Artificial Intelligence for Constructing Accurate, Low-Cost Models and Simulations
Abstract
Modeling and Simulation is an important tool in the development of the highly effective weapons systems built by the United States and its allies. However, recent initiatives to reduce the cost of weapon systems through expanded use of modeling and simulation during the development process have not always lived up to expectations. Current practice in the construction of models and simulations primarily uses a manual implementation of equations to describe the entity being modeled. After verifying correct operation, these models are then validated by comparing them to data from real world tests to insure accuracy. These equation-based models require extensive time and money in order to construct high fidelity models that accurately represent the real world. Our research explores an alternate method of creating accurate models and simulations that can be done rapidly and at much lower cost. This approach uses hybrid artificial intelligence to create the models and simulations directly from validation data sets. Test results using this method of modeling militarily representative systems such as wing lift, radar, and Forward Looking Infrared "FLIR" demonstrated a reduction of over 90% in human labor required to create the models while simultaneously achieving approximately 70% better accuracy as compared to equation-based models prior to validation. Because this method builds the models from a data set, the method can be used to construct models of activities such as human decision-making that cannot be described using an equation-based approach. Additionally, the research demonstrated that models created using this method could be fully integrated with existing equation-based models.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2005
- Accession Number
- ADA484157
Entities
People
- David P. Brown
- Richard A. Mohler
Organizations
- Defense Acquisition University