MACHINE LEARNING OF AN APPROXIMATE MORPHISM OF AN ELECTRONIC WARFARE SIMULATION COMPONENT

摘要

Electromagnetic waveforms are an essential component of radar and electronic warfare digital computer simulations. Sampled representations of radar waveforms are widely used in high-fidelity applications for their physical realism and suitability for algorithmic processing. However, this fidelity comes at a price because operations on sampled representations are often a computationally costly simulation bottleneck. In this paper, we propose a generic framework for constructing a reduced, feature-based model component derived from a given high-fidelity component, and demonstrate the approach on a simplified radar waveform model. The reduced model is related to the original through an approximate morphism. Both supervised and unsupervised machine learning are key components of the framework.