Project Highlights
    - Self-Scheduling in High Speed Data Networks

We have developed an efficient fuzzy logic predictor using optimal fuzzy logic system. This predictor is based on on-line measurements and uses the triangular membership functions. We have also implemented front ends using Matlab and Java based scripts. Based on extensive tests this algorithm performs extremely well and outperforms the traditional time series method. Regarding traffic scheduler, we have extended a dynamic bandwidth scheduler developed for two sources using fuzzy logic control. The new dynamic bandwidth scheduler can handle multiple traffic sources. We have also developed a refinement of the dynamic bandwidth scheduler which recursively and efficiently redistributes the unused bandwidth allocated by the original scheduler. The algorithm developed also provides an estimate of the line rate required to meet a set of QoS requirements imposed by the traffic sources. This will be useful in determining whether to accept or reject a new source. Extensive simulation studies indicate that the modified bandwidth allocation scheme performs extremely well. We emphasize the fact that the bandwidth allocation scheme we have developed does not assume any statistical properties of the traffic sources, and in particular adapts to drastic changes in traffic characteristics.

Regarding call admission control, we have designed a Fuzzy Connection Admission Control (FCAC) algorithm based on on-line measurements that decides whether to accept or reject a new source. This algorithm integrates two main criteria namely; the estimation of the bandwidth required to meet the set of QoS requirements imposed by the active sources and the actual use of the allotted bandwidth. We have introduced the notion of virtual line rate required and customer satisfaction factor to design the structure of the adaptive CAC. Limited experiments have shown that the FCAC algorithm performs extremely well when compared to algorithms based on effective bandwidth. The algorithm uses fuzzy logic theory. We have also developed of a graphical user interface and a simulator.