Generalizable Autonomy for Robot Manipulation

Explore generalizable autonomy for robot manipulation in this 47-minute lecture from MIT's Introduction to Deep Learning course. Delve into topics such as imitation learning, visuo-motor policies, off-policy reinforcement learning, compositional planning, and model-based RL. Learn about leveraging task structure, neural task programming, and the importance of data in robotics. Discover innovative platforms like RoboTurk for data collection. Gain insights from lecturer Animesh Garg of NVIDIA and the University of Toronto as he presents cutting-edge approaches to achieve adaptable and efficient robotic systems.

- Introduction
- Achieving generalizable autonomy
- Leveraging imitation learning
- Learning visuo-motor policies
- Learning skills
- Off-policy RL + AC-Teach
- Compositional planning
- Model-based RL
- Leveraging task structure
- Neural task programming NTP
- Data for robotics
- RoboTurk
- Summary