Â鶹ÒùÔº

While strawberry production runs from November through April in Florida, digital twin technology lets scientists simulate the growth of the fruit year-round, allowing research to proceed year-round.

Digital twins are virtual replicas of objects, systems or processes that can predict system behavior as they interact in a simulated environment.

Dana Choi and her team of University of Florida scientists have now shown that the robotic system, powered by artificial intelligence (AI), is accurate and that it saves time and labor. That's critical for the $500 million-a-year Florida strawberry industry and could also be crucial for an industry worth $2 billion annually across the United States.

A few years ago, Choi's team built a digital twin of a strawberry field that copies every row, leaf and berry at life-size. Within that virtual field, scientists let the robot drive around and take thousands of photos of a simulated commercial farm in Hillsborough County.

Newly published in AgriEngineering shows that AI trained exclusively in a digital twin environment using simulated strawberry fields achieved 92% accuracy in detecting fruit, without relying on real-world training data.

"Because the computer-simulated field never goes out of season, new berry-spotting tools can be prototyped even in the summer—speeding innovation," Choi said. "The findings also mean lower development costs. Companies can test robotic pickers or smart sprayer designs in the digital twin, first, ironing out bugs before real-life trials. That ultimately lowers the price of new technology."

The robot, trained entirely on synthetic images, also estimated real-world fruit diameter with only 1.2 millimeters of error—"good enough for commercial grading, using only synthetic, simulated data," said Choi, a UF/IFAS assistant professor of agricultural and biological engineering.

This demonstrates the potential of AI models trained in virtual environments to support commercial decision-making tasks, such as classifying fruit based on characteristics like size or quality.

If growers know precise fruit size and volume, they can predict their yields and know when to harvest.

"The study shows that a realistic digital twin can jump-start AI tool development for strawberry farms, enabling faster, more cost-effective robotics innovation," said Choi, a faculty member at the UF/IFAS Gulf Coast Research and Education Center.

"Normally, we'd have to take thousands of photos in real fields, label each one and wait for the right season," she said. "That takes a lot of time and money. But with a digital twin, we can create and label these photos instantly."

Furthermore, training in the virtual world eliminates the need to handle or label real images, saving weeks of field work.

Why does all this matter? It takes less money and time to build and improve new tools because scientists can test and fix them in a virtual setting before trying them in real life.

The digital twin platform could also support operator training and rapid prototyping of autonomous machinery, helping move agricultural technology from concept to field faster and more cost-effectively.