Custom-made passive direction finding system

Roursgaard, Erik Kold

4. semester

Electronic Systems, Master

2026

2026-06-03

Over the past decade, unmanned aerial vehicles (UAVs), commonly known as drones, have become widely available for private, commercial, and military use. This development has increased the need for new radio-frequency (RF) based monitoring methods, especially in situations where cooperative identification systems such as Direct Remote ID are unavailable, disabled, spoofed, or unreliable. This thesis investigates how to design and evaluate a low-cost, custom-built passive direction-finding system that can determine the direction of incoming RF signals in the 2.4 GHz ISM band, a frequency band commonly used by drones. A patch antenna array was designed using four individually fed (2×2) patch antenna subarrays, which together form a larger (4×4) array with sixteen patch elements. The antenna array was measured at the Stargate facility at APMS, where controlled measurements with known angles were collected to accurately relate the received signals to their direction of arrival. The measured signals were processed into magnitude and phase features and used to train four supervised machine learning models: Decision Tree, K-Nearest Neighbor, Support Vector Machine, and Multilayer Perceptron. The results show that the Support Vector Machine and Multilayer Perceptron models achieved the best overall performance, and that phase information in the signal is crucial for estimating the direction of arrival. The system should therefore be regarded as a proof-of-concept for a direction-finding subsystem rather than a complete UAV detection system.

Ubemandede luftfartøjer (UAV'er), også kendt som droner, er i løbet af de seneste ti år blevet langt mere udbredte til private, kommercielle og militære formål. Det øger behovet for nye metoder til at overvåge og spore dem ved hjælp af radiosignaler (RF), især når samarbejdsbaserede metoder som Direct Remote ID ikke er tilgængelige, er slået fra, forfalskes (spoofes) eller ikke kan stoles på. Denne afhandling undersøger, hvordan man kan udvikle og teste et billigt, specialdesignet passivt pejlesystem, der kan bestemme retningen på indkommende radiosignaler i 2,4 GHz ISM-frekvensbåndet, som også bruges af mange droner. Der blev konstrueret et antennearray bestående af fire mindre (2×2) patchantennearrays, som tilsammen dannede et større (4×4) array med i alt seksten patchantenner. Antennesystemet blev testet i Stargate-faciliteten hos APMS, hvor der blev gennemført kontrollerede målinger med kendte vinkler, så man præcist kunne se, hvorfra signalerne kom. De registrerede signaler blev omdannet til data om styrke (magnitude) og fase, som derefter blev brugt til at træne fire typer af superviserede maskinlæringsmodeller: Decision Tree, K-Nearest Neighbor, Support Vector Machine og Multilayer Perceptron. Undersøgelsen viste, at Support Vector Machine og Multilayer Perceptron gav den bedste samlede ydeevne, og at faseinformationen i signalet var afgørende for at kunne estimere signalets retning. Systemet demonstrerer derfor et funktionelt pejlesubsystem, men er endnu ikke et fuldt UAV-detektionssystem.

[This abstract has been rewritten with the help of AI based on the project's original abstract]

Radio frequency (RF) ; Machine learning ; Direction finding ; Supprt Vector Machine (SVM) ; K-Nearest Neighbor (KNN) ; Decision Tree (DT) ; Multilayer perceptron (MLP) ; Deep learning (DP) ; Antenna array ; Antenna phased array ; Angle of Arrival (AoA) ; Stargate

Download PDF
View record in AAU Student Projects