On-line optimal and adaptive control of a queuing system

A number of on-line control methods have been studied for the operational control of a queuing system. Time-series models have been used, in contrast to the probability models usual in the traditional approach to such problems. It is shown that most queuing processes can be formulated as multistage control problems to which modern control theory can be applied. The various control techniques applicable to a queuing system can be divided into two classes: decision and regulator control. In obtaining the control strategies, this thesis draws heavily from dynamic programming, least-squares estimation, the discrete maximum principle and gradient techniques. The uncertainties encountered in the queuing system can be overcome with an adaptive control method. The open-loop-feedback-optimal control technique has been stressed here due to its simplicity. Applications of the methods to various fields have also been studied. Extension of the method to long interval control is immediate in all the cases. Although the optimal control of a queuing system has been discussed, the methods are general enough to be applied to other areas.Applied Science, Faculty ofElectrical and Computer Engineering, Department ofGraduat