The application of fuzzy logic prediction in congestion control and its neural network implementation

This paper discusses a congestion avoidance and control scheme based on the application of fuzzy logic theory and its neural network implementation. It is mainly concerned with high-speed wide area networks where propagation delays can have significant effects on closed-loop traffic control. In order to overcome the detrimental effects, a fuzzy logic predictor is proposed at the switching node to estimate the queue length in advance. This information together with current queue length and the growth rate is fed into a fuzzy inference system for the generation of a traffic rate factor. This factor can be used alone or in conjunction with other schemes such as explicit rate indication for congestion avoidance (ERICA) to calculate available bit rate (ABR) traffic bandwidth allocation, and ultimately affects the explicit rate (ER) field in backward resource management (BRM) cells. This paper also discusses on the neural network implementation of the fuzzy predictor. This will greatly reduce the amount of computation while maintaining high prediction accuracy. Simulation results indicate that the overall quality of service (QoS) is also comparable with the original fuzzy logic predictor