AdHoc网络单神经自适应PID主动队列管理  

Single Neural Adaptive PID Active Queue Management in Ad Hoc Network

作　　者：孙慧玉[1] 姜文刚[1] 

机构地区：[1]江苏科技大学,江苏镇江212003

出　　处：《计算机仿真》2014年第1期307-310,共4页Computer Simulation

基　　金：国家自然科学基金(60974129)

摘　　要：针对无线自组网的网络拥塞问题,考虑网络节点的移动性以及网络拓扑结构异变性引起的局部和全局拥塞,提出了一种单神经自适应PID的主动队列管理算法。算法对现有的Ad Hoc网络移动节点的队列管理算法进行改进,利用单神经元结构简单的特点,同时结合神经网络强大的自适应、自学习能力,实现对无线节点缓冲区队列长度的有效控制。实验结果表明,新的中间节点拥塞控制算法具有响应时间短、丢包率低、网络流量吞吐率高等优点,在动态的网络环境下具有更好的稳定性和鲁棒性,综合性能优越于传统的PID控制算法。Active queue management strategies are important means for Ad Hoc internet congestion control. The paper proposed a single neuron adaptive PID active queue management by taking the mobility of nodes and the muta- tion of network topology into account. Much work was devoted to find an improved algorithm. The single neuron can effectively control the queue length with a simple structure, combined with the advantages of the neural network, pow- erful adaptive capacity and self-learning ability. Simulation results demonstrate that the new congestion control algo- rithm has a short response time, lower packet loss and higher throughput. In a dynamic network environment, the new strategy obtains better stability and robustness and its overall performance is superior to the traditional PID control algorithm.

关 键 词：拥塞控制 主动队列管理 无线自组网 单神经自适应比例积分微分算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]