Computational issues for accessibility in discrete event simulation

Abstract

Several simulation model building and analysis issues have been studied using a computational complexity approach. More specifically, four problems related to simulation model building and analysis (accessibility of states, ordering of events, noninterchangeability of model implementations, and execution stalling) have been shown to be NP-hard search problems. These results imply that it is unlikely that a polynomial-time algorithm can be devised to verify structural properties of discrete event simulation models, unless P = NP. This article considers the problem of accessibility, identifies special cases that are polynomially solvable or remain NP-hard, and discusses implications with respect to the other three problems. A local search procedure and variations of simulated annealing are presented to address accessibility. Computational results illustrate these heuristics and demonstrate their strengths and limitations.