A Multi-Layered Software Ecosystem for Precision Management of Livestock Farming

Abstract

This paper proposes Livestock IoT (LIoT), a five-layer Software Ecosystem (SECO) to improve precision livestock farming that utilizes IoT sensors, TinyML-based on-device inference, LoRa long-range communication, and a cloud layer to analyze data and provide continuous monitoring of cattle body temperature, heart rate, movements, and voice sounds. The designed neck-wearable node, based on the ESP32-S3 microcontroller, is capable of local inference using a TinyML health classifier model that has achieved around 75% accuracy in field trials. The extracted alerts and health status updates are delivered through a LoRa gateway to farm management dashboards, improving veterinary monitoring and facilitating farmers’ decisions. The proposed architecture is also envisioned with federated learning for decentralized model updates and blockchain for secure and tamper-proof log recording as future working directions. The system effectiveness was demonstrated in rural environments where LoRa connectivity was achieved at distances up to 2 km and the gateway operated continuously for the full 14-day trial without recharging. LIoT offers a low-cost, scalable approach to improving the efficiency, sustainability, and animal welfare of livestock farming in the Agriculture 4.0 era.