Adaptive Window Flow Control in MPLS Networks using Enhanced Kalman Filtering

Abstract

This paper presents an adaptive sliding window flow control protocol for MPLS networks, based on estimating the available link bandwidth using Kalman Filtering enhanced by bias estimation. An optimal control algorithm is then implemented that minimizes the variance of queue length deviations from the desired target. The simulation results show that, with bias estimation, the bandwidth estimate converges much faster than with ordinary Kalman filtering. We also achieve the goal of maximizing the bandwidth link utilization efficiency while minimizing the packet loss rate.