Contested logistics simulation output analysis with approximate dynamic programming: a proposed methodology

Purpose Rapid sensitivity analysis and near-optimal decision-making in contested environments are valuable requirements when providing military logistics support. Port of debarkation denial motivates maneuver from strategic operational locations, further complicating Simulations enable rapid concept design, experiment testing that meet these complicated logistic support demands. However, simulation model analyses time consuming as output data complexity grows with input. This paper proposes a methodology leverages the benefits simulation-based insight computational speed approximate dynamic programming (ADP). Design/methodology/approach describes simulated environment demonstrates how informs parameters required for ADP dialect reinforcement learning (aka Q-learning). Q-learning includes policy prescribes decisions each state modeled simulation. paper's methods conform to DoD modeling practices complemented AI-enabled decision-making. Findings study means state–space reduction mitigate curse dimensionality. Furthermore, massive amounts become unwieldy. work reflect inputs so behavior can compare policies. Originality/value Fast computation is attractive while divorcing evaluation scenario-based limitations. The United States eager embrace emerging AI analytic techniques inform but hesitant abandon modeling. an aid overcome cognitive limitations way satisfies desire wield combined