A Novel Flow Controller Design for High-Speed Communication Networks

A new approach to ow control in high speed communication network is proposed where the problem of ow control within a virtual circuit is modeled as a dynamic system with time delays and disturbances. In the proposed ow controller design, we apply: (1) a modiied Smith predictor to predict the backlog in each channel along a virtual circuit; (2) fuzzy CMAC neural networks to predict the time-varying feedback loop delays; (3) fuzzy CMAC neural networks to predict the channel disturbances along a virtual circuit; and (4) a mixed type of control law to improve the delay-through characteristics. The main advantage of the new design is that it can assure stability of the system as well as maintaining a certain throughput of the communication channel. The performance of the proposed ow controller design has been accessed by simulation and compared with existing designs. Simulation results show that better performance (i.e., higher throughput and/or lower delay) can be achieved by the proposed ow controller design.