对网络性能进行实时监控,从而使网络流量和网络路径达到动态平衡。同时将系统流量权重融入信息素中,利用信息素动态地控制在多条路径中选择最佳路径,提高了蚁群算法全局搜索能力。仿真结果表明,算法优化效果明显,运行速度快,并显著加快了传统算法网络流量的探索收敛速度。4.对于多约束条件的蚁群优化算法,往往约束条件的取舍及函数的设计尤为重要,针对QoS条件下蚁群算法的收敛性展开讨论,重新定义其信息素的选取,从理论上证明该算法的收敛性;并且还论证在带QoS约束条件的蚁群算法中,信息素发生变化的时间点,以及信息素的取值范围,从而证明该类蚁群优化算法收敛的可控制性,通过实验仿真说明该算法的实用性。对该算法局部和全局收敛性展开研究,提出了普遍意义下的收敛条件,为这一类约束条件下的蚁群算法进一步研究奠定了良好的基础。5.通过对网络安全的分析,借助生物聚类的机理,防范网络入侵的危害,提出了一种基于交叉融合粒子群优化算法的聚类分析,由粒子群算法形成初步的聚类中心,再由蚁群算法进行二次优化,仿真实验表明,该算法与基本聚类算法相比较,聚类组合方法能够明显改善聚类质量。论文对网络单播路由、组播路由、网络流量均衡做了较为全面深入的分析和讨论,提出了多种有效的改进措施,并证明了算法的收敛性,提出了网络安全聚类分析的方法,实现了群集智能算法在网络分析上的应用。最后对所做工作进行了总结,并提出了进一步研究的方向。更多还原

【Abstract】 In the Internet, with the increasing of multimedia communications and network video, the conflict between resource and demand of network is more and more remarkable. The balance of routing and flux in the network is most important factor in the development of network. How to restrain the congestion of network and improve the quality of network, which makes the network dynamic balance, is the focus technique research in the network.The unicast and multicast routing of network is importance capability inspection of network, which is paid more attention by people these days, especially in the multiple constrained condition of network. The quality-of-service(QoS)is adopted in the quality of network mostly. For the dynamic balance in the network, the demands of network character parameter such as delay, delay-jitter, bandwidth, packet-loss and cost are considered at the same time, which are independent each other. The routing, which are content the multi-parameter limit, is NP-complete problems in the network.In the research of the unicast and multicast routing of network simultaneously, many scholar are studied to investigate into flux of network at home and abroad presently, which are all focus on traffic engineering(TE). Recently the emphases questions of the traffic engineering focus on multiple constrained condition of routing. Considering of the multiple constrained condition, this dissertation makes the flux distributing in the network uniformity and optimizes the dynamic capability of network based on flux of network and state of resource through carrying out reasonable control.The dissertation contains the several parts content of unicast routing, multicast routing and balance of traffic network. By uses of character of intelligence algorithm, a new measure is studied by the research of unicast routing, multicast routing and control of traffic engineering as a whole. At the same time, the convergence analysis of algorithm and the safety of network are analyzed totally. The main achievements of this dissertation include:1.How to solve QoS optimizing routing problem was researched deeply by improvement technique. Firstly, A new algorithm was brought forward by multiple constraint optimization based on particle swarm amalgamation combination of ant colony algorithm, which adopts particle swarm optimization to get initialization a new solution by searching routing and avoided to be trapped into local seeking solution only by ant colony algorithm. This algorithm increased the scope of searching better routing, advanced self-adaptable capability and accurate optimizing. Secondly, a multiple constrained QoS algorithm based on chaos and ant colony optimization was proposed. By using of the properties of randomicity, regularity and ergodicity of chaos, the mixed algorithms found out the whole seeking solution quickly. Then, the mixed means improved ant colony algorithms by chaos factor and improved the searching capability. The result of searching had the advantage over the base ant colony algorithm remarkably. The experimental results show that these two new improved algorithms have high efficiency.2.Linking to the character of multicast routing in network, combination of clone and particle swarm optimization based on multiple constrained multicast routing algorithms was put forward though analyzing of multiple-constrain in the multicast routing network. The new multicast routing algorithm was studied by the change of speed and location finding multicast tree and by the Immunity Clone algorithm to search best route, which decreased the time of the local and global searching. The clone algorithm added process of clone copy,clone mutation and clone selection. In the course of clone mutation, the algorithm was high adaptability with the definite probability by changing. Then, in the course of clone selection, the algorithm avoided the degeneracy of genus regularly and enhanced the speed of convergence algorithm and the global searching capability. The simulation experimental results show that the improved algorithms have better optimization performance.3.By analyzing the relation between networks of traffic and routing deeply, the fuzzy weight value of routing controlled ant colony optimization algorithm based limited bandwidths(Fuzzy-ACO)were proposed based on the research. In the ant colony optimization algorithm based limited bandwidths, networks of traffic was controlled in the weight value by fuzzy. Since the mathematics model was founded with networks of traffic, spending of networks was decreased by large numbers of detector with grouping. By means of inspecting content of network with real time, the networks of traffic and routing were balanced dynamically. Simultaneously, the networks of traffic with the weight value were connected with pheromone, which dynamically adjusts optimal routing selected among multiple paths. The ant colony algorithm achieves globally searching ability. The simulation results show that the given algorithm was effective and high speed,in which it dramatically improved the exploring speed of convergence in network traffic by traditional networks traffic algorithm.4.For the multiple constrained condition ant-colony-optimization(ACO) algorithms, the making choice of constrained condition and designing function was very importance. The convergence of ant colony algorithm under the quality-of-service(QoS)condition was studied. By redefining the selection of pheromone, the convergence of algorithm was demonstrated by applying theory. Through the changing time of pheromone and scope of value of pheromone was analyzed in the ant colony algorithm with QoS condition, the controllability of ACO was also proved theoretically. The simulation results show that the given algorithm was practicable, by making the algorithm converge both locally and globally under a general convergence condition. This works may provide a foundation for further theoretical studies on the multiple-constrain QoS of ACO.5.By analyzing the network’s security, the threat of the intrusion on line was detected with theory of biology clustering. The clustering analysis way by combination of particle-swarm-optimization(PSO)and ant-colony-optimization(ACO)algorithm was discussed. Firstly, the center and number of clustering are determined by using the PSO, and then the above clustering results are optimized by the K-means algorithm combining with ACO. The simulated experiments show that the combining algorithm is obviously superior to some common clustering algorithms since it has obvious advantage in optimization capacity.In the dissertation, the unicast routing, multicast routing and balance of traffic network was analyzed and discussed completely. Some effective improvement methods were proposed and the convergence of the algorithm was demonstrated in this dissertation. The Cluster Analysis in the safety of network was proposed. Those all swarm intelligence algorithms were realized the application of analysis in the network. Lastly, the work of this dissertation is summarized, and further research directions were indicated.更多还原