Real-time parametric analysis of discrete event system trajectories

For a Markovian discrete-event dynamic system (DEDS), an approach has been proposed for generating an event time sequence independently from the state updating process. In simulation environments, this allows one to efficiently construct state sequences of the system under various parameter values, using only a common nominal event time sequence. The authors show how to use this approach in a real-time setting by directly observing a trajectory of a DEDS. They apply this approach to the crash time prediction problem, where they are interested in estimating the time when an M/M/1 system first reaches a given queue length as a function of different arrival rates.<
>