Title :
Analytic perturbation analysis of discrete event dynamic systems
Author :
Uryasev, Stanislav
Author_Institution :
Brookhaven Nat. Lab., Upton, NY, USA
Abstract :
This paper considers a new analytic perturbation analysis (APA) approach for discrete event dynamic systems (DEDS) with discontinuous sample-path functions with respect to control parameters. The performance functions for DEDS usually are formulated as mathematical expectations, which can be calculated only numerically. APA is based on new analytic formulas for the gradients of expectations of indicator functions, therefore, it is called an analytic perturbation analysis. The gradient of performance function may not coincide with the expectation of a gradient of sample-path function (i.e., the interchange formula for the gradient and expectation sign may not be valid). Estimates of gradients can be obtained with one simulation run of the models
Keywords :
discrete event systems; estimation theory; perturbation techniques; analytic perturbation analysis; discontinuous sample-path functions; discrete event dynamic systems; expectation gradients; gradient estimation; performance function; sample-path function; Control systems; Discrete event simulation; Discrete event systems; Finite difference methods; Laboratories; Network servers; Performance analysis; Routing; Stochastic processes; Stochastic systems;
Conference_Titel :
Computer Integrated Manufacturing and Automation Technology, 1994., Proceedings of the Fourth International Conference on
Conference_Location :
Troy, NY
Print_ISBN :
0-8186-6510-6
DOI :
10.1109/CIMAT.1994.389042