Track manipulation, locomotion and embodied systems.
Search papers, labs, and topics across Lattice.
To automate high-fatigue surgical subtasks, the authors created AGRO-SUVIDE, an agentic framework that extracts, synthesizes, and executes manipulation skills on viscoelastic substrates using just a single expert demonstration. The system uses LLM agents to program procedural skills or train policy-based modules, composing them at runtime into a verifiable, closed-loop execution graph that handles dynamic retries. Evaluated over 340 physical trials on the da Vinci Research Kit, the framework achieved an 85% single-fragment removal success rate and generalized to unseen five-fragment configurations with 80% accuracy.
Code-generating agents can now transform a single multimodal surgical demonstration into verified, self-correcting manipulation loops on real da Vinci hardware, clearing viscoelastic tissue fragments with up to 95% success with minimal intervention.
Augmented dexterity has the potential to reduce the fatigue experienced by surgeons during repetitive surgical tasks. In this paper, we propose the first AGentic RObotics framework for SUrgical VIscoelastic DEbridement (AGRO-SUVIDE), the repeated removal of small fragments attached to a viscoelastic substrate. Leveraging the self-improving and coding capability of agents, AGRO-SUVIDE adopts a modular framework. Specifically, the demonstration analysis module automatically identifies recurring skills from a single expert demonstration, using both visual and kinematic information. The construction module then builds each skill, either as a procedural model-based skill the agent codes against a scaffolded library or as a model-free policy-based skill. At runtime, the monitoring module composes the skills into a loop-style graph sized to the number of fragments it observes, then verifies pre- and post-conditions of each skill to decide whether to advance or retry. We evaluate AGRO-SUVIDE through 340 physical trials on the da Vinci Research Kit (dVRK). AGRO-SUVIDE achieves an average single-fragment removal success rate of 85%, completing consecutive three-fragment removal at 60% and at 95% with one human intervention. It further generalizes to unseen five-fragment scenarios with an average success rate of 80% for single-fragment removal. Project page: https://surgical-robotics.github.io/AGRO-SUVIDE/