ITHACA, N.Y. — Researchers have developed a robotic artificial intelligence system that could advance the development of disease-resistant crops, including mildew-resistant grapevine varieties.
The "BlackBird" robotic system, named after a military spy plane, replaces microscopes and months of human labor in analyzing mildew lab samples and makes it faster to identify which grapevines carry favorable genetic characteristics for mildew resistance.
Developing mildew-resistant vines is a big deal for the worldwide grape industry, which annually loses billions of dollars to powdery mildew in lost fruit and fungicide costs.
"(Powdery mildew) is a problem everywhere in the world that grapes are grown," said Lance Cadle-Davidson, research pathologist with USDA's Agricultural Research Service and adjunct associate professor in Cornell's School of Integrative Plant Science. He co-leads the BlackBird project.
He said having resistant varieties could lead to 90% less pesticide use per variety.
Before the BlackBird was invented, studying mildew resistance in vines was tedious.
The process involved breeders collecting leaf samples, then researchers spending hours hunched over microscopes manually scanning for mildew infection. This would take two or more months.
Now, that same work can be done by BlackBird in a single day, thanks to engineer and computer scientist Yu Jiang, assistant professor in Cornell's School of Integrative Plant Science.
Breeders collect leaf samples from vines and send them into Jiang and Cadle-Davidson's lab. The researchers punch 1-centimeter disks out of the leaves and arrange them on a tray, then infect samples. BlackBird automatically scans the leaves using an algorithm similar to facial recognition software that Jiang programmed to recognize mildew's spores and fine threads.
The robot can gather information at a scale equivalent to the resolution of a regular optical microscope, with even higher accuracy.
"In plain language, we tried to create a tool that replaces the most laborious part of this research study," said Jiang.
The software helps researchers identify phenotypes, or visible traits, of vines that carry more favorable genetic characteristics for mildew resistance.
Breeders can use this genetic information to develop mildew-resistant varieties.
Breeding is still a long, complex process that can take years, but BlackBird speeds up the first step: looking for resistance genes that breeders can utilize.
Cadle-Davidson said he believes the technology is applicable to any crops and any visible traits on the scale of micrometers to millimeters.
Breeders worldwide, public and private, are already requesting BlackBird.
"There are a lot of people who want them," said Cadle-Davidson.
The team has sent out two robots for beta tests. One is going to a USDA site in Corvallis, Ore., for lab work on spider mites and powdery mildew in hops. This robot will be shared with Washington State University hops researchers.
Before distributing the BlackBird widely, the team is fine-tuning the robot and instruction manual.
Cornell, USDA and collaborators have decided to keep BlackBird in the public domain to make it accessible and affordable.
The researchers say they believe BlackBird also has the potential to track how well specific fungicides perform against mildews.
Although BlackBird is only for lab use, Jiang and Cadle-Davidson are also collaborating on a separate project called the "phytopatholobot," a robotic system for field use.
BlackBird was made possible through the USDA funded Cornell-led VitisGen2 grape breeding project in partnership with the Light and Health Research Center.