A predictive fuzzy logic congestion avoidance scheme

Abstract

A congestion avoidance and control scheme that makes application of fuzzy set theory is described. The fuzzy logic predictor is proposed to estimate the output queue length. This information together with current queue length and growth rate is provided to a fuzzy inference system which generates a rate factor. This factor can be used alone or in conjunction with other algorithms such as ERICA to calculate ABR traffic bandwidth allocation and is ultimately influential in modifying the ER field in BRM cells. Simulation results indicate that the incorporation of a predictor reduces cell loss, queueing delay and delay variation. They further show that a fuzzy logic predictor works better than a traditional autoregression predictor. With reliable traffic prediction, congestion control algorithms are more tolerant to round-trip delay and the necessity to implement virtual source/virtual destination (VS/VD) at ATM switches is alleviated.