An IoT architecture for decision support system in precision livestock

Abstract

Sustainable animal production is a primary goal of technological development in the livestock industry. However, it is crucial to master the livestock environment due to the susceptibility of animals to variables such as temperature and humidity, which can cause illness, production losses, and discomfort. Thus, livestock production systems require monitoring, reasoning, and mitigating unwanted conditions with automated actions. The principal contribution of this study is the introduction of a self-adaptive architecture named e-Livestock to handle animal production decisions. Two case studies were conducted involving a system derived from the e-Livestock architecture, encompassing a Compost Barn production system - an environment and technology where bovine milk production occurs. The outcomes demonstrate the effectiveness of e-Livestock in three key aspects: (i) abstraction of disruptive technologies based on the Internet of Things (IoT) and Artificial Intelligence and their incorporation into a single architecture specific to the livestock domain, (ii) support for the reuse and derivation of an adaptive self-architecture to support the engineering of a decision support system for the livestock subdomain, and (iii) support for empirical studies in a real smart farm to facilitate future technology transfer to the industry. Therefore, our research’s main contribution is developing an architecture combining machine learning techniques and ontology to support more complex decisions when considering a large volume of data generated on farms. The results revealed that the e-Livestock architecture could support monitoring, reasoning, forecasting, and automated actions in a milk production/Compost Barn environment.