Research on the Causes of Dynamical Instability in Combat Models.
Abstract
Combat models are nonlinear deterministic models of decision making processes that deal with attrition of opposing forces. They contain dynamical instabilities that destroy the robustness of a simulation and interfere with a simulation's consistency fidelity, reliability, transportability, and validity. Nonlinear behavior in a model's design is a primary source of instability. Striking effects of instabilities on the performance of a combat model are found in computer arithmetic that cause global divergences of simulation runs, branchings on thresholds set by decision tables that can be triggered inadvertently by small errors in the computation of state variables, and structurally unstable decision logic. Difficulties in reducing and eliminating structural instability and structural variance arise from the use of modeling paradigms. Several paradigms are distinguished: discrete event simulations, embedded dynamical systems, and models of computer arithmetic. A dynamical systems viewpoint illuminates properties of a model and its computer simulation that are essential for verification and validation. By viewing computational processes as dynamical systems that are embedded in discrete event simulations, fundamental difficulties for models of the real world are brought to light. This report discusses examples of unstable behavior, demonstrates ways to reduce or eliminate sources of instability, and suggests strategies for designing valid combat models that are consistent, robust, and stable.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1996
- Accession Number
- ADA315079
Entities
People
- Julian I. Palmore
Organizations
- Construction Engineering Research Laboratory