Meet Torso: a cute robot that learns how to interact with objects by exploring like a child. The Motivated Multi-Task Reinforcement Learning algorithm enables the robot to dynamically choose which task to explore. The robot picks up objects and tries to move them a new way. As this video shows, humans can provide additional guidance to the robot.
More like this ➡️ here
Intrinsically Motivated Goal Exploration, Automated Curriculum Learning and Emergence of Tool Use
This video shows how the robot learns by exploring objects around it. More info is available here.
*Our articles may contain aff links. As an Amazon Associate we earn from qualifying purchases. Please read our disclaimer on how we fund this site.