Abstract:
The increased use of Unmanned Aerial Systems (UAS), or drones, requires the
development of effective detection and localization methods to mitigate escalating
privacy and security risks. Traditional active sensing techniques (visual, acoustic,
radar, RF) are susceptible to discovery or evasion. Aero Sense Phase 02 addresses
this gap by proposing an AI-enhanced passive sensing system that capitalizes on
the Extremely Low Frequency (ELF) magnetic signatures emitted by Brushless
DC (BLDC) motors. Building on the foundation of Phase I, this study extends
the detection range (beyond 10 m) and introduces a crucial localization functionality
within a predefined target area. The methodology involves designing and
testing optimized receiver antennas to capture these weak ELF signals, despite
challenges posed by high ambient noise. Multiple circular and octagonal loop
antennas were designed and experimentally characterized indoors and outdoors.
The analog front end, including a 1.65 V bias and passive 3:1 attenuation network,
reliably conditioned bipolar signals up to 10 Vpp into the 0–3.3 V ADC window
without clipping across 1–7 kHz. Drone signatures were analyzed using the Spectral
Correlation Function (SCF) computed via the FFT Accumulation Method,
which clearly separated drone-present and environment-only cases and revealed
distinct spectral patterns for different drone models. These SCF images formed
the input to CNN models, which were used to localize the drone in a real-world 5
m × 5 m grid, where a customized EfficientNetB0 reached approximately 85–98%
cell-level accuracy. The results confirm that ELF-based passive sensing combined
with SCF and deep learning is a viable alternative for short-range UAV detection
and coarse 2D localization.
