A New Active Queue Management Algorithm based on Neural Networks PI
| Authors: | Waskasi Mojtaba Yaghoubi, University of Tehran, Iran (Islamic Republic of)
Yazdanpanah Mohammad Javad, University of Tehran, Iran (Islamic Republic of)
Yazdani Nasser, University of Tehran, Iran (Islamic Republic of) |
|---|
| Topic: | 3.3 Computers and Telematics |
|---|
| Session: | Telematics Applications and Networking |
|---|
| Keywords: | Neural Networks, Queuing Network Models, Nonlinear Systems, Computer Networks, PI Controllers. |
|---|
Abstract
Active Queue Management (AQM) applies a suitable control policy upon detecting congestion in networks. In this paper, an adaptive Proportional-Integral (PI) controller based on Artificial Neural Networks (ANN) is applied to AQM for the objective of congestion avoidance and control in middle nodes. The proposed controller is simple and can be easily implemented in high-speed routers. Neural Network PI (NNPI) dynamically adapts its parameters with respect to changes in the system. It is anticipated that this results in better response compared to linear controllers due to the nonlinear nature of NNPI. We simulated our method in ns2 and compared its performance with the conventional PI controller. The simulation results show NNPI yields better performance.
