Adaptive Congestion Control of High Speed ATM Networks

Proportional control methods of controlling congestion in high speed ATM networks fail to achieve the desired performance due to the action delays, nonlinearities, and uncertainties in control loop. In this paper an adaptive minimum variance controller is proposed to minimize the rate of stochastic inputs from uncontrollable high priority sources. This method avoids the computations needed for pole placement design of the minimum variance controller, and utilizes an online recursive least squares algorithm in direct tuning of the controller parameters. The closed loop system is adaptive and robust to the uncertain network conditions and provides minimum cell loss ratio, efficient use of network resources, and fair allocation of the available bandwidth through the controlled sources.