Self-tuning Optimal PI Rate Controller for End-to-End Congestion With LQR Approach

摘要

This paper presents a self-tuning optimal PI (Proportional-Integral) rate controller for end-to-end congestion in the IP-based Internet. We employ LQR (Linear Quadratic Regulator) approach from modern control theory in the optimal controller design that would allow the user to achieve good transient performance of IP networks by selecting proper positive definite matrices Q and R. Self-tuning PI rate controller self-tunes only when the changes in network traffic have drifted the network monitoring parameter outside its specified interval. The PI rate controller is located in each router and can calculate a desirable advertised source window size (i.e. source sending rate) based on the instantaneous queue length of the buffer. Our OPNET simulations demonstrate that our network traffic control algorithm can provide the network with a good transient behavior in AQM and flow throughput.