University of Georgia

Transformers: Robots and drones accelerate plant genetic research

UGA professor of engineering Charlie Li with his team

High-tech tools commonly found in factories and on the battlefield may soon play a major role in helping feed the world’s rapidly growing population.

UGA researchers are developing a robotic system of all-terrain rovers and unmanned aerial vehicles that provide quick and accurate phenotyping—gathering and analyzing data on the physical characteristics of crops, including their growth patterns, stress tolerance and general health. Such information is vital for scientists working to radically increase agricultural productivity in a time of rapid population growth.

“By the middle of this century scientists estimate the world’s population will reach 9.1 billion people, which is a 30 percent increase in a little more than 30 years,” said Changying “Charlie” Li, professor of engineering. “This increase in population will demand that we nearly double our current food production. That’s a tall order, but one solution is to use genomic tools to develop high-quality, high-yield, adaptable plants.”

The current process for gathering data on plant characteristics is expensive and painstakingly slow, as researchers must manually record their observations one plant at a time. But the team of robots developed by Li and his collaborators will one day allow researchers to compile data on entire fields of crops throughout the growing season.

The project addresses a major bottleneck that’s holding up plant genetics research, said Andrew Paterson, Regents’ Professor and leader in mapping and sequencing flowering-plant genomes.

“The robots offer us not only the means to more efficiently do what we already do, but also the means to gain information that is presently beyond our reach,” he said. “For example, by measuring plant height at weekly intervals instead of just once at the end of the season, we can learn about how different genotypes respond to specific environmental parameters, such as rainfall.”

In addition to multispectral, hyperspectral and thermal cameras, the robots will be outfitted with Light Detection and Ranging, or LiDAR, a remote sensing method that uses light in the form of a pulsed laser to measure distances. The technology will allow the researchers to create precise three-dimensional images of plants.

During preliminary testing last year, the team collected an estimated 20 terabytes of data over the six-month growing season. Li said the team will collect 30 times that amount when the robots are fully deployed.

To analyze these massive data sets, the researchers are developing an artificial intelligence algorithm similar to the facial recognition program Facebook uses to facilitate the identification and “tagging” of people in photographs.

“Our algorithm will be able to scan an aerial photo of a large field and automatically identify the location and number of flowers on each plant,” Li said.

With fleets of autonomous vehicles rumbling through rows of crops and hovering overhead, algorithms also will play a key role in making sure the robots and drones perform their assigned tasks while staying out of each other’s way. Javad Mohammadpour Velni, assistant professor of engineering, is developing a suite of analytical tools that will allow the ground and aerial vehicles to operate independently but collaboratively to efficiently cover fields and collect data.

The researchers believe their work will provide a platform for plant geneticists to gather large amounts of data and empower advances in crops to sustain the planet’s population.

“Historically, genetics has been credited for about half of the yield gains that permitted small numbers of farmers and producers to sustain large human populations, for example during the Green Revolution,” Paterson said. “It’s realistic to envision that genetics will need to account for about half of the doubling of agricultural output that’s needed by 2050. This will require roughly doubling historical rates of progress in crop improvement, and more detailed and efficient phenotyping will be essential.”

This story appeared in the fall 2017 issue of Research Magazine. The original press release is available at