A Digital Twin Radar Environment for Testing and Evaluation of AI Applications

Modern AI (i.e., Deep Learning and its variants) is here to stay. However, its enigmatic “black box” nature presents a fundamental challenge to the traditional methods of test and validation (T&E). Or does it? In this paper we introduce a Digital Engineering (DE) approach to T&E (DE-T&E) using a Digital Twin of the RF environment that can achieve requisite mil-spec statistical validation as well as uncover potential deleterious “Black Swan” events that might otherwise not be uncovered—until it's too late. An illustration of these concepts is presented for an advanced modern radar target estimation example employing deep learning AI.