【作者】 郭晓燕；

【导师】 陈阳舟；

【作者基本信息】 北京工业大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 异步转移模式ATM(Asynchronous Transfer Mode)采用基于连接的快速分组交换方式进行数据传输。它具有多业务传输特性，能同时传送语音、视频和数据等业务。与传统的网络有区别，因此它的拥塞控制问题也与普通的网络不完全相同。本文首先介绍了ATM网络的拥塞控制理论，分析了ATM网络拥塞控制问题的特殊性，然后采用最优控制、线性二次型高斯最优随机控制和自己提出的一种算法对ATM网络信元级基于速率的拥塞控制问题进行了研究。仿真证明了算法的有效性，本论文的主要工作从下列三方面展开： 第一，对一种最优控制算法进行了改进，将改进后的算法用于带宽恒定情况下单ATM交换节点的拥塞控制问题中。采用c实现了算法和ERICA算法，在此基础上采用matlab对两种算法进行了仿真比较，仿真结果证明改进后的最优算法对既定控制目标的实现上优于ERICA算法。 第二，考虑到ATM同时传输语音、视频和数据业务的特殊性，因此ABR信源的可用带宽是一个不确定的量，将改进后的最优算法扩展为线性二次型高斯最优随机算法对单交换节点ATM网络的拥塞控制问题进行了研究。采用c和matlab进行了仿真并在相同条件下将该算法和ERICA算法进行了比较。仿真结果证明该算法总体上优于ERICA算法。另外该算法计算简单，易于在交换机上实现。 第三，对于多交换节点的ATM链状网络的拥塞控制问题，提出了一种速率分配算法。考虑到模型的复杂性和并行性，提出采用遗传算法对该算法的性能指标进行优化。最后针对一个具体的网络拓扑结构进行了分析，并给出了算法的分析和流程设计。更多还原

【Abstract】 The transmission method of ATM belongs to fast packet switching mode, and it provides several services such as voice, video and data. The pol icy to control the congestion of ATM differs from those of other networks. In this paper, ATM congestion control theory and its particularity are given, optimal control algorithm, LQG stochastic control algorithm and an algorithm put forward by the author to resolve ATM cell congestion, are introduced. The deployment of this paper is following:First, improves an optimal algorithm, and applies the modified algorithm to the congestion control problem of single ATM node whose bandwidth is invariable. We program this algorithm and ERICA, and the result shows that the improved algorithm has better performance than ERICA.Second, ATM transmits voice, video, and data at the same time, some bandwidth is reserved for CBR cells and VBR cells because voice and video cells have high priority, and thus the bandwidth available to ABR cells is variable. Take the above factors into consideration, the improved optimal algorithm is modified to LQG stochastic control algorithm for single ATM switching node whose bandwidth is variable. Both C and matlab simulation have been done, and we find that the stochastic control approach is preponderant, especially in controlling queue length of the switch node. In addition, it is easy to apply the stochastic algorithm to the switch for its simplicity.Third, this paper brings forward a rate allocation algorithm inAbstractthe case of multi-source and multi-destination transmit through multi switch. Considering complexity and correlation between variables, we decide to adopt genetic algorithm to the optimization. For a given network, we analyze and design the flow chart of implementing the algorithm in detail.更多还原