ML-Augmented Optimization of LoRa Antennas for Drone Telemetry

A compact printed monopole antenna for drone telemetry communication, operating at 433 MHz, with a gain of 2.2 dBi, is designed. The antenna is fabricated on an FR-4 substrate with dimensions of $0.116~\lambda _{0} \times 0.073~\lambda _{0}$ and is optimized for long-range communication. A Machine Learning-augmented Optimization (MLaO) method is proposed to reduce the antenna design time compared to traditional electromagnetic simulation techniques. Typically, antenna design involves complex computer simulations like CST and computationally intensive parameter sweeps. However, in this work a surrogate Artificial Neural Network (ANN) model trained on 1080 antenna designs replaces the heavy CST simulations. This ANN model is then coupled with a Simulated Annealing (SA) optimizer to generate antennas with the desired characteristics, reducing the total design time to 57% compared to traditional techniques. Three antenna designs were simulated using MLaO for different long-range (LoRa) frequency bands, with Ata1 (433 MHz) achieved a return loss (S11) of −23.3 dB, Ata2 (865 MHz) had an S11 of −30.6 dB, and Ata3 (dual band at 433 MHz and 865 MHz) with S11 of −15.6 dB and −35.4 dB respectively. The fabricated antenna (433 MHz) was mounted on a drone and tested with a 3DR-433 telemetry transceiver, recording an average Received Signal Strength Indicator (RSSI) of −57.8 dBm up to 470 m. These results demonstrate the proposed antenna’s efficiency, compactness, and the effectiveness of the MLaO approach for fast and accurate antenna design.