Teaching new tricks to a robot learning to solve a task by imitation

Abstract

Learning by imitation represent a useful and promising alternative to programming robots. This paper presents an approach for learning by imitation based on two functional elements used by humans to understand and perform actions. One is the representation of the body that contains information of the body's capabilities. The other one is a snapshot of the body and its relation with the environment at a given instant. These elements are believed to interact to generate among other abilities, the ability to imitate. Learning new tasks is then achieved by imitating the actions' goal as demonstrated. The experimental results with a humanoid robot are presented to validate the approach.