Active Queue Management Algorithm Based on RBF Neural Network Controller

摘要

Considering the problem of input saturation and UDP flows, the TCP nonlinear dynamic model is improved. On account of the nonlinearity, uncertainty and time-variability of the improved model, the nonlinear RBF neural network controller is designed by using the neural network control strategy. In simulation experiments, the effectiveness and anti-interference of the active queue management algorithm based on RBF neural network controller are verified and the controller’s superiority is shown in the comparison simulation with the active queue management algorithm based on PID controller.