Cases for Automating Realistic Threat Simulation (CARTS)

"The Missile Defense Agency does extensive system assessment based on modeling and simulation.However, it currently takes days or weeks to prepare high-fidelity threat data for each simulation run, with input from human experts consuming substantial time.To increase simulation frequency and flexibility, threat data should be generated more rapidlyideally, in-line during simulation runs. We propose to develop technology that uses Cases for Automating Realistic Threat Simulation (CARTS).CARTS will automate some of the work that experts currently do as part of threat data generation.We will apply and combine three methodologies in building CARTS: (1) knowledge acquisition, including structured interviews working through sample problems and their variants; (2) case based reasoning (CBR), to partition each expert decision space into a set of recognizable recurring challenges; and (3) behavior transition networks (BTNs), to cope with variation within each such case.During Phase I we will clarify requirements, constraints, and use cases; explore a range of expert inputs to high-fidelity data generation in order to identify the most promising areas for initial automation; develop algorithms and encode knowledge to automate selected decisions; demonstrate feasibility with a limited proof-of-concept prototype; and define a plan for Phase II."