Cao's Digital Twin Expertise Helps Global Farmers

Agriculture might be the most important industry in the world—but it is also one of the riskiest. Staple crops for millions of people can be threatened by a single pest, like the Fall Armyworm, which has devastated maize (corn) crops as it spread across Africa, Asia, and the Americas over the last decade.

Food production also uses about 70 percent of global freshwater, while overapplication of nitrogen fertilizer enters the atmosphere, contributing about 12 percent of human-sourced greenhouse gases every year.

Farmers across the globe are eager to increase agricultural sustainability and reliability, but making the wrong decision—like under-watering, under-fertilizing, or applying pesticide at the wrong time—could have dire consequences.

But what if they had a way to practice?

"An agricultural digital twin is a virtual replica of a real farming environment that can simulate crop conditions in real time," said Charles Cao, an associate professor in the Min H. Kao Department of Electrical Engineering and Computer Science at the University of Tennessee, Knoxville. "The twin brings data from drones, ground sensors, and satellite imagery into a single platform where we can model what's happening on a farm, predict what might happen next, and even run 'what-if' scenarios."

Cao is the University of Tennessee's lead principal investigator on Holistic AI-powered Agricultural Response Validation and Early Prediction System across Territories (HARVEST), an international effort led by the Missouri University of Science and Technology (Missouri S&T) to create next-generation tools that help farmers identify and manage agricultural threats.

The project aims to reduce crop losses from pests and disease by 20 to 30 percent, and decrease fertilizer and water usage 20 percent, by developing advanced tools farmers can use to detect threats early, manage resources more efficiently, and make better-informed decisions in the field.

HARVEST is one of six projects funded under the National Science Foundation's inaugural AI-ENGAGE program, which brings together researchers from the United States, Japan, India, and Australia (the Quad nations) to apply artificial intelligence to some of agriculture's most pressing challenges.

Cao is spearheading development of the HARVEST AI-driven digital twin in collaboration with several researchers from UT's Institute of Agriculture (UTIA), building an integrated system that connects AI and engineering expertise with real-world agricultural research.

"This isn't just about building better algorithms," Cao said. "It's about putting practical tools in the hands of farmers around the world—tools that help them grow more food with fewer resources and less risk."

Sharing Expertise; Maintaining Privacy

Members of HARVEST hailing from the US, including Cao, anchor the program's efforts in AI, precision agriculture, and molecular diagnostics.

The HARVEST team also includes experts in digital twins and Internet of Things networks from Japanese institutions, including Osaka University; experts in edge AI computing and insect ecology from the Indian Institute of Science in Bangalore, India; and experts in sensor networks and plant biology from the University of Melbourne in Australia.

At UT, Cao's team is also leading the effort to maintain each nation's privacy, building AI models that can learn from distributed data sources across international borders while keeping the underlying information secure.

"Because HARVEST spans four countries with different agricultural practices, data regulations, and environmental conditions, the project must enable collaboration without exposing sensitive data," Cao said. "Our approach creates an AI system that gets smarter from global experience without any single country or farm having to share its raw data."

Helping Every Farm Access Digital Twins

Cao's collaborators at UTIA include Ethan Parker, the director of the East Tennessee AgResearch and Education Center, and Shigetoshi Eda, who has spent more than a decade developing rapid, on-site molecular diagnostics for agricultural diseases.

In a previous collaboration, Cao, Parker, and Eda developed a novel diagnostic method called DNAzyme-LAMP, which can identify fungal infections in corn in minutes—right in the field.

As part of HARVEST, the team will now integrate DNAzyme-LAMP with AI analysis and drone-based hyperspectral imaging (which captures infrared, visible, and ultraviolet light) to create a complete diagnostic pipeline: aerial drones scan fields to flag potential problem areas, humans on the ground collect samples to identify the cause of the problem, and the digital twin integrates the results to guide effective treatment.

"This combination of aerial surveillance, molecular diagnostics, and AI-driven decision support represents a shift from reactive pest management to proactive, precision intervention," Cao said.

The digital twin pipeline will be enhanced with large language model (LLM) technology that translates complex sensor data into plain-language recommendations, making advanced analytics accessible to farmers who may not have a technical background.

Cao's team, including Missouri S&T Professor Sajal Das, presented their initial work on the LLM-orchestrated digital twins at the First International Association for the Advancement of Artificial Intelligence (AAAI) Workshop on AI in Agriculture in Singapore this January. The team's first sensor prototypes are preparing for field deployment later this spring.

At the same time, Cao's lab is also working to create an edge computing system that will allow the drones to analyze crop health, detect pests, and assess soil conditions independently—without a cloud or internet connection.

"The drones communicate through a network designed for low-power, long-range data transfer, enabling coverage even in remote agricultural areas," Cao explained. "This makes the technology practical not just for large commercial operations but also for smallholder farms in developing regions."

Provided by University of Tennessee at Knoxville