A simple technique in Markovian control with applications to resource allocation to resource allocation in communication networks (Q1085077): Difference between revisions

Certain problems occurring in optimal sharing of memory among processors or in optimal allocation servers with blocking can be seen as instances of a general semi-Markov decision problem with exponentially distributed transition times. The paper considers an improved value-iteration algorithm for a discrete semi-Markov decision model. The idea is to create sparse transition matrices by introducing fictitious decision epochs. Numerical experiments are also presented for practical applications with a large number of states.