Evaluation of the mean cycle time in stochastic discrete event dynamic systems

TL;DR

This paper analyzes the mean cycle time in stochastic discrete event dynamic systems modeled with linear equations over an idempotent semiring, providing solutions under various probabilistic assumptions on matrix entries.

Contribution

It extends previous results by deriving formulas for mean cycle time with different exponential distributions and zero entries in the system matrices.

Findings

Explicit formulas for systems with independent exponential entries.

Extension to matrices with non-identical exponential distributions.

Solutions for systems with a single random entry and deterministic others.

Abstract

We consider stochastic discrete event dynamic systems that have time evolution represented with two-dimensional state vectors through a vector equation that is linear in terms of an idempotent semiring. The state transitions are governed by second-order random matrices that are assumed to be independent and identically distributed. The problem of interest is to evaluate the mean growth rate of state vector, which is also referred to as the mean cycle time of the system, under various assumptions on the matrix entries. We give an overview of early results including a solution for systems determined by matrices with independent entries having a common exponential distribution. It is shown how to extend the result to the cases when the entries have different exponential distributions and when some of the entries are replaced by zero. Finally, the mean cycle time is calculated for systems with matrices that have one random entry, whereas the other entries in the matrices can be arbitrary nonnegative and zero constants. The random entry is always assumed to have exponential distribution except for one case of a matrix with zero row when the particular form of the matrix makes it possible to obtain a solution that does not rely on exponential distribution assumptions.