Researchers at Johns Hopkins University have developed a robotic surgery system capable of performing procedures with the skill level of experienced human surgeons by using imitation learning, a machine-learning technique. This system trains robotic devices to mimic surgical techniques by watching videos rather than relying on hand-coded instructions for each movement.
The research, published in a peer-reviewed study and presented at the Conference on Robot Learning in Munich, leverages the da Vinci Surgical System—a commonly used surgical robot platform. Although widely used, the da Vinci system’s precision has traditionally been limited, requiring meticulous programming for each movement. Instead, the Johns Hopkins model uses imitation learning to train the robot on relative movements, allowing it to replicate actions such as needle manipulation, tissue lifting, and suturing by interpreting video data collected from surgeries worldwide.
Senior author Axel Krieger highlighted the ease of the new system, which interprets input from wrist-mounted cameras on da Vinci robotic arms to determine the necessary movements for each procedure. This model adapts the same machine-learning architecture that powers language models such as ChatGPT, translating the intricacies of motion into kinematic data, which can be applied to robotic motion control. Lead author Ji Woong “Brian” Kim noted that even a few hundred video demonstrations can enable the robot to generalize and perform in new settings.
During trials, the team’s model enabled the robot to autonomously adjust and continue its task when disruptions occurred, such as dropping a needle—a capability the model learned without specific programming for error recovery. Researchers envision this system as a way to train robots for a wide range of surgical tasks rapidly, potentially completing training in days rather than years, as was previously required.
The Johns Hopkins team, including PhD student Samuel Schmidgall, Associate Research Engineer Anton Deguet, and Associate Professor Marin Kobilarov, is now advancing the system to support full surgical procedures autonomously.
