RF-Based UAV Detection and Identification Enhanced by Machine Learning Approach

Abstract

This paper introduces the design and implementation of an RF-based system for detecting non-cooperating unmanned aerial vehicles (UAVs). The system comprises an RF module, an automated Pan and Tilt Unit (PTU), and IoTs. The integrated RF module is highly sensitive and capable of detecting signal level up to −120 dBm in the frequency range of 2 to 6 GHz. It is designed primarily to receive frequencies used for civil and commercial UAV applications, specifically those at 2.4 and 5.8 GHz. The PTU enables azimuth rotation scanning from 0 to 360 ° and elevation scanning from 0 to 75 °, effectively covering the surveillance space and locating the direction of RF radiation maxima. The system processes the signal data to extract key features and aiding in differentiating emitter types such as Wi-Fi, Bluetooth, or UAV control signals. Machine learning algorithms are trained to make decisions based on these extracted features. Comprehensive testing validates the system’s successful performance, meeting predefined criteria. The efficacy of the system is underscored by the discernible RF fingerprints received from distinct emitters and their respective spatial parameters. In general, this paper contributes to the field of UAV detection by presenting an integrated system that combines hardware and software components, offering reliable and efficient identification and tracking of UAV signals in the presence of other RF emitters.

Keywords

• Internet of Things (IoT)
• machine learning
• radio frequency (RF)
• signal processing
• software defined radio (SDR)
• transceiver