New Grant Gets Proactive with Pigs

By Laura Ivanitch

Inside a barn used for commercial swine production, upwards of 2,000 pigs spend their days eating, breeding and growing. Creating an environment within these massive pig houses to promote their inhabitants’ health and wellbeing presents farmers with both opportunities and challenges. 

In theory, barns control the temperature, humidity and air circulation to maintain conditions where pigs thrive. But the reality is more complex, says  Lingjuan Wang-Li, interim department head and William Neal Reynolds Distinguished Professor in NC State University’s Department of Biological and Agricultural Engineering.  

While enhanced technology gives growers a leg up when it comes to raising healthy pigs, limitations lead to suboptimal growth conditions, which result in more than $900 million in losses annually in the U.S. swine industry. 

Funded by a $1.15 million grant from the National Institute of Farming and Agriculture (NIFA), Wang-Li is leading a team to take a new approach to controlling these indoor environments. She and her team of 11 researchers, hailing from a broad range of specialties across three universities, are embarking on the Proactive Pig Project (P3). Their plan: Take advantage of artificial intelligence to glean pig data that can help manage barn conditions proactively. 

Data-informed Farming 

Since its inception in the early 1980s, precision livestock farming (PLF) has enabled farmers to make informed decisions on animal welfare based on data collected from the animals. For example, pigs wear sensors that alert farmers to illness or injury.  

Meanwhile, barns are monitored by a separate sensor system that notifies farmers when conditions like barn temperature fall outside of a predetermined range. These ranges are based on optimal pig growth and development. Once alerted that they are out of “ideal range,” farmers can make decisions about increasing or decreasing airflow, temperature and the like.  

Pigs are individuals, however, and with each animal’s unique genetic makeup comes an equally distinct response to the environment. Singular pigs have specific needs regarding air temperature, air circulation, and other factors, just like humans.  

Wang-Li compares it to a group of people sharing an office — and a thermostat. “During summertime I prefer 70 degrees,” she says. “A lot of my colleagues prefer 60 degrees.” 

Like humans, pigs also get stressed in uncomfortable spaces. That stress can lead to physiological changes affecting their digestion, weight, reproduction, immunity and more. Stress can also affect the quality of meat.

From Data to Action 

In order for PLF to account for the needs of individual animals, it must first use monitors to collect data from pigs and their environments — things like body temperature, heart rate and respiration rate, and eating and drinking behaviors.  

This body of data will help the swine industry move to a proactive model.  

“If we don’t use the data to build AI, whatever we do using sensors based off of environmental controls is just reactive to whatever the sensor detects,” Wang-Li says. 

By adding data from pigs to the data created by monitoring barn environments, artificial intelligence can make recommendations for individualizing living conditions. Farmers might turn on a fan or adjust the thermostat to lower a pig’s stress before that stress threatens the pig’s welfare. 

The research involved with the P3 project will tap researchers and faculty with a broad range of experience and expertise. “We just could not do the study without a transdisciplinary approach,” says Wang-Li. “It’s not only research activity, but also, some faculty may focus on research, others on extension and teaching. There are a lot of pieces.”

Faculty involved in the P3 project include: 

• Suzanne Leonard, NC State Animal Science Assistant Professor in PLF, will lead efforts in collecting and investigating live pig data with PLF technologies, which includes thermal cameras, microphones, radio frequency identification, weight estimation cameras, water flow meters, and RGB+IR cameras that detect both visible and infrared light; Eric van Heugten, NC State Animal Science Professor, who specializes in swine feed and nutrition; and Christian Maltecca, also NC State Animal Science Professor, working in swine data driven genetics and genomics.
• Lirong Xiang, Assistant Professor of Biological and Agricultural Engineering (BAE) in sensors, controls, automation and applied AI, will work to modify technologies currently used in PLF to collect pig data and then to develop offline AI models, which can operate with limited or no internet access, to automatically detect animal behavior and model the statistical interaction of their wellbeing with environmental conditions for stress classification.
• Xipeng Shen, an NC State Computer Science Professor specializing in dependable AI and augmented deep reinforcement learning, will integrate the offline AI models with a deep reinforcement learning-based AI system and controls for indoor pig environments to create a proactive management system.
• Anna Johnson, Iowa State University professor of animal behavior and welfare, and Glen Almond, a professor at NC State’s College of Veterinary Medicine, will work to define pig stress levels on a three-point scale: moderate stress, severe stress, and life-threatening stress. 
• Chadi Sayde, an associate professor of biological and agricultural engineering at NC State, will use his expertise to provide insight into environmental sensing and CFD (Computational fluid dynamics) modeling. Mahmoud Sharara, an associate professor of biological and agricultural engineering at NC State, will weigh in on sustainable waste management.
• Mulumebet Worku, professor of animal molecular genetics and biotechnology at North Carolina A&T, will focus on outreach and Extension.

Meeting Global Needs  

With growing populations, the United Nations projects the need for animal-based foods could increase 70% by 2050. As demand for food grows, so will the importance of minimizing agriculture’s harmful effects on the natural world. This, combined with larger farm sizes and fewer farm workers, points to a greater need for efficiency.  

Combining PLF’s real-time monitoring technologies with AI will allow farmers to determine optimal amounts of food, water and other inputs needed to produce pork products. This could reduce waste, lower cost and improve outcomes for farmers while contributing to improved animal welfare. As a result of the group’s effort to create the P3 project and financial support through the NIFA grant, Wang-Li’s team will begin the live pig experiments in early December.

“I’m personally so proud of this team,” she says.