A novel Iterative Optimization-based Simulation (IOS) framework: An effective tool to optimize system's performance
We describe a metaheuristic algorithm for simulation optimization. Traditionally, discrete event simulation optimization is carried out by multiple simulation runs executed <i>sequentially.</i> At the end of each simulation run, the run is evaluated (using model output - black box approach) by an objective function. If we carry out simulation runs <i>simultaneously</i>, then we can evaluate (using model internal data - white box approach) different simulation runs during their execution before the end is reached. Thus, we can eliminate the inferior runs early and allow only the most promising runs to continue to the end. We explore this parallel competition of simulation models on a single processor computer. Applications of the algorithm to traveling salesman and job shop scheduling problems are presented. In conclusion, our results suggest that the algorithm is a suitable approach for solving some combinatorial problems, and it represents a promising "nonsequential" avenue for simulation optimization.