On the Spring Festival evening in Beijing, a group of robots completed the performance of "making dumplings with skillful hands" in front of a national audience for the first time.They knead the dough at a speed of 0.8 seconds each, and the movements are clean and neat, and can even leave clearly identifiable fingerprints on the thin dumpling skin. Behind this is the precise cooperation of 21 degrees of freedom of the joints.
Realman's RealBOT wheeled folding robot shows stronger scene adaptability.When it takes out the soy sauce bottle from the refrigerator, it relies on the 6D visual positioning system to identify the reflective surface of the glass with an accuracy of 0.1 mm, and then adjusts the grip force in real time through the pressure sensor in the flexible gripper. This combination of technology allows the robot to break through the world-wide problem of "grabbing transparent objects" for the first time.
The data shows that this performance is not a simple "show off skills."Tests by the Robotics Institute of Beijing Institute of Technology show that the current success rate of robot dumplings has reached 92.7%, far exceeding the 43% in the laboratory stage in 2024.
Why is dumpling making called the "Mount Everest" of the robotics world? The Intelligent Robot Research Center of Tsinghua University explains that processing a flexible dough with a thickness of only 0.3 mm requires simultaneously solving three major problems: material deformation prediction, multi-modal sensing fusion, and real-time motion planning.
On the Spring Festival Gala stage, the "three-finger twisting" technique demonstrated by the Ruilman robotic arm,What it actually relies on is a bionic tactile array sensor - 256 pressure-sensing units distributed per square centimeter, which can sense subtle stress changes in real time as the skin expands.
The truly disruptive breakthroughs are hidden behind the scenes. The RynnBrain project of Alibaba Damo Academy has given these robots the ability to "interrupt memory": when it is necessary to pause making dumplings to open the refrigerator door, the system will automatically save the 128 deformation parameters of the current dough, and then seamlessly connect again.
This spatiotemporal memory technology enables a robot to have human-like multi-threaded housework processing capabilities for the first time, and its underlying architecture has triggered extensive discussions in the GitHub open source community.
