A finite algorithm for \(\epsilon\)-optimal solutions of adaptive queueing control (Q1099591)

A method for the solution of adaptive queueing control is proposed. It is obtained by generalization of the non-adaptive approach and backward induction introduced by \textit{L. C. Thomas} in order to get an \(\epsilon\)- optimal solution for the problem of adaptive inventory control [A finite algorithm for \(\epsilon\)-optimal solutions of the infinite horizon adaptive inventory model, Notes in Decision Theory No.71, Department of Decision Theory, University of Manchester (1979]. To overcome the difficulty of unbounded cost the author uses a weighted supremum norm due to \textit{J. Wessels} [J. Math. Anal. Appl. 58, 326-335 (1977; Zbl 0354.90087)]. The relationship between the adaptive queueing control model and the non- adaptive one is discussed. A finite algorithm for calculating the \(\epsilon\)-optimal cost and for determining a \(2\epsilon\)-optimal policy is given. A formula allowing to calculate the number of backward induction steps is established. Numerical results are given. The paper is restricted to the case of Bernoulli queue but the author is stating that the ``the idea and method can be generalized in more general queueing control problem''.