Unitree Humanoid Performs Live Pig Cholecystectomies in Nature Study
Unitree's robot, which once appeared on the Spring Festival Gala, has now entered the operating room. A recent paper in Nature reports that a UC San Diego research team used a humanoid robot developed by Unitree to perform two cholecystectomies on live pigs.
The first author and corresponding author, Zekai Liang, is from China. He earned a bachelor's degree in mechanical engineering from Huazhong University of Science and Technology and is now pursuing a PhD at UC San Diego.
This is the first time a humanoid robot has completed live surgery. Surgical robots are not new to clinicians: the da Vinci system was approved for clinical use in 2000 and has been used for over 20 years. The Nature study's first specifically refers to the humanoid form.
The team chose Unitree's G1 as the hardware platform and developed a custom end effector, teleoperation system, and surgical control algorithms. Laparoscopic surgery requires the instrument to maintain a fixed remote center of motion around the abdominal wall trocar. Specialized robots such as da Vinci guarantee this pivot mechanically, but a humanoid lacks such a mechanism and must estimate it in real time through vision and algorithms, demanding higher control precision.
Results showed the humanoid approached specialized robot performance only under specific port layouts. Straight-line error was about 1.3 mm, similar to specialized robots, but complex curved-trajectory error rose to about 10.4 mm, roughly 10 times specialized systems. Teleoperation latency was about 156 ms, near the 150 ms threshold generally considered acceptable.
Surgeons and ordinary subjects then tried the humanoid, a specialized surgical robot, and traditional laparoscopy. The specialized robot scored best at 97.67, the humanoid 85.39, and traditional laparoscopy 70.47. Subjectively, the humanoid's operating burden fell between the specialized robot and manual laparoscopy.
In live-animal experiments, the team completed laparoscopic cholecystectomy on two live pigs. Both surgeries finished key steps such as Calot's triangle dissection and gallbladder detachment, with no conversion to open or traditional laparoscopy. The process was not easy: animal breathing and slight robot-base drift caused the software-calculated remote center of motion to shift. The two surgeries required 8 and 4 pause calibrations, with long interruptions exceeding 3 minutes occurring 2 and 3 times. Postoperative questionnaires showed surgeons generally agreed the system could complete surgery, but the median clinical maturity score was only 2.5 out of 5. Even operating surgeons felt it remained far from clinical entry.
Why put a humanoid on the surgical table? The researchers gave three reasons. First, no need to modify the surgical environment: a humanoid's height and arm span are close to a human's, so it could theoretically stand in an existing operating room and use existing laparoscopic instruments, without replanning the layout as da Vinci requires. Second, lower cost and broader use: da Vinci has high purchase and maintenance costs and proprietary consumables, usually available only in well-resourced large hospitals; a humanoid is relatively inexpensive and could theoretically do other work. Third, smaller footprint and less obstruction: compared with da Vinci's bulky bedside tower arms, a humanoid is more compact, potentially reducing crowding and leaving more room for the bedside assistant.
Yet safety and accuracy matter most in surgery. The humanoid's drawbacks are clear: its precision is nearly an order of magnitude worse than a specialized surgical robot. G1's arm span is about 450 mm, while an adult's is generally 1.6 to 1.8 m, limiting workspace. Sterility is also hard to solve; it currently needs human-like sterile gloves. The paper's purpose is not to put humanoid robots into operating rooms now, but to use surgery, the highest-threshold and least fault-tolerant test, to examine the current ceiling of humanoid capability. The results are honest: humanoids have made significant breakthroughs but still have not reached the clinical threshold. A Science Robotics study had a robot autonomously perform eight ex vivo pig cholecystectomies without human intervention, with 100% success. But that, too, remains far from clinical use.