WSC 2004 Final Abstracts
Monday 10:30:00 AM 12:00:00 PM
Chair: Jennie La (University of Calgary)
Bayesian Methods for Discrete Event Simulation
Stephen E. Chick (INSEAD)
Bayesian methods are now used in a variety of ways in discrete-event simulation. Applications include input modeling, response surface modeling, uncertainty analysis, and experimental designs for field data collection, selection procedures, and response surface estimation. This paper reviews some fundamental concepts of subjective probability and Bayesian statistics that have led to results in simulation applications.
Monday 1:30:00 PM 3:00:00 PM
Stochastic Petri Nets
Chair: David Munoz (ITAM)
Stochastic Petri Nets for Modelling and Simulation
Peter J. Haas (IBM )
Stochastic Petri nets (SPNs) have proven to be a powerful and enduring graphically-oriented framework for modelling and performance analysis of complex systems. This tutorial focuses on the use of SPNs in discrete-event simulation. After describing the basic SPN building blocks and discussing the modelling power of the formalism, we present elements of a steady-state simulation theory for SPNs. Specifically, we provide conditions on the SPN building blocks that ensure long-run stability for the underlying marking process (or for a sequence of delays determined by the marking process) and the validity of estimation procedures such as the regenerative method, the method of batch means, and spectral methods.
Monday 3:30:00 PM 5:00:00 PM
Kriging Interpolation in Simulation
Chair: Natalie Steiger (University of Maine)
Kriging Interpolation in Simulation: A Survey
Wim C.M. Van Beers and Jack P.C. Kleijnen (Tilburg University)
Many simulation experiments require much computer time, so they necessitate interpolation for sensitivity analysis and optimization. The interpolating functions are ‘metamodels’ (or ‘response surfaces’) of the underlying simulation models. Classic methods combine low-order polynomial regression analysis with fractional factorial designs. Modern Kriging provides ‘exact’ interpolation, i.e., predicted output values at inputs already observed equal the simulated output values. Such interpolation is attractive in deterministic simulation, and is often applied in Computer Aided Engineering. In discrete-event simulation, however, Kriging has just started. Methodologically, a Kriging metamodel covers the whole experimental area; i.e., it is global (not local). Kriging often gives better global predictions than regression analysis. Technically, Kriging gives more weight to ‘neighboring’ observations. To estimate the Kriging metamodel, space filling designs are used; for example, Latin Hypercube Sampling (LHS). This paper also presents novel, customized (application driven) sequential designs based on cross-validation and bootstrapping.
Tuesday 8:30:00 AM 10:00:00 AM
Verification, Validation, and Accreditation
Chair: Young Lee (IBM)
Quality Assessment, Verification, and Validation of Modeling and Simulation Applications
Osman Balci (Virginia Tech)
Many different types of modeling and simulation (M&S) applications are used in dozens of disciplines under diverse objectives including acquisition, analysis, education, entertainment, research, and training. M&S application verification and validation (V&V) are conducted to assess mainly the accuracy, which is one of many indicators affecting the M&S application quality. Much higher confidence can be achieved in accuracy if a quality-centered approach is used. This paper presents a quality model for assessing the quality of large-scale complex M&S applications as integrated with V&V. The guidelines provided herein should be useful for assessing the overall quality of an M&S application.
Tuesday 10:30:00 AM 12:00:00 PM
Network Traffic Modeling
Chair: John Charnes (University of Kansas)
More ``Normal'' Than Normal: Scaling Distributions and Complex Systems
Walter Willinger (AT&T Labs-Research) and David Alderson, John C. Doyle, and Lun Li (California Institute of Technology)
One feature of many naturally occurring or engineered complex systems is tremendous variability in event sizes. To account for it, the behavior of these systems is often described using power law relationships or scaling distributions, which tend to be viewed as ``exotic'' because of their unusual properties (e.g., infinite moments). An alternate view is based on mathematical, statistical, and data-analytic arguments and suggests that scaling distributions should be viewed as "more normal than Normal''. In support of this latter view that has been advocated by Mandelbrot for the last 40 years, we review in this paper some relevant results from probability theory and illustrate a powerful statistical approach for deciding whether the variability associated with observed event sizes is consistent with an underlying Gaussian-type (finite variance) or scaling-type (infinite variance) distribution. We contrast this approach with traditional model fitting techniques and discuss its implications for future modeling of complex systems.
Tuesday 1:30:00 PM 3:00:00 PM
Inside Simulation Software
Chair: K. Preston White (University of Virginia)
Inside Discrete-Event Simulation Software: How it Works and Why it Matters
Thomas J. Schriber (University of Michigan) and Daniel T. Brunner (Systemflow Simulations, Inc.)
This paper provides simulation practitioners and consumers with a grounding in how discrete-event simulation software works. Topics include discrete-event systems; entities, resources, control elements and operations; simulation runs; entity states; entity lists; and entity-list management. The implementation of these generic ideas in AutoMod, SLX, and Extend is described. The paper concludes with several examples of “why it matters” for modelers to know how their simulation software works, including coverage of SIMAN (Arena), ProModel, and GPSS/H as well as the other three tools.
Tuesday 3:30:00 PM 5:00:00 PM
Chair: Nilay Argon (University of Wisconsin, Madison)
Dependence Modeling for Stochastic Simulation
Bahar Biller (Carnegie Mellon University) and Soumyadip Ghosh (IBM )
An important step in designing stochastic simulation is modeling the uncertainty in the input environment of the system being studied. Obtaining a reasonable representation of this uncertainty can be challenging in the presence of dependencies in the input process. This tutorial attempts to provide a coherent narrative of the central principles that underlie methods that aim to model and sample a wide variety of dependent input processes.
Wednesday 8:30:00 AM 10:00:00 AM
Chair: Andrew Seila (University of Georgia)
Simulation Output Analysis: A Tutorial Based on One Research Thread
Bruce W. Schmeiser (Purdue University)
In this tutorial we discuss simulation output analysis: the problem of evaluating and reporting the quality of a given stochastic simulation experiment. We advocate the use of micro/macro replications based on fixed sample sizes, batch sizes based on mean squared error, and avoiding confidence intervals and analysis of variance. In addition, we discuss the problem of evaluating and comparing confidence-interval procedures and the issue of how to report point estimates concisely.
Wednesday 10:30:00 AM 12:00:00 PM
Military Applications of Agent-based Models
Chair: Bill Biles (University of Louisville)
Military Applications of Agent-Based Simulations
Thomas M. Cioppa, Thomas W. Lucas, and Susan M. Sanchez (Naval Postgraduate School)
Navy personnel use the REMUS unmanned underwater vehicle to search for submerged objects. Navigation inac-curacies lead to errors in predicting the location of objects and thus increase post-mission search times for explosive ordnance disposal teams. This paper explores components of navigation inaccuracy using discrete event simulation to model the vehicle’s navigation system and operational per-formance. The simulation generates data used, in turn, to build statistical models of the probability of detection, the mean location offset given that detection occurs, and the location error distribution. Together, these three models enable operators to explore the impact of various inputs prior to programming the vehicle, thus allowing them to choose combinations of vehicle parameters that reduce the offset error between the reported and actual locations.