Towards an Imitation System for Learning Robots

摘要

This paper proposes an imitation system for learning robots that attempts to model the functional role of pre-motor brain areas and in particular mirror neurons, i.e. neurons that are believed to form the fundamental basis for imitation in primates. Mirror neurons were found in the macaque monkey brain and are active during both observation and manual execution of an interaction. The mirror system (previous work) is concerned with the activation of structures in response to both the observation and the execution of interactions. Previous experiments show that observed interactions are represented, learnt, and in turn reproduced. These interactions however remain limited due to a rather crude modelling of neighbouring brain areas outwith the mirror system. The imitation system (current work) is an extension to the mirror system that brings neighbouring areas into play to exploit the learning capacity and diversity of the premotor cortex.