单神经元自适应PID控制AQM算法的改进

Improved single neuron adaptive PID control AQM Algorithm

针对PID主动队列管理(AQM)中参数固定、不能自整定、无法适应复杂的非线性网络环境等缺点,提出了一种改进的单神经元自适应PID控制智能主动队列管理算法—GSNAPID算法。将神经元比例系数进行了在线自调整以及修改加权系数学习修正部分。基于NS2的仿真结果表明,改进的单神经元自适应PID控制智能主动队列管理算法具有更好的收敛性,能将队列长度更快速的收敛到期望值附件,在复杂的网络环境里能更好的保持鲁棒性和稳定性,比传统PID控制和单神经元自适应PID控制主动队列管理算法优越。

Focus on the disadvantages of PID active queue management（AQM）,such as fixed parameters,cannot self-setting,cannot adapt to complex nonlinear network environment and so on,an improved single neuron adaptive PID control intelligence AQM algorithm is presented,named GSNAPID.This improved algorithm makes the neuron proportional coefficient to auto-tuned online and modifies the part of weighting factor correction learning.The simulation results based on NS2 demonstrated that the GSNAPID algorithm has better convergence,can quickly converge the queue length near to expectations,and also has better stability and robustness in the complex network environment.Compared with other active queue management algorithms such as the traditional PID algorithm and the single neuron adaptive PID（SNAPID） algorithm,the GSNAPID algorithm is advantageous.