Design of hardware and software platform for intelligent automation of livestock farming using internet of things

Abstract

The 4th generation industrial livestock farming reduces livestock losses, increases fertility rates, reduces operating costs, manages human resources, and generally increases productivity. In this research, a set of wearable sensors including a cattle collar and a leg mounted sensor was designed for automation of livestock farming. A LoRaWAN based internet of things network is designed using a set of custom gateways in three livestock farms. An intelligent livestock big data analysis framework that uses edge and cloud computing is designed for processing and modelling of the behaviour of the cattle using the collected sensor data. A decision support system for estrous cycle management, stress and health control and cattle behaviour modelling is designed using machine learning based modelling of this data. The proposed system monitors the cattle and provides the vital signs such as body temperature, mobility, feeding behaviour and estrous behavior for management and veterinarian decision support. The accuracy of the KNN algorithm for modelling livestock behavior is 78% and the accuracy of convolutional deep neural networks is 84%. However, due to the simplicity of the KNN algorithm, this method increases the battery life of the system by 4.5 times and therefore, it is a more appropriate choice for commercial livestock farming.