Real-Time Auto-Monitoring of Livestock: Quantitative Framework and Challenges

The use of automated sensors has grown rapidly in recent years, with sensor data now routinely used for monitoring in a wide range of situations, including human health and behaviour, the environment, wildlife, and agriculture. Livestock farming is a key area of application, and our primary focus here, but the issues discussed are widely applicable. There is the potential to massively increase the use of empirical data for decision-making in real time, and a range of quantitative methods, including machine learning and statistical methods, have been proposed for this purpose within the literature. In many areas, however, development and validation of quantitative approaches are still needed in order for these methods to effectively inform decision-making. Within the context of livestock farming, for example, it must be practically feasible to repeatedly apply the method dynamically in real time on farms in order to optimise decision-making, and we discuss the challenges in using quantitative approaches for this purpose. It is also crucial to evaluate and compare the applied performance of methods in a fair and robust way-such comparisons are currently lacking within the literature on livestock farming, and we outline approaches to addressing this key gap.