摘要
基于神经网络理论中的神经元模型与学习算法,设计了一种主动队列管理算法SNAPI(Single Neuron-based Adaptive PI controller).控制器根据系统误差在线调整PI控制器的控制参数,以适应动态变化的网络参数.运用Nyquist稳定判据给出了系统在平衡点附近的局部稳定条件.最后通过仿真检验了SNAPI,并比较了它与使用固定控制参数的PI算法的性能.
Based on the neuronal model and learning algorithm of neural network theory, an active queue management (AQM) algorithm called SNAPI (Single Neuron-based Adaptive PI controller) is presented. SNAPI employs single neuron to adjust control parameters of the Proportional-Integral (PI) controller online according to the system error to adapt the dynamically changing network parameters. Using the Nyquist stability criterion, this paper gives the local stability conditions around the system equilibrium point. Finally, simulations are made to verify the SNAPI algorithm and to compare its performance with that of the fixed-parameter PI algorithm.
出处
《信息与控制》
CSCD
北大核心
2008年第5期565-570,575,共7页
Information and Control
关键词
拥塞控制
主动队列管理
PI控制器
单神经元
congestion control
active queue management (AQM)
PI controller
single neuron