Millimetre Wave Radar Target Classification using UAVs in an Adversarial Electronic Warfare Environment

Abstract

Millimetre wave radar promises to revolutionise the application of radar in the military setting which will lead to a corresponding change to how electronic warfare (EW) systems are applied. This type of sensor is potentially small, light, low-power, cheap and offers excellent radar imaging performance. Furthermore, the prevalence of unmanned and autonomous systems will see this type of sensor applied to dynamic and distributed scenarios. This thesis considers the imaging capability of a small, low-power 94GHz (W-band) sensor in terms of high resolution range profiles (HRRPs), a common form of radar imaging. The thesis addresses the problem of imaging of maritime targets using such a sensor, ultimately performing target classification using a number of classifiers such as naïve Bayes, support vector machines (SVMs) using linear, polynomial and radial basis function (RBF) kernels, classical correlation and linear discriminant analysis (LDA). Modelling and simulation is undertaken to establish training and test sets which are affected by noise, multipath propagation and ship dynamics. A W-band radar system was constructed and a significant trials program undertaken to record HRRPs from targets of opportunity as well as from hired boats. This data was then used to investigate two key areas of the problem: trajectory optimisation of a simulated, autonomous, imaging unmanned aerial vehicle (UAV) and susceptibility of such a sensor to electronic attack (jamming). The key new contributions made in this thesis are the development of an algorithm that utilises classifier training data to determine the path flown by a simulated, radar imaging UAV in 2D space. All previous works investigating target classification relied on pre-determined paths of best-practice. Furthermore, applying electronic attack to such as sensor has not been carried out in open literature before.