Grounded Relational Reinforcement Learning using Interactive Perception

This video demonstrates the use of Interactive Perception to gather and generalize manipulation knowledge.
On top of our real-world robust skill for extracting kinematic information about articulated objects,the robot uses Relational Reinforcement Learning to capture its experience. As a result, our robot is able to gather and generalize knowledge about the best ways to interact with new objects.

Category:

Science & Technology

Tags:

• Simulation
• Interactive Perception