Reallocation and Cross-Domain Optimization using Relationally-Consistent Semantic Extensions (ReCOURSE)

ABSTRACT: Real world uncertainties all but guarantee that execution will deviate from even carefully considered plans, and therefore any effective military command will assume that resources must be reallocated to achieve desired outcomes. Unfortunately, reallocation is typically a manual, reactive, and ad hoc process that results in sub-optimal use of resources and overall force effectiveness. To provide a system that can dynamically reallocate resources across air, space, and cyber domains, Charles River Analytics proposes to design, demonstrate, and evaluate Reallocation and Cross-Domain Optimization using Relationally-Consistent Semantic Extensions (ReCOURSE). Our approach consists of three primary thrusts: (1) build a semantically rich, logically grounded representational layer that can be used to model all world and plan states in a consistent fashion; (2) extend the representation layer with a computational inference layer that can input heterogeneous information and output analyses and projections of world states; and (3) build a multiobjective optimization layer as an extension of the inference layer to generate options and identify tradeoffs among possible plan elements. Together, these three layers form what we call a Semantic Type Engine. During Phase I we will instantiate our Semantic Type Engine in a demonstration prototype to illustrate the feasibility and benefits of our approach. BENEFIT: The research performed under this effort will have immediate benefit to Air Operations Center Weapon System (AOC WS) and its configuration across multiple regional and functional AOCs. Additionally, advances in this area of mission planning could help in other military domains across air, maritime, ground, and space operations. This research will also have direct application to enhance our commercial EAToolkit product, a software development kit for optimization using evolutionary algorithms.