Suboptimal congestion control laws for TCP-RED dynamics in computer networks

Abstract

Based on a dynamic model that simulates the behavior of TCP flows and active queue management (AQM) scheme two specific congestion avoidance modes, feedback control law and real time control law, have been presented respectively in the thesis. The control laws proposed monitor the status of buffers and multiplexor of the router and detect incipient congestion by sending warning signals to the sources. In our work, the AQM scheme can be modeled as a feedback control system governed by stochastic differential equations driven by a doubly stochastic point process with intensities being the controls. Besides, the AQM scheme can also be modeled as a real time control system in which the state of the router is monitored online and an optimal control must be executed in real time for each of sampled time intervals. Key Words: Feedback Control, Real time Control, RED, Dynamic Model, TCP, AQM.