Robotic Grasping in Cluttered Environments
A particularly notable obstacle to the adoption of robotics in the everyday household is the difficulty of teaching robots to execute new tasks in unforeseen environments–this is currently a task only approachable by those with plenty of time and technical expertise. I am working on a system that aims to make teaching robots as simple as providing human demonstrations. It will use ideas from artificial intelligence to enable robots to generalize demonstrations to new situations and learn from failures as they interact with the world. My work will focus on deformable objects, such as rope and cloth, which are among the most difficult for robots to reason about and manipulate.
Message to Sponsor
- Major: Electrical Engineering and Computer Science
- Sponsor: Rose Hills Foundation
- Mentor: Pieter Abbeel, Electrical Engineering and Computer Science