Browse by author
Lookup NU author(s): Professor Kevin Glazebrook
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
When controlled stochastic systems have performances which satisfy generalised conservation laws (GCL), an objective which is linear in the performance is optimised by a Gittins index policy. We develop measures of the extent to which a system fails to satisfy GCL and derive suboptimality bounds for suitable index policies in terms of such measures. These bounds are used, inter alia, to explore the robustness in performance of cμ-type rules for a multiclass G/G/1 queueing system to departures from an assumption of exponential service times. We also study Gittins index policies for parallel processor versions of the classical undiscounted and discounted multi-armed bandit problems. In the undiscounted case, the cost of an index policy comes within a constant of the optimal cost - this constant being independent of the number of projects submitted for scheduling. In the discounted case, under fairly mild conditions, Gittins index policies come within an O(1) quantity of optimality and are hence average reward optimal when the discount rate is small enough.
Author(s): Glazebrook KD, Garbe R
Publication type: Article
Publication status: Published
Journal: Annals of Operations Research
Year: 1999
Volume: 92
Issue: 0
Pages: 19-43
Print publication date: 01/01/1999
ISSN (print): 0254-5330
ISSN (electronic): 1572-9338
Publisher: Springer New York LLC
URL: http://dx.doi.org/10.1023/A:1018934714800
DOI: 10.1023/A:1018934714800
Altmetrics provided by Altmetric