From machine learning to digital twin integration for livestock production and research

TL;DR

This paper explores how combining machine learning and digital twins can improve livestock production and research by predicting animal behavior and optimizing farming practices.

Contribution

The paper introduces the novel integration of machine learning and digital twin technologies to dynamically simulate and predict animal states in livestock systems.

Findings

ML-DT integration can dynamically predict animal physiological and behavioral states in real-time.

The combination offers predictive tools for animal welfare and production efficiency.

ML-DT models provide deeper insights compared to traditional snapshot-state simulation models.

Abstract

Globally, climate change, economic crises, and increased food demand pose significant challenges to the stability of agricultural production systems, underscoring the urgent need for more innovative approaches and tools to advance livestock production science. Machine Learning (ML) development supported the Digital Twin (DT), a digital replica of a real-world entity, as a game-changer in modern livestock science, enabling the prediction, optimisation, and simulation across various research environments. At the same time, it has been shown that synergism between ML and Digital Twin (DT) can mimic animals' physiological and physical state and behavior based on input data, leading to a better understanding of animal behavior, nutritional requirements, physiological status, or environmental stressors to investigate responses and suggest precise decisions. Moreover, such animal simulation models can offer deeper insights and predictive analytical tools that support animal welfare, forecast production efficiency, and sustainability. Although traditional simulation models are mainly snapshot-state models that indicate what should happen on average, ML-DT integration serves as a living mirror, dynamically predicting what is happening right now and what will happen to each animal under various changes. This integration can be a versatile tool for introducing solutions in the research domain; however, its augmentation remains complex and poses significant ethical, economic, and governance challenges. This review discusses recent ML-DT synergism applications in both barns and labs, highlighting their potential to reform both industry and research.

Figures