IoT-Enabled Remote Health Monitoring System for Horses Using Edge Computing

Abstract

The increasing demand to have a continuous and non-invasive equine health surveillance has stimulated the creation of intelligent surveillance systems that can function in real life scenarios. In the paper, an IoT-based remote health monitoring system on horses based on edge computing is introduced, which works to record physiological and behavioural data and convert them into valuable health information. An indigenous Indian horse breed was put under a long-term data collection of a wearable multi-sensor IoT device in farm and semi-rural settings. Body temperature and tri-axis accelerator were measured in the acquired data, which indicated the physiological condition and the behaviour in the activities. The acquired data were processed and analysed by machine learning to determine abnormal health conditions, especially heat stress-related and behavioural suppression conditions. Various regression algorithms were tested such as Linear Regression, Ridge Regression, and Random Forest Regression. Random Forest Regression was the best performer and showed better predictive results because it has the capability to capture non-linear relations and can work with noisy sensor data in the real world. Combination of behavioural characteristics with physiological measurements greatly minimised misclassification as well as enhanced early detection of anomaly compared to individual-parameter monitoring strategies. Experiments verify that proposed system is strong, stable and can be implemented in the real-time scenario on the edge computing platforms. The paper demonstrates the usefulness of the multi-sensor fusion and ensemble learning based on IoT to have a scalable and context-aware equine health monitoring.