NVIDIA Research presented eight new papers at the International Conference on Robotics and Automation (ICRA) detailing advancements in simulation-to-real transfer for robotics. These papers showcase methods for robots to perceive, reason, plan, and act in dynamic environments, moving beyond scripted automation. Key developments include ScheduleStream for multi-arm coordination, COMPASS for generalizable navigation policies, Grasp-MPC for adaptive object grasping, and Deformable Cluster Manipulation for handling tangled materials. Additionally, SPARR and Refinery frameworks improve precision assembly, while PEEK and SEAL enhance action models for complex tasks and reasoning. AI
IMPACT These advancements in sim-to-real transfer and embodied AI are crucial for developing more adaptable and reliable robots, potentially accelerating their deployment in complex, real-world scenarios.
RANK_REASON The cluster details multiple research papers and frameworks presented by NVIDIA Research at a major robotics conference, focusing on technical advancements in simulation-to-real transfer.
- Carnegie Mellon University
- cuRobo
- Deformable Cluster Manipulation
- GraspGen
- Grasp-MPC
- International Conference on Robotics and Automation (ICRA)
- NVIDIA
- NVIDIA Isaac Lab
- NVIDIA Jetson
- NVIDIA Omniverse NuRec
- NVIDIA Research
- PEEK
- Refinery
- ScheduleStream
- SEAL
- University of Sydney
- University of Utah
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
