Humanoid robots have moved beyond factory demonstrations and research laboratories into one of the most demanding environments imaginable: the operating room.
Researchers and surgeons at the University of California San Diego have used modified Unitree G1 humanoid robots to conduct two laparoscopic gallbladder-removal procedures on live pigs. The preclinical study, published in Nature, is described by the team as the first in vivo demonstration of a general-purpose humanoid robotic system performing surgery.
The milestone is significant, but it requires an important qualification. The robots did not decide how to perform the operations and were not working autonomously. Experienced surgeons controlled their movements remotely through a teleoperation console, while medical personnel handled anesthesia, initial access, port placement and several bedside tasks.
A world-first preclinical trial
The researchers completed two laparoscopic cholecystectomies, the surgical removal of the gallbladder, at the UC San Diego Center for the Future of Surgery.
Both procedures were performed on live porcine subjects under general anesthesia and with institutional animal-care approval. The operations followed the standard sequence used in robot-assisted gallbladder surgery, including tissue retraction, dissection around the gallbladder, identification of critical anatomical structures and separation of the gallbladder from the liver bed.
According to the published study, both procedures were completed without converting to conventional laparoscopic or open surgery. The humanoid system recorded 56 minutes and 15 seconds of active console operating time during the first case and 31 minutes and 59 seconds during the second. The number of robot deployments or major repositioning events also fell from eight in the first procedure to four in the second.
The experiment represents an important step for the wider humanoid robot industry. Rather than testing locomotion, warehouse handling or choreographed movements, the researchers evaluated whether a commercially available humanoid could support a safety-critical clinical workflow under realistic physiological conditions.
Image Credits: UC San Diego JacobsHow the Unitree G1 surgical system worked
The research team transformed the commercially available Unitree G1 into a surgical platform nicknamed Surgie.
At the control station, a surgeon used two master manipulators, a foot pedal and a stereo video headset displaying the laparoscopic camera feed. The system translated the surgeon’s hand movements into coordinated movements of the G1’s arms and surgical instruments.
Custom wrist adapters allowed the humanoid to hold commercially available wristed laparoscopic tools. A servo-controlled interface opened and closed the instrument grippers, while an inverse-kinematics system calculated the robot wrist positions needed to reproduce the surgeon’s commands.
One of the central engineering challenges was maintaining a remote center of motion. During laparoscopic surgery, each instrument must pivot around the small entry point where it passes through the patient’s abdominal wall. Purpose-built surgical robots control this mechanically. The G1, as a general-purpose humanoid, required software, optical markers and its head-mounted camera to estimate and maintain that pivot point.
A safety harness was also attached to the robot during the procedures as a precaution against instability, although the researchers said it was not required for normal system operation.
Humans remained firmly in control
Despite the striking images of humanoid robots standing beside an operating table, this was not autonomous robotic surgery.
A senior surgeon controlled the primary G1 from the console throughout both procedures. Human surgeons and clinical fellows also performed abdominal access and trocar placement before positioning the humanoid beside the operating table.
Because a compatible robotic clip applicator was unavailable, the clipping of the cystic duct and ligation of the cystic artery were performed using conventional laparoscopic techniques. The humanoid then resumed the robotic dissection and separation of the gallbladder from the liver.
UC San Diego’s public announcement describes one procedure as a human-robot operation and the other as involving two humanoids working together. The study’s methods provide a more detailed account: a second G1 was used briefly for camera holding and tissue retraction during the first case, while human assistants provided most bedside assistance in both procedures.
The most accurate interpretation is therefore that the Unitree-based system successfully enabled surgeons to perform key stages of two live animal operations—not that autonomous humanoid robots independently completed surgery.
Why researchers selected the Unitree G1
The Unitree G1 is a compact, commercially available humanoid originally designed for robotics research, embodied AI and general manipulation tasks.
According to Unitree Robotics, the standard G1 starts at $13,500 and can be configured with between 23 and 43 joint motors. The platform includes a depth camera, 3D lidar, an eight-core processor and optional force-controlled dexterous hands.
Its relatively low entry price and human-like body shape made it an attractive experimental platform. Unlike a purpose-built surgical machine, a humanoid can stand in a workspace designed for people, approach a conventional operating table and interact with tools originally intended for human hands.
The G1’s commercial price should not, however, be confused with the total cost of a clinically deployable surgical system. The study required custom instrument adapters, an operating console, imaging hardware, control software, tracking equipment and extensive engineering support. No commercial price has been announced for the complete Surgie platform.
The project nevertheless demonstrates how an accessible humanoid from Unitree Robotics can be adapted for an application far beyond its original demonstrations.
Could humanoids make robotic surgery more accessible?
Existing surgical robots deliver extremely high precision, but they are specialized systems that require dedicated equipment, trained teams and operating rooms with sufficient space.
A general-purpose humanoid could theoretically offer a different model. The same robot might assist with an operation, transport equipment, hold a camera, retrieve instruments or help prepare and clean the room.
Researchers believe this flexibility could eventually make robotic assistance easier to deploy in rural hospitals, emergency-response environments and regions that cannot support a permanent specialized robotic surgery installation.
Michael Yip, one of the study’s senior authors, said remotely operated humanoids could potentially expand access to surgeries in communities facing shortages of skilled medical personnel. The team has also identified field medicine, disaster response and long-distance surgical assistance as possible future applications.
These remain long-term possibilities rather than proven clinical use cases. The study involved only two animal procedures at a specialist surgical simulation center with engineers and experienced surgeons present.
Why the G1 is not ready to operate on humans
The trial revealed a clear divide between technical feasibility and clinical readiness.
Robot drift and changes caused by the animals’ breathing required the team to pause and recalibrate the system. Instrument exchanges and major repositioning events generally added at least three minutes to the workflow.
Researchers also reported limitations involving latency, reach, joint strength, calibration accuracy and intermittent overheating. The G1’s compact proportions restricted its surgical workspace, sometimes forcing the team to reposition the robot or adapt the trocar configuration.
Sterility presents another major obstacle. The researchers covered the robot’s arms with gloves during the animal procedures, but current commercial humanoids do not have fully autoclavable components designed for human operating rooms. Future systems will require validated sterile coverings that do not interfere with sensors, joints or calibration.
The G1 has not received regulatory approval as a surgical device, and the study does not demonstrate that the platform is safe for use on human patients.
A new test for general-purpose humanoids
Most leading humanoids are being developed around manufacturing, logistics or domestic assistance. Models such as the 1X NEO, Tesla Optimus and Figure 03 are primarily presented as platforms that could eventually perform a broad range of physical tasks.
The UC San Diego experiment introduces a more demanding benchmark. Surgery requires controlled movements, reliable communication, predictable behavior and the ability to stop safely when conditions change.
The trial does not prove that general-purpose humanoids can replace established surgical robots. It does show that their human-compatible form can be useful in environments and workflows originally designed around people.
That distinction may become increasingly important as manufacturers compete to develop robots that can move between tasks rather than remain permanently assigned to a single workstation.
What happens next
The research team plans to improve motion accuracy, reduce teleoperation latency and develop more robust methods for tracking the surgical entry points. Future trials will need to involve longer procedures, additional surgical tasks and larger numbers of cases.
A more immediate role for humanoids may be assisting rather than operating. A robot could hold an endoscopic camera, maintain tissue retraction, pass instruments or perform physically repetitive tasks under the supervision of medical professionals.
Greater autonomy may follow, but only after extensive technical validation, clinical testing and regulatory review.
For now, the Unitree G1 trial should be viewed as a proof of feasibility rather than the arrival of an autonomous robot surgeon. It shows that a general-purpose humanoid can be adapted to perform delicate, medically relevant manipulation inside a living body while controlled by a trained surgeon.
That alone represents a notable expansion of what today’s commercially available humanoid platforms can do—and a glimpse of how robots designed for human environments might eventually enter some of their most specialized workplaces.
Readers can explore the official Humanoid Surgeon project, review the peer-reviewed Nature study or compare humanoid robots and their specifications on Anton Robots.