Perceptual Robot Learning - David Held, CMU

Explore perceptual robot learning in this 57-minute lecture by David Held from CMU, presented at the Paul G. Allen School. Discover how integrating computer vision and robot planning can enhance robots' capabilities to interact with non-rigid, transparent, and reflective objects found in daily life. Learn about novel perception and planning algorithms for manipulating cloth, articulated objects, and challenging materials. Gain insights into designing state representations that consider both perception and planning for improved performance on complex robotic tasks. The lecture covers topics such as cloth folding, articulated object manipulation, and cloth grasping, offering a comprehensive look at the future of robotics and its potential to handle diverse objects in real-world scenarios.

Introduction
What is a robot
Can we just use a larger land vector
First approach
Overview
Results
Clarification
Folding Cloth
Articulated Object Manipulation
Cloth Grasping
Insights
Questions
Flow
Representation