Trends in Cognitive Sciences
ReviewIs imitation learning the route to humanoid robots?
Section snippets
Movement imitation
Movement imitation is familiar to everybody from daily experience: a friend demonstrates a movement, and immediately we are capable of approximately repeating it. For the purpose of this review, only visually mediated imitation will be considered, although, at least in humans, verbal communication can supply important additional information. From the viewpoint of motor learning, a teacher’s demonstration as the starting point of one’s own learning can significantly speed up the learning
Imitation from the viewpoint of behavioral and cognitive sciences
In infant and animal studies, the ability to imitate is usually concluded from the subject’s increased tendency to execute a previously demonstrated behavior. However, other causes can equally account for a higher probability of the subject’s behavior, in particular priming, emulation, and response facilitation (see Glossary); such causes are not to be mistaken for true imitation8, 9. True imitation is present only if (1) the imitated behavior is new for the imitator, (2) the same task strategy
Imitation from the viewpoint of neuroscience and cognitive neuroscience
An essential prerequisite for imitation is a connection between the sensory systems and the motor systems such that percepts can be mapped onto appropriate actions. This mapping is a difficult computational process as visual perception takes place in a different coordinate frame from motor control. This process is also more complex than pure object recognition because it requires integration of multiple objects (i.e. several limbs), their spatial relationships, their relative and absolute
Symbolic approaches to imitation learning
At the beginning of the 1980s, the idea of imitation learning started to find increasing attention in the field of manipulator robotics as it seemed to be a promising route to automate the tedious manual programming of robots. Inspired by the ideas of artificial intelligence, symbolic reasoning was commonly chosen to approach imitation, as outlined in the following discussion (e.g. 48, 49, 50, 51, 52). During a training phase, several example movements were generated under manual robot control
Is imitation learning the route to humanoid robots?
In the introduction of this article, we appealed to a pragmatic view of imitation learning as a means to speed up learning in complex high dimensional motor systems, such as humanoid robots. This view emerged from the lack of theories of motor learning that are able to work efficiently in high dimensional spaces. Interestingly, the apparently simple idea of imitation opened a Pandora’s box of important computational questions in perceptual motor control. None of the approaches described in this
Outstanding questions
- •
Learning perceptual representations: How can appropriate representations of the identity and movement of others be developed in an automated fashion from visual input? Is it necessary that such representations develop simultaneously with the motor representations?
- •
Movement primitives: Is there a basic set of primitives that can initialize imitation learning? How complex are the most elementary primitives in this set? How can new primitives be learned, and old primitives be combined to form
Acknowledgements
Earlier versions of this paper were significantly improved as a result of comments from Mitsuo Kawato, Nicolas Schweighofer, and three anonymous reviewers. Research described in this paper was made possible by Award No. 9710312 of the National Science Foundation, the ERATO Kawato Dynamic Brain Project funded by the Japanese Science and Technology Cooperation, and the ATR Human Information Processing Research Laboratories.
References (90)
Does immediate imitation influence long-term memory for observed actions?
J. Exp. Child Psychol.
(1991)- et al.
Fixation behavior in observation and imitation of human movement
Cognit. Brain Res.
(1998) Premotor cortex and the recognition of motor actions
Cognit. Brain Res.
(1996)- et al.
Hierarchical organization and functional streams in the visual cortex
Trends Neurosci.
(1983) Visual analysis of body movements by neurones in the temporal cortex of the macaque monkey: a preliminary report
Behav. Brain Res.
(1985)- et al.
Motion sensitive cells in the macaque superior temporal polysensory area: response discrimination between self-generated and externally generated pattern motion
Behav. Brain Res.
(1996) Grasping objects: the cortical mechanisms of visuomotor transformation
Trends Neurosci.
(1995)Do imagined and executed actions share the same neural substrate?
Cognit. Brain Res.
(1996)- et al.
Language within our grasp
Trends Neurosci.
(1998) - et al.
Mirror neurons and the simulation theory of mind-reading
Trends Cognit. Sci.
(1998)
Computational approaches to motor control
Trends Cognit. Sci.
Multiple paired forward and inverse models for motor control
Neural Networks
Getting to know each other: artificial social intelligence for autonomous robots
Robotics and Autonomous Systems
A Kendama learning robot based on bi-directional theory
Neural Networks
A tennis serve and upswing learning robot based on bi-directional theory
Neural Networks
Internal models in the cerebellum
Trends Cognit. Sci.
Modeling parietal-premotor interactions in primate control of grasping
Neural Networks
Bio-robotic approaches to the study of motor systems
Curr. Opin. Neurobiol.
Modeling and Control of Robot Manipulators
Current and future perspective of Honda humanoid robot
The development of Honda humanoid robot
Reinforcement learning: a survey
J. Artif. Intell. Res.
Reinforcement Learning: An Introduction
Learning to predict by the methods of temporal differences
Machine Learn.
Learning tasks from a single demonstration
Cultural learning
Behav. Brain Sci.
Mental Evolution in Animals
Animal intelligence: an experimental study of the associative process in animals
Psychol. Rev. Monogr.
Play, Dreams, and Imitation in Childhood
Imitation of facial and manual gestures by human neonates
Science
Newborn infants imitate adult facial gestures
Child Dev.
The Thinking Ape
Imitative learning of actions on objects by children, chimpanzees, and enculturated chimpanzees
Child Dev.
A comparison of imitation strategies in observational learning of action patterns
J. Motor. Behav.
Preliminaries to a theory of action with reference to vision
Information, Natural Law, and the Self-Assembly of Rhythmic Movement
Origins and early development of perception, action, and representation
Annu. Rev. Psychol.
Infant’s understanding of people and things: from body imitation to folk psychology
Explaining facial imitation: a theoretical model
Early Development and Parenting
Neural Networks for Pattern Recognition
Social signals analysed at the single cell level: someone is looking at me, something touched me, something moved!
Int. J. Comp. Psychol.
Frameworks of analysis for the neural representation of animate objects and actions
J. Exp. Biol.
Integration of form and motion in the anterior superior temporal polysensory area (STPa) of the macaque monkey
J. Neurophysiol.
Functional organization of inferior area 6 in the macaque monkey: II. Area F5 and the control of distal movements
Exp. Brain Res.
Cited by (1232)
Imitation learning enabled fast and adaptive task scheduling in cloud
2024, Future Generation Computer SystemsAn online reinforcement learning approach for HVAC control
2024, Expert Systems with ApplicationsTask-unit based trajectory generation for excavators utilizing expert operator skills
2024, Automation in ConstructionPiagetian experiments to DevRobotics
2024, Cognitive Systems ResearchHuman-inspired autonomous driving: A survey
2024, Cognitive Systems Research