Numerical recipes in C (2nd ed.): the art of scientific computing
Numerical recipes in C (2nd ed.): the art of scientific computing
Alternative implementations of multitrajectory simulation
Proceedings of the 30th conference on Winter simulation
| Hi-index | 0.02 |
This paper investigates the dimensionality characteristics of the outcome space of a combat simulation. The independent state variables of all of the outcome states for a simulation run for given event management policies were analyzed using techniques based on Principal Component Analysis and Singular Value Decomposition, to give metrics for dimensionality. The number of dimensions in the outcome space is indicative of variety of possible outcomes, a property potentially important in hierarchical simulation. Events were managed using random choices, multitrajectory methods designed to give greater preference to high probability trajectories, and by various methods guided by analysis of the impact of the various events. The number of dimensions could not be increased greatly by the event management techniques used for selecting event outcomes for multitrajectory resolution. Beyond 1000 replication runs, the size of the state space did not even strongly influence the metrics.