Traffic predictions have been demonstrated with the capability
to improve netwon efficieDey and QoS in broadband ATM networks.
Recent research shows that fuzzy logic prediction outpeforms conventional
autoregression predictions. The application of fuzzy logic also has a
potential to control traffic more effectively_ In this paper, we propose the use
of the fuzzy logic prediction on connection admission control (CAC) and congestion
control on high speed networks. We first modeled traffic charactoristies using an on-line fuzzy logic predictor on CAC. Simulation results show
that fuzzy logic prediction improves the efficiency of both conventiooal and
measurement-based CAC. In addition, the measurement-based approach incorporating
fuzzy logic inference and using fuzzy logic prediction is shown
to achieve higber netwok utilization while maintaining QoS. We then applied
the fuzzy logic predictor to cingestion control in which the ABR queue
is estimated one round-trip in advance. Simulation results show that the
fuzzy logic control scheme significantly reduces convergence time and overall
buffer requirements as compared with conventional schemes.