Complex systems do not readily submit to reductionist analysis where an individual element is examined in isolation from other elements. This reductionist approach ignores the linkages that exist between system elements, and the fact that the linkages cannot be deduced from individual analysis. Complex systems exhibit a rich variety of behaviors, including many that are counter-intuitive. To capture these effects, complex systems should be analyzed from a holistic rather than a reductionist point of view.
Sandia has developed various models that utilize an agent-based approach along with discrete-event simulation, system-dynamics, and sequential-modular approaches. An agent is an autonomous, computational entity that can be viewed as perceiving and acting upon its environment and is autonomous in that its actions or behavior depend at least partially on its own experience. This simulation technology capitalizes on technological advances in evolutionary learning algorithms and massively parallel computing. Interactions among elements are modeled individually by intelligent software agents representative of real world decision-makers. The agent-based model provides a picture of systems that can be examined both computationally and analytically, offering new ways of experimenting with and theorizing about the impact of perturbation or shocks.
We have found that complex systems many times are best modeled using a hybrid approach that allows one to use agent-based, discrete-event, system dynamics, and sequential-modular approaches to represent various aspects of a model. With this motivation, we utilize a software systems development environment that allows such hybrid models to be constructed.
Research Areas
Adaptive Intelligence Aggregation, Mark Boslough Agent-Based Optimization, John Siirola Behavior-Aware Decision Support Systems, George Backus Economic Systems, George Backus Healthcare System Policy Analysis, George Backus and David Schoenwald Integrating Modeling and Simulation in Large Organizations, Laura McNamara Interacting Systems, David Schoenwald Logistics, David Schoenwald Hybrid Multi-Agent Simulation, John Siirola Robotic Behavior, Mark Boslough Soldier Behavior, Mark Boslough Verification and Validation Methodologies for Intelligence Applications, Laura McNamara
Presentations & Publications
2008 2007 2006 2005 2004 Pre-2004