A Task-Level Safety Framework for Autonomous Surgical Robots

Surgical robotics has moved in four decades from a repurposed industrial manipulator to teleoperated platforms in routine clinical use, and laboratory systems have now executed complete surgical steps autonomously. The instruments that govern this transition grade a device by the highest autonomy it can exhibit, yet the safety of an autonomous action depends on which task is automated, how quickly injury follows failure, and whether the supervising surgeon can regain control in time. This paper analyses the taxonomic, evaluative and regulatory instruments currently in use, identifies five gaps they leave open, and derives from them CASR, a context-aware assurance framework. CASR characterises autonomy per surgical task, combines it with a clinical criticality index and a supervisory margin relating take-over time to the time-to-harm horizon, discounts evidence by its transfer distance, composes task classes across a procedure, and maps the result onto five assurance classes carrying graded obligations that interoperate with existing instruments. Applied to four published exemplars it separates cases that a levels-of-autonomy analysis conflates, including a teleoperated procedure whose burden is raised by network latency alone. The framework is a calibratable proposal, and a validation roadmap is set out.

کلیدواژه ها:

surgical robotics ، artificial intelligence ، levels of autonomy ، telesurgery ، assurance case ، human–robot interaction ، patient safety ، medical device evaluation