NANJING, March 30 (Xinhua) -- At a strawberry farm in Jiangning District of Nanjing, east China's Jiangsu Province, a tracked robot the size of a supermarket shopping cart slowly makes its way between strawberry plants.
A camera mounted on it rapidly scans and precisely locks onto a ripe strawberry. A robotic arm then slowly brings the flexible gripper at its end to the target. Then, silicone tentacles on the gripper gently envelop the strawberry like a sea anemone, giving a light tug and twist. A whole, stem-free strawberry is thus neatly picked and gently placed into a collection basket.
This is the new-generation strawberry-picking robot developed by Wang Xiaochan's team from Nanjing Agricultural University. It takes an average of 20 seconds to pick one strawberry, with a grasping success rate of 84 percent. The harvested strawberries remain in almost perfect condition, similar to those picked by human hands.
Many agricultural products are harvested by machines to improve efficiency and reduce costs. However, harvesting soft and easily damaged fruits like strawberries had long remained a challenge.
Previous picking robots usually functioned by using scissors to cut the strawberry stems, always leaving a section of stem attached to the fruit. This residual stem easily punctured or scratched adjacent strawberries during transport from farm to processing workshop, causing significant fruit loss.
Even more troublesome, workers had to manually trim the stems before boxing the strawberries. Thus, the use of robots had not fully liberated human labor.
Wang's team drew inspiration from how sea anemones feed and developed a biomimetic soft gripper powered by pneumatics.
Made entirely of flexible silicone material, the gripper is highly elastic and deformable, ensuring it will not crush the strawberries. It can completely wrap around strawberries of various sizes and shapes, grasping with both stability and gentleness.
The bottom of the gripper features a pneumatic chamber that enables intelligent extension and retraction of the tentacles via air pressure regulation. The gripper can gently grasp the strawberry inward or spread outward to release the fruit, perfectly adapting to its soft, easily damaged nature.
To solve the residual stem problem, the team designed a "drag-and-rotate" motion for the robot that mimics human picking. After the gripper seizes the strawberry, the robotic arm first gives a slight upward tug and then completes a small arc rotation, allowing the stem connection point to break naturally. This means the picked strawberry is without any stem, ready to be boxed and sold, with no human processing required.
Additionally, the team employed a deep learning model to develop a strategy for handling plant blocking and adjusting the grasping pose. This allows the robot to accurately identify ripe strawberries.
Through a comprehensive scoring mechanism, the robot can evaluate the spatial relationships among clustered strawberries, giving higher scores to less blocked ones for priority picking. Relying on visual feedback from the camera, the robot can also dynamically adjust its grasping posture.
After picking each strawberry, the system recalculates the next picking target. Data shows that this strategy boosted the robot's grasping success rate from 66 percent to 84 percent.
Wang said that in the future, his team will establish a cloud data platform. Standardized picking data from planting bases will be uploaded to the cloud, allowing ordinary farmers to complete picking operations by simple commands without having to train the machine by themselves.
More importantly, this approach is not only applicable to strawberries, but also sheds new light on the intelligent harvesting of other soft fruits and vegetables such as tomatoes and cherries, helping to advance the intelligent upgrading of agriculture in China. ■
