Fiscal Year	2021
Division	Research & Sponsored Programs
Department	NPS Naval Research Program
Investigator(s)	Ballard, Marcus A.
Sponsor	Office of the Chief of Naval Operations (Navy)
Summary	Wargames are an essential tool for education, training and formulation of strategy. They are especially important in the evaluation of threats from, and strategies against, trained adversaries who present significant risk to friendly forces. We propose to develop a wargame adversary trained to defeat the current strategy of friendly forces, thereby allowing the evaluation of alternate strategies against an intelligent, simulated opponent. We will investigate the use of deep neural network (DNN) algorithms to solve a constrained stochastic reward-collecting path problem. Agents from a friendly (blue) team and an adversarial (red) team will be placed within a discrete environment. The blue team will be challenged to obtain a reward by achieving a fixed goal using a pre-determined strategy. Then, reinforcement learning will be used to train the red team to overcome the blue team's current strategy. Having thus trained a competent red team, the blue team's strategy can be altered to evaluate the efficacy of new strategies. This research will seek to evaluate the ability of different DNN algorithms to train the red team against various blue team strategies, in terms of both efficacy and efficiency, and the resiliency of the trained red team to subsequent changes in blue team strategy. We anticipate the results of this research to be summarized in a research poster and executive summary, in addition to a presentation and full technical report deliverable to the Topic Sponsor.
Keywords
Publications	Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal
Data	Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal