Location: Livestock Bio-Systems
Project Number: 3040-31320-001-010-N
Project Type: Non-Funded Cooperative Agreement
Start Date: Feb 28, 2026
End Date: Feb 28, 2031
Objective:
Labor shortages are a critical issue for the swine industry, particularly in the farrowing barn where continuous 24h real-time monitoring is essential for improving piglet and dam survival during farrowing and throughout lactation. In addition, continuous monitoring of other health and welfare issues throughout the production cycle is essential for early detection of issues, particularly sow lameness, to identify problematic animals for faster intervention to improve welfare and reduce economic losses. Precision Livestock Management (PLM), the use of sensors, information technologies, and data analytics to monitor and manage livestock production in real-time, provides a significant new approach to improve production efficiencies for the swine industry. Therefore, the primary objective of this project is to further develop and optimize our previously established PLM system utilizing a multi-modal swine behavioral monitoring system to monitor animal activity/events and develop predictors and corresponding invention technologies using large language models (LLM) to maximize lifetime productivity in swine, particularly during farrowing/lactation and the identification of sow lamen.
Approach:
Our previously established multidisciplinary research team, consisting of engineers and computer scientists across several United States universities and swine research scientists at ARS (ARIS, NCFA#58-3040-1-004N, 11/01/2020-10/30/2025), has developed and optimized a robust behavior monitoring PLM system that utilizes video, accelerometry (wearables), and ground vibration (seismological technology, geophones) technologies in conjunction with developing predictor/invention tools using LLM. This system has been validated in swine farrowing barns as capable of monitoring health of dams through heartbeat and respiration to predict stressful events, predicting the onset of farrowing based on the activity of the dam, and evaluating piglet growth throughout lactation via activity monitoring. Recently, our group has begun to evaluate the effectiveness of this system to actively monitor and detect the onset of sow lameness. For this project, all animal facilities, pigs, and animal production data will be provided by ARS scientists at the U.S. Meat Animal Research Center (USMARC). Geophones will be produced and provided by our Cooperators at Stanford University, the University of Illinois, and the University of Michigan to generate ground vibration data in farrowing barns and for detecting sow lameness. These geophones will be attached under metal farrowing pens at USMARC strategically located to capture the movement of piglets and dams within each pen. Activities to be monitored in the farrowing barn include but are not limited to events around parturition, nursing, and aberrant behavior resulting in piglet crushing. For monitoring and predicting sow lameness, geophones will be placed directly on concrete walkways or within strategic areas (e.g., around electronical sow feeders) of gestation pens at USMARC to capture sow gait and identify potential abnormalities in gait to predict lameness. To create predictive models for vibration data, multiple types of data will be collected and analyzed by machine learning capabilities using LLM provided by our Cooperators at Stanford University, University of Illinois, and University of Michigan. Video data will also be collected using cameras provided by our Cooperator at the University of Nebraska. These cameras will be mounted above farrowing pens, walking hallways and throughout gestation pens. In addition, accelerometers to measure specific activity of sows in the farrowing barns and for assessing sow lameness will be provided by our Cooperators at the University of Nebraska and accelerometers for individual piglet activity will be provided by ARS scientists. Video data and accelerometry data will be used to provide ground truth data for comparison of activities identified by geophones. Although this project will primarily focus on monitoring and predicting events in the farrowing barns and for identifying sow lameness, the utility of this system could be expanded to other potential areas throughout the production life-cycle to maximize efficiencies in swine.