【作者】 杨亮；

【导师】 古钟璧；

【作者基本信息】 四川大学 ， 模式识别与智能系统， 2001， 硕士

【摘要】 在ATM网络中，业务量不可预期的统计波动和网络内的故障都会造成业务阻塞。如果产生阻塞，已建立连接的网络性能就会下降。ATM网络一般采用预防性的阻塞控制机制，如用法参数控制(UPC)，连接允许控制(CAC)来避免阻塞状况的发生。ATM网络中众多的业务源属于不同的业务流，具有不同服务质量，因此业务管理功能对于控制网络内的业务流十分必要。通过在呼叫建立阶段的信令过程，网络和业务源就一些业务特征参数达成协定。如果业务源违背了业务参数，阻塞的可能性就增加了。网络必须控制业务流，并检测到违背协定的信元。所以快速检测出违约信元是一个好的业务量控制器的最重要的性能之一。在本文中，提出了一种具有高检测违约信元能力的模糊控制器。该模糊控制器是基于漏桶机制的，具有两个输入量，当前峰值信元速率和当前缓存中信元数，输出量是用来调节令牌产生速率的比例系数，从而调节信元释放到网络的速率。对于控制器核心的模糊规则，利用一种新的自适应遗传算法进行优化。以开关模型的声音信号包作为信源的对该模糊漏桶控制器进行仿真实验，结果证明本文提出的模糊漏桶机制较之传统的漏桶机制有着更快的违约信元检测能力，同时也表明本文提出的自适应遗传算子在对模糊规则的优化上是有效的。更多还原

【Abstract】 In Asynchronous Transfer Mode (ATM) networks, congestion can be caused by unpredictable statistical fluctuations of traffic flows and fault conditions within the network. If congestion happens, then the network performance for the already established connection will decrease. ATM networks use the preventive congestion control mechanisms such as Usage Parameter Control (UPC) and Connection Admission Control (CAC) to avoid congested conditions. Knowing that many sources in ATM networks have variable traffic stream with different QoS characteristics, traffic management functions become necessary to control the traffic flows within the network. By using the signaling procedures at the call setup phase, the network and source reach an agreement for some traffic characteristic parameters. If the source violates the traffic parameters, then the probability of congestion increases. So the network must control the source traffic streams and detect well the violating cells. Therefore, fast detection of any violating source is one of the most important characteristics of a goodtraffic policer. In this paper we propose a fuzzy traffic policer with high ability in detection of violating sources. Our proposed fuzzy controller has two inputs, current peak cell rate and the current number of the buffer, the output of fuzzy controller is a factor .which is used to adjust token rate so as to change the cell release rate. As the core of fuzzy controller, the fuzzy rule was also optimized by a new proposed self-adaptive genetic algorithm. Simulation results obtained from packetized voice sources based on on-off model, show that the proposed fuzzy leaky controller has better selectivity than the conventional leaky bucket. It is observed that the proposed controller has better ability to detect the violation cell, especially in the detection of peak cell rate. In the meanwhile, the proposed genetic operator was effective in the process of optimizing fuzzy rules.更多还原