Optimization of Polling Systems by Means of Gradient Methods and the Power-SeriesAlgorithm

摘要

The paper is devoted to numerical optimization of polling systems with a generalservice order table and with Bernoulli service disciplines at each visit of the server to a queue. The optimization is based on the use of the power-series algorithm (p.s.a.), a tool which can be applied for the numerical evaluation of performance measures for these models. The applicability of the p.s.a. is extended towards the numerical computation of first order partial derivatives with respect to the Bernoulli parameters. The extension strongly improves the efficiency of procedures for optimization purposes.