基于独立神经元的自适应主动队列管理算法

Adaptive AQM Algorithm Based on Independent Neurons

针对主动队列管理中PI(Proportional-integral)算法的不足,设计了一种基于独立神经元的自适应PI控制器INAPI(Independent neurons-based adaptive PI controller)。控制器利用神经网络理论中的神经元模型与学习算法,2个独立的神经元根据系统状态采用最速下降法在线调整PI控制器的控制参数,以适应动态变化的网络参数。仿真结果表明,INAPI的性能要优于使用固定控制参数的PI和FLC(Fuzzy logic controller)算法。

To improve the performance of PI （Proportional-integral）, a new AQM （Active queue management） algorithm, called the INAPI （Independent neurons-based adaptive PI controller）,is presented. INAPI uses the neuronal model and learning algorithm of neural network theory, two auto-tuning neurons by steepest descent method to adjust the control parameters of the PI controller according to the system state and to adapt to the dynamic varieties of network parameters. Simulation results show that the algorithm has a better performance than the fixed parameter controller PI and FLC（Fuzzy logic controller） in a large-scale network.