摘要

Internet congestion control system is complex, uncertainty and nonlinear. An Improved PID Neural Network (I-PID-NN) controller with changing integration rate and incomplete derivation in hidden layer is applied in active queue management (AQM). The adjustment of neural network parameters are implemented by using gradient algorithm as learning the rules, and the probability of packet loss can achieve adaptation. The performance of the I-PID-NN controller is verified by NS2 simulation results, which shows higher robustness and link utilization than PID-NN.