Skip to Main Content
One of the major challenges nowadays when managing IP networks is to guarantee proper quality of service by using network infrastructure in an optimized way. One of the recently proposed solutions is the so called traffic engineering with MPLS. In this paper, we propose a novel approach for adaptive fuzzy model-based flow control in MPLS networks. Current research treats MPLS as the Internet's solution to high performance network. A nonlinear predictive controller is designed on the basis of a Takagi-Sugeno fuzzy model. By online adaptation of the fuzzy model, high control performance can be achieved for network traffic processes multiplexed in a single buffer server. Our developed fuzzy predictor consists of a two-step algorithm: an adaptive training with covariance resetting and a gradient-based learning algorithm for refining the first part of the learning procedure. The effectiveness and online applicability of the proposed approach are demonstrated by simulations using real network traffic traces. In fact, the proposed approach is capable of providing better system performance than some existing flow control schemes.