Artificial Intelligence in Livestock Robotics: A Review of Research Advances and Application Prospects

Abstract

Livestock farming is a cornerstone of the global agricultural economy. Yet as farms expand, experience-based management falls short. It lacks efficiency and precision, along with animal welfare. Precision Livestock Farming (PLF) addresses these limits. It uses continuous, automated, individual-level monitoring and management. Livestock robots serve as key execution platforms. These platforms are increasingly enhanced by artificial intelligence (AI). We examine computer vision and deep learning for animal detection. We also cover behaviour recognition and health assessment. We review the shift from fixed-path navigation. We look at reinforcement-learning-based dynamic planning. We also cover multi-modal perception fusion. Our work highlights three research frontiers. These include large language models (LLMs) and embodied intelligence. They also include model lightweighting and edge deployment. Multi-agent coordination with UAV herding is the third. Our work also assesses general-purpose robots. It covers round-the-clock health inspection. Natural-language interaction is included too. We outline the lab-to-farm deployment gap. We cover human-animal-robot interaction. Data standardisation is addressed. Economic viability is also discussed.