Updates from NVIDIA's Seattle Robotics Lab - Task and Motion Planning, Visuomotor Transformers, and Fine-grained Robot Manipulation
Explore cutting-edge robotics research from NVIDIA's Seattle Robotics Lab in this 55-minute colloquium talk. Delve into three key areas: combining Task and Motion Planning (TAMP) with teleoperation and imitation learning for improved visuomotor policies, the introduction of RVT (a multi-view transformer) for efficient object manipulation, and the IndustReal project's simulation-first approach to contact-rich assembly. Learn how these advancements are pushing the boundaries of robotic manipulation, from enhancing planning models to achieving successful sim-to-real policy transfers. Gain insights from research scientists Caelan Garrett, Ankit Goyal, and Iretiayo Akinola as they discuss their work on TAMP, 3D computer vision, and sim-to-real techniques in robotic manipulation.