Unitree G1 Performs World's First Live Microsurgery, Study in Nature

What actually happened in the operating room?

A research team led by Liang Zekai, a doctoral student at the University of California San Diego, used a Unitree G1 humanoid robot to perform laparoscopic cholecystectomies on two live pigs. The paper, titled "In vivo feasibility study of a humanoid robot for surgical procedures," was published in Nature, making it the first peer-reviewed account of a general-purpose humanoid robot completing live microsurgery. Both operations succeeded without conversion to open surgery.

The setup required real engineering work. The team built a custom fixture so the G1 could grip a commercial manual-wristed laparoscopic forceps, a tool designed for human hands. A senior surgeon operated remotely from a console wearing a stereoscopic headset, with hand movements scaled and mapped to the robot's wrist. A second surgeon stood at the bedside to handle camera control, tissue traction, and instrument cleaning.

The numbers tell a clear improvement story across the two procedures. Console operation time dropped from 56 minutes 15 seconds to 31 minutes 59 seconds. Robot redeployments fell from eight to four. The team attributes the gains to growing proficiency, not hardware changes.

The real competitive advantage of humanoid robots in surgery is not raw precision but the ability to walk into an existing operating room and pick up tools already designed for human hands.

What are the honest limitations?

The research team did not soften the problems. Latency in the teleoperation link disrupted surgical flow. The robot's workspace is narrow, forcing frequent recalibration mid-procedure, which is a major reason total times remain far above those of the da Vinci system. Most critically, the G1's components cannot survive the high-temperature, high-pressure sterilization that human surgery requires. The team covered the robot's hands with sterile gloves, a workaround that falls well short of clinical sterility standards.

These are not minor footnotes. Latency and sterility are regulatory and safety barriers, not just engineering inconveniences. Any path to human trials runs directly through solving both. The paper itself notes that the first da Vinci laparoscopic cholecystectomy took six hours; the same procedure now takes 30 minutes. That precedent suggests the current G1 results, while rough, are not disqualifying.

Why does this matter for the robotics industry?

The da Vinci surgical system costs tens of millions of yuan to procure, requires proprietary instruments, and demands a modified operating room. Hospitals in lower-resource settings, rural areas, or field medical environments simply cannot access it. The G1, by contrast, used standard commercial laparoscopic instruments and was wheeled to the operating table on its own legs. That portability and compatibility with existing tools is the economic argument the paper is really making.

Unitree's robot was originally built for tasks like boxing and dancing demonstrations. The fact that a doctoral student's lab adapted it for live surgery with a custom fixture points to a broader shift: general-purpose humanoid platforms are becoming a substrate that researchers can modify for specialized tasks without starting from scratch. That lowers the barrier for medical robotics research considerably.

The study also lands at a pointed moment. Elon Musk has publicly predicted that Tesla's Optimus will surpass human surgeons within three to four years. Unitree, a Chinese company with a much lower public profile, has now put peer-reviewed animal data on the table first. The competitive geography of humanoid robotics in medicine is no longer purely speculative.

What needs to happen before this reaches human patients?

Several concrete problems must be solved before any regulatory body would consider human trials. Sterilizable components are the most immediate hardware requirement. Latency reduction, likely through dedicated low-latency network links or on-site operation rather than true remote surgery, is a close second. The recalibration problem points to a need for better spatial awareness and arm positioning software.

Beyond hardware, the workflow itself needs redesign. The current setup required close coordination between a surgical team and an engineering team just to position the robot correctly. Clinical adoption demands that deployment be fast and reliable without engineering support on site.

The paper's authors frame the work as a feasibility study, not a clinical proposal. That framing is accurate and important. What the study establishes is that the concept is not blocked by any fundamental physical barrier. The remaining obstacles are engineering and regulatory, which means they are, in principle, solvable.