【作者】 任敏；

【导师】 王万良；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 随着Internet的飞速发展，新型网络应用不断涌现，用户数量迅速增加，使得网络流量急剧增加，由此引发的网络拥塞已经成为制约网络发展和应用的瓶颈问题。拥塞容易造成传输延迟和吞吐量等QoS性能指标下降，导致网络性能下降、网络资源利用率降低，从而无法提供有效的QoS保证。因此，有效地解决拥塞问题对于提高网络性能具有重要意义，如何更好地预防和控制网络拥塞也成为近年来网络研究领域的热点问题。 第一章在介绍Internet中的TCP／IP协议以及QoS路由的基础上，系统地介绍了TCP／IP网络拥塞控制的研究进展。特别是从系统的角度，分析了控制理论在拥塞控制应用中的优点和存在的问题。 第二章首先介绍了Internet网络中拥塞现象及其产生的原因，然后介绍了TCP／IP协议中传输层和网络层的各种拥塞控制策略，对主要的拥塞控制策略进行了分析，讨论了已有方法的不足及其改进方案。 第三章作为本文的研究重点，首先介绍了QoS路由的基本概念和原理，讨论了路由的图模型、状态信息和QoS路由策略，研究了路由器的拥塞控制策略。然后重点介绍了本文提出的一种基于双倍体遗传算法的组播路由算法，仿真结果表明该算法能够在参数(种群规模、浙江工业大学硕士学位论文摘要交叉率及变异率)选择适当的情况下，取得较好的收敛效果，并可以确定最优路由。最后，深入研究了基于神经网络的组播路由算法，提出了基于混沌神经网络的组播路由算法，仿真结果表明该算法能够避免陷入局部极小，并提高收敛速度，具有较好的性能。 第四章从复杂系统分析与控制理论的角度分析了计算机网络拥塞机理，建立了网络拥塞控制系统的数学模型，研究了基于PID、预测控制等控制理论方法的计算机网络系统拥塞控制方法。 第五章对本文所做的工作做了总结，对基于Qos的网络拥塞控制问题的解决进行了展望。 拥塞控制本身是一个极其复杂的问题，任何单一的拥塞控制机制都不能完整地解决这个问题，必须采用多种策略，从网络的各个部位、多角度全方位对拥塞加以控制，才能保证网络高效、.稳定运行。更多还原

【Abstract】 With the rapid development of Internet, widespread use of computernetworks, as well as the appearance of varied network applications hasmade network congestion control a significant problem. The networkcongestion has become a bottleneck problem, which restricts thedevelopment and application of network. Congestion often results in thedecline of QoS (Quality of Service) parameters such as transmission delayand throughput, the network performance and resource utilization are alsoaffected seriously, effective QoS guarantee cannot be provided accordingly.Therefore, it is very important to solve the congestion problem effectivelyfor improving the network performance. How to avoid and control networkcongestion has become an active issue in the field of computer networkresearch.In chapter one, a general introduction of TCP/IP protocol and QoSrouting is given. The development of TCP/IP network congestion control is introduced in detail. The advantages and limitations of control theory used in congestion control are analyzed in the system aspect.Chapter two is the foundation of the whole thesis. Firstly, the cause and situation of current Internet network congestion are introduced. Secondly, various congestion control strategies of transport layer and network layer in TCP/IP protocol are analyzed detailedly. Finally, different control strategies are compared, the disadvantages of existed methods are discussed and improved proposals are presented.Chapter three is the emphasis of the thesis. Firstly, the fundamental concepts and principle of QoS routing including graph model of routing, status information and QoS routing strategies are introduced. Router congestion control strategies are also proposed. Secondly, a new multicast routing algorithm based on double chromosomes genetic algorithm is realized. The simulation results show that this algorithm can get better convergent effect and select best routing under proper system parameters. Finally, the multicast routing algorithm based on neural network is studied deeply, and the multicast routing algorithm based on chaotic neural network is presented. The simulation results show that this kind of algorithm can avoid getting into local minimum and improve convergent effect.In chapter four, the computer congestion control mechanism isanalyzed from the point of complicated system and control theory, the network congestion control system mathematical models are constructed, computer network system congestion control methods based on control theories such as PID and predictive control are researched.In chapter five, the whole study work of this dissertation is summarized and the prospect of solving the network congestion control problem based QoS is looked forward.Congestion control is a sophisticated task that cannot be absolutely resolved by a single protocol or an algorithm. The only way to guarantee the steadiness and efficiency of network is to use various strategies to control the congestion in multiple aspects of the whole network.更多还原