A neuron-based active queue management scheme for internet congestion control

To deal with nonlinear and complex problems of internet congestion control, an intelligent scheme is required, which can learn the traffic pattern of the network. In this paper, we design a robust AQM scheme called neuron-based AQM (N-AQM) to efficiently control the complex network congestion problem and achieve QoS. In N-AQM, a neural network is used to predict the future value of current queue length and estimate the differential queue length error and use it to define the packet drop probability. Our simulation result demonstrates that N-AQM is stable, robust and outperforms other AQM schemes. From the result section, it is observed that N-AQM is more efficient in stabilising the queue length around the target with faster settling time and incurs lower oscillation than others. Copyright © 2020 Inderscience Enterprises Ltd.