Tactical Drone Detection and Tracking using Machine-Learned Radio Interferometry: Lessons from Ukraine

Overview

The utility of inexpensive radio-controlled drones has never been clearer than during the present conflict in Ukraine. Remote-controlled aircraft—quadcopters, small planes, and even balloons—provide a cheap and disposable source of reconnaissance, surveillance, and munitions-deployment. As a result, the need arises to accurately detect and track drones in an ever-changing and chaotic environment. Current solutions rely on unreliable visual detection or expensive, cumbersome, and inflexible radio-based tracking. In response to the increase in prevalence and agility of drone-based threats, new tracking methods are required. Using off-the-shelf hardware, I designed and tested a low-cost drone detection and geolocation system using radio interferometry. I built custom antenna-switching circuitry using a high-speed RF switch IC, discrete components, and Arduino microprocessor. Antennas in an array (4-dipole UHF grid) are rapidly switched to an open-source software-defined radio. I wrote a custom signal-processing algorithm in python, which runs on Linux. I wrote and tested manual signal-processing algorithms—using phase-sample counting, k-means clustering, and density-based grouping (DBSCAN)—as well as regressive neural networks to compute detection attributes. Using the principles of radio interferometry, my algorithms analyze the phases and magnitudes of UHF radio signals from a drone, in the array of antennas. The system estimates drone bearing, range, and altitude—alerting users and graphically displaying crucial information. I tested my system in the field using drone hardware identical to that used in Ukraine. Results confirm a minimum bearing accuracy of 5°, an elevation accuracy of 10 m, and a minimum detection range of 150 m. Using machine learning algorithms, my system can perform using geometrically complex, non-standard antenna arrays. This adaptability is crucial towards the feasibility of the integration of this system. I designed and tested a flexible drone detection and tracking system for under $50.

Awards (1)

Category Award: 1

Competition history

CSEF 2023 Electronics & Electromagnetics · Entry S1005

Resources

Related projects

ISEF · 2023