Detecting Multiple Perturbations on Swine using Data from Simulation of Precision Feeding Systems

Abstract

Industry 4.0 brings transformation to all industries, including agriculture. Smart livestock has been replacing traditional livestock as a trend of the agricultural industry in the world. Precision feeding is one of the areas of smart husbandry that combines many modern multidisciplinary technologies which are prominent such as AI, IoT, Big Data, etc. To obtain that for pigs, a precision feeding system needs to be implemented. Components of the system include automatic feeders connected to a computer system to collect and process data on daily feed intake of fishes and animals, and/or from ambient sensors. Perturbations such as heat stress or sanitation issues have a significant impact on the nutritional profile of group housed pigs. However, perturbation is often detected only after it has occurred and is recognized late by the consequences left on the animal. Although the cause of perturbations might be unknown, the effect on the animal can be observed early throughout the data of voluntary feed intake. By the precision feeding system, the data are processed and analysed based on mathematical models following a two-step approach: (1) estimation of target trajectory of cumulative feed intake using linear and quadratic functions, and (2) detection of perturbations based on deviations from the target cumulative feed intake. However, implementing such a system requires huge costs and is often beyond the capabilities of farms, production households and small/medium laboratories. In this paper, we introduce an agent-based modeling approach to simulate precision feeding systems for swine, whose data can be used to early detect multiple perturbations which may have appeared. Experiments were carried out on GAMA simulation platform to demonstrate the efficiency in detecting multiple perturbations of group housed pigs, and also prove the usefulness of simulation of precision feeding systems.