A Self - Tuning Regulator for Adaptive Overload

The problem of designing eecient feedback overload control mechanisms has received considerable attention in the literature and it continues to be a major research area. In this paper, we interpret the design of such mechanisms as an adaptive feedback control problem. We illustrate how this analogy can be used to design congestion control mechanisms for high-speed wide-area networks. Speciically, we design a control mechanism that combines a least-squares parameter estimator with a minimum prediction error control law. Preliminary simulation results indicate that the controller yields high throughput and low loss probability even for large control delays and heavy overload conditions .