Digital twins as decision-support tools for automation in agriculture: A case study on robotic vaccination

Agriculture faces increasing pressure to improve efficiency, ensure animal welfare, and reduce dependence on manual labour, yet traditional operations often lag in adopting Industry 4.0 and 5.0 solutions. Digital technologies, par- ticularly digital twins, offer a transformative approach by enabling virtual representation, real-time simulation, predictive analytics, and performance validation prior to physical deployment, thereby providing a cost-effective proof of concept. This paper introduces a framework that not only show- cases a robotic vaccination system for feedlots but also establishes a proof-of- concept pathway to demonstrate the potential value and practicality of such technologies for agriculture. One of the latest state-of-the-art development environments, NVIDIA Isaac Sim, provides a robust platform for building these digital twins, allowing robotic systems to be tested and optimized in realistic, dynamic environments. Using this environment, we outline a robotic vaccination system that combines a robotic arm-based injection mechanism with reinforcement learning agents and a Detectron2 deep learning model these systems allow for precise neck muscle segmentation and accurate vac- cine site targeting. The system was evaluated on two robotic platforms: Isaac Sim’s built-in Franka Emika Panda arm and a customized manipulator, which may offer realistic and cost-effective solutions for feedlots. Simulation results demonstrate achievable positional accuracy, robust control, and reliable task execution across both platforms. This digital twin-based framework reduces reliance on early-stage physical prototyping, enhances safety, measures op- erational efficiency, and serves as a decision-support tool, highlighting the critical role of digital simulation in enabling practical, scalable automation in modern agriculture.