STELLAR SCIENCE LTD. CO. — Department of Defense SBIR Phase I: AF181-063

Description

The challenge is to reduce the effort of integrating sensing component models into System of System (SoS) simulations by streamlining integration of component models into a master Simulation Environment (SIMEN) which can be used to conduct SoS studies.Stellar Science has previously developed and fielded a number of government-owned sensing model integration systems based on modular architectures, enabling component models to be readily added and removed. We propose to design and prototype Varied Interface & Phenomenology Engineering Relationship Suite (VIPERS), an environment for SoS simulation that leverages these existing investments.Component integration will be automated using the Machine Learning technique of Surrogate Modeling (SM), which constructs high-fidelity emulations from model data. SM enables 1) very fast execution; 2) unified software interfaces; and 3) a universal approach for replacing complex models in integrated simulations. Our approach uses intelligent closed-loop adaptive sampling methods, enabling the construction of high-fidelity SMs with reduced computation and analytic effort. SM takes real-world experimental data into account more readily than traditional component models, with appropriate weighting of confidence and uncertainty. SMs provide universal, data-driven models that capture highly detailed trends with accurate prediction power and known uncertainties, while simultaneously eliminating most integration overhead.