On the construction of sample paths of discrete event systems in parallel

Considers the problem of using observations of a given discrete event system for constructing sample paths of alternative discrete event systems. The author formulates the problem in the setting of generalized semi-Markov schemes. The author views the system as an input-output system, where the input is a sequence of event lifetimes, and the output is the resulting sequence of states, events, and transition epochs. The system is observed via an observation map. The author addresses the problem from the point of view of inversion, defined as the extraction of event lifetimes from observations of the output. The extracted event lifetimes are used to drive parallel simulations of alternative systems, thereby generating multiple sample paths simultaneously. The author discusses necessary and sufficient conditions for invertibility of the system. The author also considers the problem of implementing his inversion algorithm in parallel, by decomposition of the state transition structure