Scalable Techniques for Center-of-Mass and Course-of-Action Analytics for Intelligence Preparation of the Battlefield

Intelligence Analysts (S2) and Operations Planners (S3) are responsible for tracking adversaries on the battlefield.  These users must know where the opposing force (OPFOR) units are currently located and where they are going next.  To perform such analysis, an S2 or S3 might take all the current sensor data for a given Area of Operations (AO) and plot it on a map.  However, the user will be quickly overwhelmed with “sea of red” dots, icons, and mound of data that is difficult to understand.  To address this problem, Phoenix proposes the development of two analytics that use Machine Learning (ML) techniques to roll up individual pieces of equipment into units by echelon type, then assesses potential attack vectors. The conceptual strategy for a Center-of-Mass analytic will analyze a set of disparate inputs (i.e., equipment shown as dots), build a Graph Convolutional Neural Network (GCNN), and finally produce a set of unit symbols by echelon type. The second analytic, Course-of-Action Validation, evaluates OPFOR unit locations and movement in relation to friendly forces. It uses Gradient Boosted Decision Trees to assess potential attack vectors and identifies the likely enemy course of action.