Adversarial Modeling for DBM Planning (AMDP)

We propose a Phase II effort to extend the game-theoretic technology to apply to the DARPA Distributed Battle Management (DBM) program. Our proposed AMDP (Adversarial Modeling for DBM Planning) system augments DBMs capabilities with a predictive model of adversarial uncertainty and action.We seek to answer the question How can we predict enemy responses and positions, and update those predictions based on incoming dataAMDP uses models of various aspects of enemy COAs including decoys and mobile units and various related uncertainties including: enemy positions, asset types, and available actions in a SEAD-type (Suppression of Enemy Air Defenses) mission against ground targets.AMDP fuses prior, incoming, and predicted world knowledge to generate courses of action that maximize informational gain metrics, and thus minimize uncertainty, in order to ensure correct actions and protect against worst-case outcomes due to unexpected events.The targeted end result is a simulation assessment showing improved mission efficacy (measured by speed and accuracy in identifying and engaging true targets in a field of decoys) in high-uncertainty scenarios, within a third-party air battle simulation environment such as AFSIM or SIMAF.