China’s Crop Robot Hits 92.8% Flower Recognition In Pollination Test
GEAIR 2.0, a Chinese humanoid crop-breeding robot, uses two arms and AI flower recognition to automate hybrid pollination after improving stigma-recognition accuracy from 85% to 92.8%.
A Beijing crop-breeding robot now reaches 92.8% stigma-recognition accuracy and uses two arms to move through plants, Interesting Engineering reported, turning a laboratory pollination system into a more capable embodied-AI experiment for agriculture.
GEAIR 2.0 was unveiled at the 33rd China Beijing Seed Industry Conference as an autonomous humanoid system for hybrid pollination, one of the most repetitive and time-sensitive steps in crop breeding.
A genetics and developmental biology institute within the Chinese Academy of Sciences built the system for dense crop settings where constant human control is impractical.
The upgrade changes both the machine and the plant-breeding workflow around it.
Earlier work used a flexible single-arm system.
The new version has one hand to hold a pollen tube and another to use a pollination brush, allowing it to avoid leaves and branches while keeping access to the flower structure it needs to target.
That design is tied to a measurable recognition gain.
The robot’s ability to identify the stigma, the female reproductive organ of a flower, rose from 85% to 92.8%.
In a launch demonstration with tomato plants, the system completed each pollination in less than 10 seconds, while one brush dip into pollen could support about 50 pollinations, with China Daily credited for the demonstration detail.
The project’s name stands for Genome Editing combined with AI-based Robotics, and its method depends on both parts.
The plant side of the design uses genome editing so male-sterile flowers present their stigmas more openly, reducing the petal-opening work that people would otherwise do before pollination.
The robot then uses AI and robotics to locate the exposed reproductive structure and carry out the cross-pollination task.
That pairing reflects what the developers call crop-robot co-design.
A 2025 Cell study described automated tomato cross-pollination that achieved efficiency comparable to manual work before researchers extended the approach to soybean.
The finding matters because hybrid pollination can require precise timing, repeated movements and large amounts of manual labor across breeding programs.
The operational case is not only speed.
Automating pollination could reduce labor needs, lower breeding costs and let breeding teams run repetitive tasks for longer periods under changing conditions.
Dense leaves, branches, needle-sized stigmas and narrow pollination windows remain the practical test for whether the robot can move from demonstration settings into larger breeding operations.
GEAIR 2.0 also fits a broader Chinese push to modernize agriculture with biotechnology and AI.
In 2025, Chinese farms produced a record 715 million tons of grain, while national planning documents set a roughly 725-million-ton comprehensive production-capacity goal for 2030.
The robot does not solve that target by itself, but it shows how embodied AI is being aimed at a specific agricultural bottleneck rather than a general farm automation promise.
The clearest result is the shift from a one-arm research system to a two-arm platform with higher flower-recognition accuracy and a timed tomato-pollination demonstration.
For crop breeders, the next proof point is whether those 92.8% recognition and sub-10-second pollination figures can hold when the system faces broader plants, field variation and longer operating runs.
