Smarter Agriculture Takes Root

Engineers Revolutionize Farming with AI, Drones and Data.

Be it in the field with robotic sprayers, on a supercomputer with artificial intelligence or in the air with sharp-eyed drones, University of Florida researchers are bringing innovation and cutting-edge technology to farming.

Smart Agriculture targets efficiency and the environment. These technologies save time and money, as well as reduce waste and pesticides. UF researchers and graduate students are tending multiple test fields in Florida, from Quincy in the Panhandle to Immokalee in Southwest Florida.

Much of that work touches the Herbert Wertheim College of Engineering, particularly with AI use and digital applications.

Here are some samples of UF’s Smart Agriculture work:

Internet of Things 4 Ag

David Arnold, UF’s George Kirkland Engineering Leadership Professor with Electrical & Computer Engineering (ECE), is a lead for the multi-university, National Science Foundation-funded precision agriculture research program called IoT4Ag (IoT stands for the Internet of Things). The $50 million partnership with UF, the University of Pennsylvania, Purdue University, and the University of California Merced is designed to increase crop production by advancing technology.

The UF team is developing novel sensors, carrying out agricultural research leveraging the unique data from these sensors and developing new AI algorithms to process and understand the data collected automatically.

“UF is providing (A) deployable sensing systems, either stationary or mounted on ground robots, and also underground sensor networks; (B) new power and energy solutions – wireless power for charging sensors, fast-charging batteries for robots and solar-powered energy stations that are deployed in fields; (C) AI-image processing algorithms; and (D) AI- and data-informed crop growth models and digital twins,” noted Arnold, also the director of Florida Semiconductor Institute.

Arnold is the UF site director/innovation ecosystem director for the IoT4Ag partnership. Working in UF’s test fields in Quincy and Live Oak, UF’s IoT4Ag facilitates early detection and mitigates diseases and mold-related toxins in peanut crops. The technologies include automated crop monitoring using aerial and ground-based robotic imaging, AI methods for detecting toxins from hyperspectral images of peanut pods and kernels, and predictive models to process data to inform crop management.

The team consists of Arnold, Ph.D. (ECE), Carlos Messina, Ph.D., Ian Small, Ph.D., Katie Stofer, Ph.D., Ramdeo Seepaul, Ph.D., Charlie Li, Ph.D., Vivek Sharma, Ph.D. (all of IFAS), Alina Zare, Ph.D., Sanjeev Koppal, Ph.D., William Eisenstadt, Ph.D. (all of ECE), Brent Sumerlin, Ph.D. (chemistry), research coordinator Marcy Lee and graduate and undergraduate students.

A harvest of networks

UF researchers, including Jose Dubeux, Ph.D., Chang Zhao, Ph.D. (both from IFAS), Joel Harley, Ph.D., and Zare are developing mechanisms for data-collection networks and measuring plant biodiversity across the state to determine the relationships between plant biodiversity and land use. They collect the data from field measurements, flying drones equipped with hyperspectral cameras and satellite-imagery analysis.

The project’s goal is to quantify the ecosystem services provided by natural or agricultural lands to help incentivize certain management practices and better understand the impacts of land management practices on the overall environment and ecosystem.  

AI for nutrient recommendations

Another multi-college UF team is developing methods for nutrient recommendations of agricultural fields (with a focus on potatoes) using data, crop models and artificial intelligence. The project is led by Zare, Harley, Lincoln Zotarelli, Ph.D. (IFAS), Lakesh Sharma, Ph.D. and Marcio Resende, Ph.D. (both from IFAS).

The team, including students and postdocs, developed AI algorithms and advancements to crop modeling (that leverages and embeds AI into the processing) for nutrient recommendations in agricultural fields, Zare said. With the use of UF’s supercomputer, HiPerGator, this project helps farmers pinpoint optimal timing and amounts of fertilizer use, which could reduce hefty fertilizer costs.

Switchgrass into biofuel

UF is one of nine universities and research facilities studying how to make switchgrass — a fast-growing perennial native to the U.S. — into biofuel to enrich soil for crops, optimize climate adaptation and improve sustainability. 

Zare is this project’s UF team member. The idea is to increase biofuels to meet the global demand, as current biofuel production requires more water, fertilizer and other energy sources compared to switchgrass.

Working from a Department of Energy grant, the team transplanted various switchgrass in 10 fields in multiple states, from Texas to South Dakota. Examining the same plants in different locales allowed the team to see how the plant’s genes interact with different environments.

In nature, switchgrass sequesters carbon underground in its roots, produces cellulose that can be used to make ethanol and typically grows in soils that are unsuitable for food crops — all characteristics that make it a great candidate for biofuel. Current biofuels come primarily from agricultural crops or feedstocks such as corn.

Meanwhile, in Immokalee …

At UF’s Southwest Florida Research and Education Center in Immokalee – one of 12 UF agricultural research fields in the state – there is a smart tree sprayer that uses artificial intelligence to apply pesticides efficiently and only to the existing trees (as opposed to a steady stream that hits all spots, including vacant patches).

The work at this 320-acre facility is part of UF professor Yiannis Ampatzidis’ AI-Enhanced Precision Nutrient Management for Tree Crops program. A key component of Ampatzidis’ work is Agroview, a cloud-based application that processes information gathered by drones taking aerial pictures; it can provide nutrient information on crops, calculate the number of gaps in citrus tree rows, tree height, canopy area and leaf density.

An assistant professor with UF’s Southwest Research & Education Center, Ampatzidis contends Agroview’s reporting is 95% accurate and certainly more cost-effective than traditional crop-management applications. In a study of 10 citrus orchards, the number of trees detected in drone photos was thousands more per orchard than those detected by airplane photos.

Ampatzidis and his team also developed Agrosense, a now-patented, AI-enabled and ground-based sensing system that provides plant-based analytics such as tree crop counting, canopy density classification, tree-height estimation, and fruit counts, all critical for precision orchard management. Agrosense was also used on a tree crop sprayer to optimize agro-chemical usage by applying the right amount of chemicals to individual trees based on canopy size and leaf density.

The team is also on their third prototype of an AI-enabled robot sprayer in Immokalee. It, too, is designed to spray only where necessary to minimize chemicals and waste.

Additionally, the UF team in Immokalee is working with automated tree-trunk injection methods, which have needles on a support-frame apparatus affixed to a field vehicle equipped with a depth-sensing camera and a touch-screen monitor. Trees with signs of citrus greening are injected with antibiotic materials.

Tubes and hyperspectral signals

UF researchers from ECE and IFAS received a $1.2 million grant in 2024 from the United States Department of Agriculture to use machine learning, hyperspectral cameras and minirhizotron (MR) tubes (clear plastic tubes inserted into the ground next to a plant) to study plant root systems over time. SiTS: Hyperspectral Signal in Soil is led by Zare with Sanjeev Koppal, Ph.D., from ECE, Chris Wilson, Ph.D., and Diane Rowland, both from Agronomy, and Stefan Gerber from Soil and Water Sciences at IFAS.

Researchers have used minirhizotron tubes for years. A camera is inserted into the tube to take pictures of the roots as they grow. But this method has its challenges: Analysis is tedious and time-consuming and the cameras provide limited types of data.

The SiTS project tackles these limitations by developing AI algorithms and tools to automatically trace the roots and pull out important root features using AI methods. Second, the grant supports work to develop a hyperspectral camera to be installed into the MR tubes. Hyperspectral imaging can obtain much more information than traditional RGB cameras — for example, carbon levels in the soil, chemical composition in soil and in the roots.

High-tech peanut care

In Quincy, UF professor William Eisenstadt, Ph.D., of ECE is leading a team honing high-tech communications systems to monitor peanut and cotton crops.

It’s difficult to communicate across a large farm area,” he said. “We’re trying to improve the technology so that we move data across the farm to a central point and then process it.”

Currently, the UF team is using LoRaWAN devices placed around the fields to communicate with overhead drones about temperature and other numerical readings that can indicate crop health. But this technology cannot transmit large chunks of data, such as multiple high-resolution photos from drones.

The team is looking at DECT NR+, a powerful 5G wireless technology.

“It’s supposed to be able to move about 2MB of data in about a second. Of course, these are all optimal conditions, and it can go up to about two miles,” Eisenstadt said.

A flying drone, for example, would automatically survey a farm and push images back to a central point. That will allow real-time decisions for the farmers.

“You send the images back to a central point, and then you need to put an AI machine-learning system in there to help make automated decisions – ‘Oh, we look dry here. We may have plant diseases there. We have to look closer, take some more data,’” Eisenstadt said.

Smart sprayers

UF researcher Nathan Boyd is using an AI smart sprayer to reduce herbicide sprayed on crops.

“We are building multiple AI-powered herbicide-application units, each of which works in different ways,” said Boyd, a UF/IFAS weed scientist and horticultural sciences professor at the Gulf Coast Research and Education Center (GCREC).

Targeted herbicide applications kill weeds by applying the chemicals only where the weeds grow.

Boyd and his colleagues developed a precision-spraying system designed to make sure herbicide goes through the holes that scientists punch in the plastic mulch and into the soil where tomatoes are growing.

With the AI smart sprayer, researchers at the GCREC showed that the new technology found the punch holes 86% of the time, which translates to a savings of more than 90% on herbicide use in tomato fields at the research center. 

“This equipment applies herbicide on the soil wherever there is a hole in the plastic, because that is the only place that weeds can emerge, rather than over the entire bed top,” Boyd said.

Satellites and soil

Earlier this year, the National Institute of Food and Agriculture awarded UF engineering professor Joel Hartley a Sustainable Agricultural Systems grant for his project “Hybridizing Ecological Data for Mapping Soil Organic Carbon from Satellite Imagery.”

The research will use AI/machine learning to predict carbon in soil satellite imagery. It will explore opportunities to promote winter cropping systems in the Southeast.

“This team will use new AI strategies to overcome the scarcity of ecosystem service data, focusing on one particular service: carbon sequestration,” according to the National Agricultural Data Producers Cooperative. “They will integrate artificial intelligence, satellite imagery, geographic information systems, and in situ measurements to predict, track, and visualize ecosystem services at scale.”

“The hypothesis,” Hartley noted, “is that the overall ecosystem (vegetarian, soil, nearby water sources, etc.) will all play a role in helping us predict the soil carbon levels.”