Radar Detection of Loitering Drones Via Artificial Neural Networks: A Generative AI Software Support

Abstract

The focused effort of this study refers to developing an ANN-based generative AI software towards radar detection of loitering drones in a combat area. Relevant heuristics can provide an adjunct support to a surveillance radar system that remains vulnerable to threats on attacks by low-observable loitering drones. Detection of such drones is necessary for proactive counter actions. In general, a loitering drone implies a suicide munition deployed as an exploding device mimicking kamikaze-style operation at war zones. It is intended to crash on to a vulnerable target such as, a surveillance radar. The present study envisages specifying a compatible radar-specific algorithm towards robust, true-positive detection of menacing drones in a combat zone and adopt an ANN thereof  to get trained and predict the estimation of drone-location with an acceptable probability of detection. That is, a test ANN is trained with a set of acquisitioned data pertinent to backscattered signal (echo) received from the drone illuminated by the radar. Compatible ANN of perceptron architecture with backpropagation is indicated and details on compiling necessary data vis-à-vis radar detection algorithm towards training and prediction phases (of the ANN) are outlined. Simulation experiments are explained and results are presented to illustrate the efficacy of the test ANN as a generative AI software.