Imitation Learning in Robotics
Nowadays robots are able to perform complex tasks in structured
worlds by the implementation of sophisticated software systems and
the control of skilled human operators or programmers. However,
high complexity and task specificity are limiting factors for
large scale uses. A possible solution is the
implementation of imitation mechanisms. In this way, human
user can explain the execution of a task by showing it to the
robot without any knowledge of complicated programming
languages.
This fascinating research area involves different fields, from Robotis and Artificial Intelligence, to Neuroscience and Psychology. Basically, three types of imitation approaches have been proposed:
- Sensory-motor imitation: aims to the exact reproduction of the perceived visual system onto the robot
- Goal-level imitation: aims to the reproduction of the effects of actions performed
- Social imitation: aims to predict and explain other’s emotions, behaviors and other mental states, and to formulate appropriate responses based on this understanding
The research conducted at the RoboticsLab in Palermo is focused on the goal-level imitation. Our ultimate goal is to implement an imitating mechanism which will lead to the development of artificial agents, endowed with powerful perceptual, cognitive and action capabilities. In this intent we proposed ConSCIS (Conceptual Space based Cognitive Imitation System), a cognitive architecture for Imitation Learning which tightly links low-level data processing with high-level knowledge representation.
Videos
This section shows our system at work. For more information see our recent articles, and consult the following page. NOTE: Some videos require camtasia codecs available here.
Robot observes a human while doing some tasks and builds its internal model
(download video#1) (download video#2) (Output of the system)
Robot uses its internal representation to find the best task to imitate
Robot imitates
Code
You can (download) the code of the scoring algorithm used for imitation (the code is in Prolog language).
Publications
International Journals & Transactions
A. Chella, H. Dindo, I. Infantino, “Imitation Learning and Anchoring through Conceptual Spaces”, in Applied Artificial Intelligence, Volume 21, Issue 4 & 5, April 2007, pp. 343-359
A. Chella, H. Dindo, I. Infantino, “A Cognitive Framework for Imitation Learning”, in Robotics and Autonomous Systems, 54(5), pp. 403-407, May 2006 (pdf)
H. Dindo, A. Chella, I. Infantino, “Anchoring by Imitation Learning in Conceptual Spaces”, Lecture Notes in Computer Science, vol. 3673/2005, pp. 495-506 (pdf)
I. Infantino, A. Chella, H. Dindo, I. Macaluso, “A Cognitive Architecture for Robotic Hand Posture Learning”, in IEEE Transactions on Systems, Man, and Cybernetics: Part C, 35, pp. 42-52, February 2005 (pdf)
A. Chella, I. Infantino, H. Dindo, I. Macaluso, “A Posture Sequence Learning System for an Anthropomorphic Robotic Hand”, in Robotics and Autonomous Systems, 47(2-3), pp. 143-152, June 2004 (pdf)
Book Chapters
H. Dindo, I. Infantino, “Representation, Recognition and Generation of Actions in the Context of Imitation Learning”, in Springer Tracts in Advanced Robotics - STAR, pp. 65-77, Berlin/Heidelberg, Springer, Germany, 2006 (pdf)
International Conferences
A. Chella, H. Dindo, I. Infantino, “Learning high-level tasks through imitation”, in Proc. of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Beijing, China, October 9-14, 2006
A. Chella, H. Dindo, I. Infantino, “Learning high-level manipulative tasks through imitation”, in Proc. of the 15th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN), University of Hertfordshire, Hatfield, United Kingdom, September 6-8, 2006
A. Chella, H. Dindo, I. Infantino, “A Cognitive Framework for Learning by Imitation”, in Proc. of the IEEE ICRA Workshop on The Social Mechanisms of Robot Programming by Demonstration, Barcelona, April 22, 2005 (pdf)
I. Infantino, A. Chella, H. Dindo, I. Macaluso, “A Posture Sequence Learning System for an Anthropomorphic Robotic Hand”, in Proc. of the IEEE/RSJ IROS Workshop on Robot Learning by Demonstration, Las Vegas, October 31, 2003 (pdf)
A. Chella, I. Infantino, H. Dindo, I. Macaluso, “Visual Control of a Robotic Hand”, in Proc. of The 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), Las Vegas ,October 27-31, 2003 (pdf)
Infantino, A. Chella, H. Dindo, I. Macaluso, “Human Hand Posture Reconstruction for a Visual Control of an Anthropomorphic Robotic Hand”, in Proc. of the Third IEEE International Conference on Humanoid Robots, TU München and U Karlsruhe, Germany, October 1-3, 2003 (pdf)
National (Italian) Conferences
I. Infantino, A. Chella, H. Dindo, I. Macaluso, “A Cognitive System for Human Interaction with a Robotic Hand”, in Proc. of the First RoboCare Workshop, Rome, Italy, October 30, 2003.