Policies trained from data instead of hand-written controllers.
Low-cost bimanual teleoperation hardware and ACT policy learning.
