An improved adaptive active queue management scheme with fairness

In this paper, an improved adaptive active queue management scheme based on neuron gradient learning is presented, which integrates the flow identification mechanism of CHOKe with neuron-based adaptive AQM algorithm. This scheme proposed can identify both responsive and unresponsive flows, and simultaneously penalize unresponsive flows to provide fairness for all kinds of flows. On the other hand, the parameters of neuron-based AQM controller are tuned adaptively for time-varying load of networks, which can keep good performance on stability and robustness of queue dynamics at router. Finally, simulation results via NS-2 simulator show the effectiveness of the proposed scheme.