Estimation of Delays in Non-Regenerative Discrete-Event Stochastic Systems P e t e r J. H a a s IBM Almaden Research Center San Jose, CA peterh@almaden.ibm.com 1 The Estimation Problem analytically or numerically, but must be estimated using simulation. This paper deals with estimation methods for timeaverage limits. A delay in a GSMP usually is computed as the length of a corresponding "delay interval" (that is, a random time interval) whose start (left endpoin0 and termination (fight endpoin0 each coincide with a state-transition epoch. We focus on sequences of delays that are specified using the method of"start vectors" introduced by Haas and Shedler [7]. This method uses a recursively-generated sequence o f random vectors to provide the link between the starts and terminations of individual delay intervals. Heuristically, the nth such "start vector" records the starts of delay intervals for all ongoing delays and newly started delays at the nth state transition. The values of the starts and the order of the starts in the start vector together summarize the history of the GSMP and comprise sufficient information to measure individual lengths. At each state transition, the current time may be inserted, starts may be deleted, and the
/lp/association-for-computing-machinery/estimation-of-delays-in-non-regenerative-discrete-event-stochastic-A7V3jQiImw