Advancing Healthcare Robotics with Physical AI and Foundation Models

This session explores how physical AI, foundation models, and advanced simulation are addressing critical challenges in healthcare robotics. Industry leaders and researchers discuss innovations in surgical autonomy, data generation, and robotic systems designed to improve efficiency, safety, and accessibility in medical procedures.

■ Surgical Data and Simulation Innovations
- Leveraging video language models for efficient surgical data annotation and ontology creation
- Using Cosmos for synthetic data generation to bridge gaps between real-world and simulated environments
- Training surgeons and validating autonomous capabilities through high-fidelity simulations

■ Autonomous Robotic Surgery Frameworks
- Implementing imitation learning and hierarchical policies for complex tasks like suturing and tissue manipulation
- Utilizing large-scale datasets (OpenH, SRT) to train generalizable models across different robotic platforms
- Demonstrating autonomous performance in ex vivo models for procedures such as cholecystectomy

■ Next-Generation Robotic Systems and Deployment
- Optimizing arm deployment and workspace configuration using simulation and edge computing
- Developing humanoid architectures for hard-tissue surgery with synchronized vision and dual arms
- Integrating closed-loop feedback mechanisms for continuous improvement and patient safety

■ Future Directions and Q&A
- Addressing the growing demand for surgical services through automation and AI augmentation
- Discussing the role of Jetson Thor and runtime vs. training infrastructure for startups
- Exploring applications beyond the hospital, including home health care and infection control