Dynamic computing budget allocation for design ranking in stochastic complex systems
Simulation has been widely used to evaluate and optimize stochastic complex systems such as manufacturing,telecommunication,and healthcare systems,which are known as cyber-physical systems.However,research on design ranking in simulation optimization is limited.This research considers the complete ranking of k designs,whose performance can only be evaluated through simulation.Using the Bayesian framework,an optimal dynamic computing budget allocation model is developed,which maximizes the posterior probability of correct rankings.The computing budget allocation model is reformulated as a stochastic dynamic programming problem.This paper derives the optimal dynamic computing budget allocation rule and confirms its asymptotic optimal property.Numerical experiments and a case study demonstrate that the proposed simulation procedure considerably increased the simulation optimization efficiency for design ranking problems.
simulation optimizationranking and selectioncyber-physical systemMarkov decision processBayesian estimation