Here the iCub has finished solving the "uncover the toy" problem through cooperation with the human user. The black and white toy was under the blue and brown box, and together, the human lifted the box while the robot then took the toy and placed it on the left, with the human finally replacing the box in the middle. This is illustrative of the kinds of cooperative tasks performed by human children that are investigated at MPG.

This image displays elements of the integrated system. As the robot observes the interaction work surface, its internal representation is displayed on the screen on the right of the figure.

This first video (ORO-SPARK-Grasp.MOV) shows the user asking the robot to grasp the toy, while at the same time he covers the toy, rendering it invisible. The robot detects that it cannot see the toy and thus cannot perform the grasp. The human then uncovers the toy, it becomes visible (as demonstrated by its appearance in the display of the robot's representation of the workspace), and the robot then grasps it.

This video (ZOOMshard-putGOOD.MOV) displays the use of a new action that the robot has just learned (put, consisting of concatenated grasp and release) within a previously acquired shared plan that involves the uncover the toy task.

iCub - "put OBJECT at LOCATION" command

A crucial aspect of the CHRIS architecture is that the integrated system is portable to the different robot development sites. A second and perhaps more exciting aspect, is that as information is acquired by one robot through human interaction, that knowledge can be shared with other robots that embody the CHRIS architecture. In this video we wee the "super internet mind transfer from INSERM to IIT". More technically, the ability to perform the "put OBJECT at LOCATION" command was learned by the icub in Lyon, where it could then be used in a shared plan for "uncover the toy". Later, in Genoa, the IIT iCub is able to use that information which was learned and acquired in Lyon, in order to perform the new task in Genoa. Thus, knowledge acquired by one robot can be used by another (iit_inserm_exp0.wmv).

CHRIS 2nd year demo (full version)

BERT Videos

Recognizing Familiar Learning New and Memorizes Action Sequence

Object Learning

Learning Touch Action

Learning Get and Put Actions

BERT Points