Extended perturbation analysis of discrete event dynamical systems
Shu Li · 1988
This thesis provides a further attempt to develop concepts and results from a control theoretic view in the area of Discrete Event Dynamical Systems. The basic approach is the recent technique of Perturbation Analysis. Motivated by the inability to deal with many different classes of problems by early versions of perturbation analysis, the thesis presents a new method in the light of perturbation analysis, which considerably extends the technique of perturbation analysis to general discrete event dynamical systems. Central to the thesis is the new idea of the so called invariance under cut-and-paste property of Markov state. Inspired by this idea, many different versions of new perturbation algorithms are proposed. Both theoretical proofs and extensive experimental results are given to validate the legitimacy of the new theory. The basic model used in the thesis is the Generalized Semi-Markov Process. State space descriptions, sample path analysis, as well as general extended perturbation analysis algorithms are given for such a general model of discrete event dynamical systems.