Parallel numerical solutions of a class of Volterra integro-differential equations

Publication date 2 February 1998

discrete-event stochastic system generalized semi-Markov process Markov chain multi-dimensional integrals parallel computation system of Volterra integral equations Volterra integro-differential equations

Mathematics Subject Classification ID

Systems of nonsingular linear integral equations (45F05) Integro-ordinary differential equations (45J05) Continuous-time Markov processes on discrete state spaces (60J27) Probabilistic methods, stochastic differential equations (65C99) Numerical methods for integral equations (65R20) Parallel numerical computation (65Y05)

Review text

A discrete-event stochastic system makes state transitions when events associated with the occupied state occur; events occur only at an increasing sequence of random times. The underlying stochastic process of a discrete-event system has to record the state of the system as it evolves over continuous time. The generalized semi-Markov process provides a framework for specifying and analyzing stochastic processes underlying most kinds of discrete-event systems. A generalized semi-Markov process is a continuous-time stochastic process that makes a state transition when one or more ``events associated with the occupied state occur. In previous work, we introduced a numerical method for computing limiting distributions for finite-state generalized semi-Markov processes with exponential and deterministic events. The method rests on observation, at equidistant time points, of the continuous time Markov process and analysis of the resulting general state space Markov chain (GSSMC). This embedded GSSMC and the continuous time Markov process have the same limiting distribution, provided that limits for the Markov process exist. In general, the method for steady state analysis of generalized semi-Markov processes with exponential and deterministic events requires the numerical solution of a system of Volterra integral equations of the second type which comprises multi-dimensional integrals. While a considerable amount of results for the efficient numerical solution of such Volterra equations with one-dimensional integrals is available, to the best of our knowledge, the numerical solution of such Volterra integral equations with multi-dimensional integrals and their parallelization has not been considered yet. In this paper, we describe a parallel implementation for the numerical solution of the stationary equation of the GSSMC which allows their efficient numerical solution on a cluster of modern workstations or a multi-processor system. The key observation is that we can employ a wavefront approach for solving the set of systems of linear equations that result from discretizing the integral expressions in the Volterra equations by an appropriate quadrature rule. For the entire collection see [Zbl 0870.00040].

Recommendations

Cited in

(8)

This page was built for publication: Parallel numerical solutions of a class of Volterra integro-differential equations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4357557)