Shanghai
A slender robotic arm navigates its way through tomato plants using its 3D vision, which scans the surrounding environment. When the arm stops at a flower, sensors mounted on its fingertips assess the viscosity of the pollen, and algorithms process the data in a fraction of a second.
Once the right conditions are established, the pollen is broken up by high-frequency vibrations and deposited precisely on neighboring flowers, completing the artificial pollination process with amazing precision. The robot then navigates its way to the next plant automatically, avoiding obstacles with a level of awareness comparable to that of a human.
Developed by researchers at Fudan University, the smart agricultural robot system integrates multiple advanced technologies covering areas such as 3D visual perception, autonomous navigation, cloud-based decision-making, and deep learning with AI, representing the culmination of a remarkable model of embodied AI.
Its seamless operations are the result of four years of development and four generations of prototypes—the work of a research team led by Shang Huiliang, an associate professor at the university.
The multifunctional robot handles the entire tomato cultivation process, including pollination, leaf pruning, fruit thinning, and harvesting. Unlike previous single-function devices, this “thinking” machine can mimic human cognition, decision-making, and task execution.
The team’s research journey began in 2021, identifying agricultural robotics as a key development area. “We realized that precision automation in agriculture presents enormous challenges and opportunities,” Shang said.
However, technical obstacles were daunting, and machines struggled to perform some tasks, such as harvesting covered fruit or navigating dense foliage, despite their simplicity for humans.
Faced with these complex challenges and limited staff, Shang’s team adopted a systematic approach to research and development. At the beginning of the project, the team leveraged the university’s academic capabilities and collaborated with optics and algorithm experts to solve the problem of plant cover on fruit, while materials scientists and mechanical engineers developed flexible electronic arms.
The team gradually built an integrated, multidisciplinary research unit spanning mechanical engineering, electronics, automation, software development, and artificial intelligence.
After four years of effort, the team successfully transformed industrial robotic arms into fourth-generation autonomous machines. Li Ruijiao, a researcher on the team, said the current model coordinates “eyes, brain, hands, and feet” through embodied artificial intelligence, maintaining a pollination success rate of over 90 percent even under difficult natural conditions.
The model, which demonstrates remarkable efficiency, is currently undergoing field testing at a Bright Food Group farm in Chongming, where one robot can replace six workers. This combination of performance and affordability has generated significant interest in the industry.
The team follows a demand-driven innovation model, with members conducting extensive field research on 20 farms in Shanghai, Qinghai, Guangdong, and Hainan.
“We’ve turned the traditional research and development process upside down, starting from the actual pain points of farmers,” Li added, noting that developing physical technology requires continuous problem-solving under real-world conditions.
