Learning Concepts in Real World Embodies Agents
Service robots, personal assistance and other intelligent computers which operate in the real world inhabited by humans, deal with problems of interacting in a "natural" way with humans about their environment.
In this project we develop novel learning methods that can be used for concept formation in a human-robot interaction.
Research group members
S.H.G ten Hagen
This project is funded by NWO.
Our agent is a mobile robot equipped with omni-directional vision and it categorizes visual inputs into concepts related to target locations and places in the environment.
We use an hierarchical architecture that allows the robot to learn about concepts at different abstraction levels from ``move forward 3 centimeters'' to ``go to the robot lab''.