Design of an active queue management technique based on neural networks for congestion control
Sukant Kishoro Bisoy, Pratik Kumar Pandey, Bibudhendu Pati · 2017
In this work, an adaptive neural based active queue management (AQM) technique named ANB-AQM is proposed as network algorithm supporting TCP flows to control congestion and achieves quality of service (QoS). The proposed technique tunes the parameters based on self-learning to deal with the nonlinearity of network systems. It helps in predicting the future value of the packet drop probability by considering the past history of error in queue length. The parameters of the proposed technique adjusted online. The performance of the proposed technique is analyzed with existing AQM such as PI, IAPI and NNPI. The result shows that the proposed technique achieves stability with faster settling time (transient response). Moreover, it achieves smaller delay and smaller jitter than others.