Parallel computation in the design and stochastic optimization of discrete event systems

Overviews the issue of constructing multiple sample paths of a stochastic discrete event dynamic system (DEDS) under changes in some parameter value or its actual structure in the context of parallel processing. The basic sample path constructability problem is formulated and some solution techniques are reviewed. Assuming that the ultimate objective of this parallelized sample path construction is to select good designs and to make good decisions, and, ultimately, to optimize the performance of a DEDS, the authors also outline the issues involved in using parallel computation for this purpose and describe some previously proposed techniques based on "ordinal" optimization.<
>