A framework for cognitive robots to learn behaviors through imitation and interaction with humans

Abstract

This paper proposes a general learning framework for robots to learn behaviors through imitation and interaction. A modified codebook based method is used for robots to segment and recognize new objects in the environment. Task related semantic information is learned by robots through the speech communication with humans. Dynamic Movement Primitive method is used to generate similar behaviors to complete similar but slightly different tasks. Experimental results are given to verify the effectiveness of this framework.