Application of Kalman filter in high-speed networks

Abstract

The problems of traffic management and control in multiple-service networks are subjects of considerable interest. We investigate traffic observation and prediction techniques in high speed heterogeneous services networks, which carry large traffic volumes and have large delay-bandwidth products. Such techniques are necessary for feedback flow control. The state dynamics and network perturbations are modeled for the feedback flow control application. Models are linear and derived from assumptions of classical queueing theory. Kalman filtering is used for traffic estimation. Estimation errors are obtained through computation and simulation.