Stochastic optimization algorithms for regenerative DEDS

In this paper we are concerned with the application of stochastic programming techniques for the parametric optimization and analysis of Discrete Events Dynamic Systems (DEDSs). In particular, we consider optimization of steady state behavior of DEDSs which sample path depends discontinuosly on control parameters. This is a difficult problem, since in order to evaluate the steady state performance measure it is necessary to make an observation on the infinite time horizon, which is impossible. On the other hand, in the general case any observation on the finite time interval gives a biased estimates of the performance measure. Here we consider an important subclass of DEDSs for which we developed a new algorithm for optimizing their steady state performance, by adapting the methods of stochastic optimization. Such subclass is the set of regenerative DEDSs