Compensators and derivative estimation for queueing systems
- 1 January 1990
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 173-175 vol.1
- https://doi.org/10.1109/cdc.1990.203570
Abstract
A fairly general method for sensitivity analysis of simulations is proposed involving the use of compensator identities. The case of a GI/GI/1 queue is discussed in some detail. Expressions for the derivatives of state probabilities that provide direct simulation estimates are given for single class closed queueing networks with Markovian routing. The method uses compensator identities in conjunction with infinitesimal perturbation analysis (IPA) techniques to provide low variance unbiased estimates at the expense of additional computational requirements.Keywords
This publication has 11 references indexed in Scilit:
- Sensitivity Analysis for Simulations via Likelihood RatiosOperations Research, 1989
- Sensitivity Analysis and Performance Extrapolation for Computer Simulation ModelsOperations Research, 1989
- Derivative estimates from discontinuous realizations: smoothing techniquesPublished by Association for Computing Machinery (ACM) ,1989
- Replication Schemes For Limiting ExpectationsProbability in the Engineering and Informational Sciences, 1989
- Perturbation Analysis Gives Strongly Consistent Sensitivity Estimates for the M/G/1 QueueManagement Science, 1988
- Smoothed (conditional) perturbation analysis of discrete event dynamical systemsIEEE Transactions on Automatic Control, 1987
- Optimization of stochastic systemsPublished by Association for Computing Machinery (ACM) ,1986
- Convergence of parameter sensitivity estimates in a stochastic experimentIEEE Transactions on Automatic Control, 1985
- Perturbation analysis and optimization of queueing networksJournal of Optimization Theory and Applications, 1983
- Point Processes and QueuesPublished by Springer Nature ,1981