Stochastic fluid models with upward jumps and phase transitions (Q6164834)

This work studies a numerical scheme to compute the steady state of a so-called Markov-modulated fluid flow. The model is the following: a reservoir is filled with a fluid whose level evolves linearily at a rate which depends on an underlying finite-space Markov chain; when the fluid level hits zero, it jumps to a fixed positive level, and the Markov chain also possibly jumps. The steady state of the joint process (fluid level \(+\) underlying chain) solves an explicit set of ordinary differential equations and can thus be expressed explicitly in terms of a finite number of values which solve a system of linear equations. However, evaluating these expressions leads to numerical problems due to some exponential terms. The authors propose an alternative formulation of the solutions which avoids this issue. The efficiency of the method is illustrated on some numerical examples.