An adaptive AIMD-type scheme for congestion control derived using virtual equilibria

The Chiu-Jain or additive-increase multiplicative-decrease (AIMD) congestion control algorithm is approached from a variable structure control perspective, leading to a better understanding and consequently a new choice of parameters of this classical algorithm. This new parameter choice, in the absence of delays, ensure that the users´ injection rates always attain a neighborhood of the fair, full-capacity equilibrium with enhanced smoothness and responsiveness, using only one bit of information, as well as knowledge of the fair share. If the fair share is unknown, an adaptive scheme that overcomes this lack of information is proposed. Simulations show that the proposed choice can result in convergence of the injection rates to a fair, full utilization equilibrium, while maintaining improved responsiveness and smoothness, in the presence of delays that are not too large.