SRT-H: A Hierarchical Framework for Autonomous Surgery via Language Conditioned Imitation Learning
Ji Woong Kim, Juo-Tung Chen, Pascal Hansen, Lucy X. Shi, Antony Goldenberg, Samuel Schmidgall, Paul Maria Scheikl, Anton Deguet, Brandon M. White, De Ru Tsai, Richard Cha, Jeffrey Jopling, Chelsea Finn, Axel Krieger
2025-07-10
Summary
This paper talks about SRT-H, a new robotic system that can perform complex surgeries on its own by using a two-level approach where one part plans the surgery using language instructions and the other part controls the precise movements of the surgical tools.
What's the problem?
The problem is that real surgeries are very complicated and require robots to handle many detailed steps over a long time, while also dealing with changes in human tissues. Most previous robotic systems can only do simple tasks or need human help for complex surgeries.
What's the solution?
The researchers designed SRT-H with a high-level system that creates step-by-step plans using language commands, and a low-level system that makes the robot’s movements to follow the plans accurately. This way, the robot can handle the whole surgery by correcting mistakes and adapting as needed.
Why it matters?
This matters because it is a big step toward fully autonomous surgical robots that can perform real operations without human intervention, which could make surgeries safer, more precise, and accessible to more patients.
Abstract
A hierarchical framework with high-level language-based task planning and low-level trajectory generation achieves full autonomy in ex vivo cholecystectomy.