Application of Cooperative Agents for Routing in Large-scale High-speed Networks

在这项工作中，我们提出了一个基于协作代理的智能呼叫接纳控制(CAC)和路由框架。提出的框架基于分布式人工智能(DAI)方法，该方法处理人工代理的设计以开发智能系统。我们介绍了两种类型的代理：简单代理和智能代理。智能代理基于模糊逻辑(FL)和遗传算法(GA)。拟议的战略运作如下。在基于FL的GAG代理已经决定接受网络中的连接之后，激活基于GA的IntraD路由代理以找到可行路径。所提出的路由算法是源路由和分布式路由的结合。它使用域内的源路由和域间的逐跳路由。在这份报告中，我们提出了模糊接纳控制器(FAC)、IntraD代理和ARGAQ方法，并通过仿真评估了它们的性能。从评价结果来看，我们得出如下结论。The Perf…对所提出的FAC方案进行了更多的评估，结果表明，与传统方法相比，模糊等效容量估计器(FEES)具有更好的等效容量估计；模糊等效容量估计器和平稳逼近相结合给出了更准确的等效容量估计；FAC方案具有比等效容量方法更好的接纳区域。提出的基于代理的路由和CAC框架具有CAC和路由一体化、源路由和分布式路由相结合、搜索空间小、自适应、灵活、智能、能避免洪泛、能避免根环、可伸缩等特点。与GLBR算法相比，IntraD算法具有以下特点：采用了一种新颖的基因编码方法，搜索效率高，遗传操作简单。ARGAQ算法的性能评估结果表明，与GLBRQ算法相比，ARGAQ算法具有更快的响应时间；ARGAQ算法具有简单的遗传操作，可以支持两个服务质量参数；ARGAQ算法能够找到比ARGA方法更好的服务质量路径。作为下一步的工作，我们希望研究FAC方案和基于遗传算法的路由算法在实际环境中的实现问题。我们打算在并行GA体系结构中实现IntraD代理。较少

In this work, we propose an intelligent Call Admission Control (CAC) and routing framework which is based on cooperative agents. The proposed framework is based on Distributed Artificial Intelligence (DAI) approach, which deals with design of artificial agents to develop intelligent systems. We introduce two types of agents : simple and intelligent agents. The intelligent agents are based on Fuzzy Logic (FL) and Genetic Algorithm (GA). The proposed strategy operates as follows. After a FL based GAG agent has decided to accept a connection in the network, a GA based IntraD routing agent is activated to find a feasible path. The proposed routing algorithm is a combination of source and distributed routing. It uses source routing inside a domain and hop-by--hop routing for inter-domain. In this report, we proposed and evaluated by simulations the performance of the Fuzzy Admission Controller (FAC), IntraD agent and ARGAQ method. From the evaluation results we conclude as follows. The perf … More ormance evaluation of proposed FAC, scheme showed that the Fuzzy Equivalent Capacity Estimator (FECE) had a good equivalent capacity estimation compared with conventional methods ; the combination of FECE and stationary approximation give a more accurate estimation of equivalent capacity ; and FAC scheme has a better admission region than the equivalent capacity method. The proposed agent-based routing and CAC framework have the following characteristics : integration of CAC and routing, combination of source and distributed routing, reduction of search space, is adaptive, flexible and intelligent, can avoid flooding, can avoid rooting loops, and is scalable. The IntraD algorithm has the following features : it uses a novel gene coding method, has an efficient search, and has easy genetic operations compared with GLBR. The performance evaluation of proposed ARGAQ method showed that ARGAQ has faster response time compared with GLBRQ method ; ARGAQ has simple genetic operations and can support two QoS parameters ; and ARGAQ can find better QoS routes than ARGA method. As future works, we would like to investigate the implementation issues of FAC scheme and GA-based routing algorithms in real environments. We intend to implement the IntraD agent in a parallel GA architecture. Less