ARTIFICIAL NEURAL NETWORKS AS META-MODELS OF COMBAT PROCESSES - APPLICATIONS TO LINE-OF-SIGHT COMPUTATIONS

Modeling high resolution combat is a computationally intensive activit y that often requires compromise in the completeness or fidelity of th e model to accommodate existing computer technology. This trade-off wi ll always be necessary, but implicit modeling of some processes can re duce the computational load at run time so CPU cycles may be devoted t o other areas of the model. Unfortunately some costly processes, such as intervisibility calculations, are even more expensive (in terms of storage) to model implicitly. This paper examines the potential of art ificial neural networks to serve as efficient meta-models for line-of- sight determination.