基于神经元学习的自适应AQM算法及仿真研究

摘要随着网络规模的增大，用户迅速增加，同时大量应用业务和实时多媒体数据的涌现，也使网络流量急剧增加，网络带宽资源受限从而导致网络拥塞和系统性能的下降。近年来，主动队列管理机制（AQM）受到了国内外学者的关注。它是一种有效的拥塞控制方法，能够有效的抑制被动式管理机制中的满队列、死锁和全局同步等问题。本课题主要针对现有的AQM算法进行仿真研究，并从队列长度、丢包率等性能指标对RED、REM、PI等算法进行仿真对比。本文主要针对基于神经元网络的自适应AQM算法进行仿真研究，详述了该算法的原理，给出了稳定性分析。对于NS-2软件也有简介，并与前述三种算法仿真对比，仿真结果表明，该算法具有更好的收敛性，队列长度波动也更小，稳定性增强。64503

毕业论文关键词网络拥塞主动队列管理仿真 NS-2

毕业设计说明书（论文）外文摘要

Title An Adaptive AQM Algorithm Based on Neuron Learning and Its Simulation

Abstract As the rapid increase of network size and users, while a large number of application services and real-time multimedia data emerging, made the network traffic increase dramatically. The limited network bandwidth resources lead to network congestion and system performance degradation. In recent years, active queue management mechanism (AQM) has received the attention of scholars both at home and abroad. It is an efficient congestion control method capable of effectively suppressing full queue, deadlock problems and global synchronization of passive queue management mechanisms. This paper is mainly aimed at simulation research of the existing AQM algorithm, and gives the simulation comparison from the performance indexes such as the queue length and packet loss rate of RED, REM and PI algorithm. This paper focuses on simulation research of adaptive neural network AQM algorithm. The principles of the algorithm and the stability analysis are described in this paper. This paper also has a brief introduction of NS-2, and compared with the foregoing three algorithm. Simulation results show that this algorithm has better convergence and less fluctuation of queue length and enhanced stability.

Keywords Network congestion Active Queue Management Simulation NS-2

1 绪论 2

1.1 网络拥塞 3

1.2 拥塞控制 4

1.2.1 源算法 4

1.2.2 链路算法 5

1.3 本章小结 8

2 几种经典AQM算法 8

2.1 引言 9

2.2 RED算法 9

2.3 REM算法 10

2.4 PI控制器 12

2.5 本章小结 13

3 基于神经元网络的自适应AQM算法及稳定性分析 13

3.1 引言 13

3.2 基于PID神经元网络的TCP/AQM系统模型基于神经元学习的自适应AQM算法及仿真研究:http://www.751com.cn/zidonghua/lunwen_71711.html