A Deep Learning Approach for Accurate Path Loss Prediction in LoRaWAN Livestock Monitoring.

Area of Science:

• Agricultural technology
• Wireless communication systems
• Artificial intelligence in agriculture

Background:

• The agricultural sector is integrating sensing, communication, and AI for an industrial revolution.
• The Internet of Things (IoT) is crucial for remote livestock monitoring, but faces challenges in field communication, coverage, and data transmission.
• Monitoring livestock in mountainous regions presents unique communication hurdles due to terrain and land cover.

Purpose of the Study:

• To assess LoRa communication for livestock monitoring in the Italian Alps.
• To analyze LoRa path loss prediction in diverse land-cover types and optimize gateway deployment for reliable coverage.
• To develop and validate a deep learning model for accurate path loss estimation in challenging mountainous terrains.

Main Methods:

• Empirical assessment of LoRa communication performance in mountainous pastures.
• Utilizing remote sensing for land-cover recognition.
• Developing a deep learning model, Bidirectional Long Short-Term Memory (Bi-LSTM), for path loss prediction.

Main Results:

• The proposed deep learning approach significantly reduces path loss estimation errors, achieving less than 5 dB error.
• The model demonstrates a 2X improvement over state-of-the-art methods in path loss prediction accuracy.
• Experimental data validates the model's effectiveness in complex, rugged landscapes.

Conclusions:

• The study advances IoT-driven livestock monitoring by providing robust communication solutions for mountainous environments.
• The developed deep learning model offers precise path loss prediction, enhancing the reliability of LoRa networks.
• This research addresses critical communication challenges, paving the way for more effective remote livestock management.