RESOURCE ALLOCATION UNDER UNCERTAINTY VIA STOCHASTIC CONTROL METHODS.

Abstract

For a class of resource allocation problems under uncertainty, a sizable improvement in performance or saving in resources can be achieved when the allocation is made using stochastic control methods that have the closed-loop property. This property consists of the following: at every stage the decision is done using the past measurements on the results of the previous actions and the statistics of the future measurements. In this way the present decision takes into account the value of future information. A suboptimal algorithm having the above property and modest computational requirements has been derived, and its effectiveness shown in an example.