Objective
Recent research suggests that imitation, as evident in the individual development of human cognition and motor skills, is a paradigm that can endow artefacts with the ability for skill growth and life-long adaptation without programming. To demonstrate the power of this paradigm, we propose to develop a fully functional artefact with complex cognition and behaviour: a living artificial hand that learns structurally new complex action sequences by imitation.
This will be achieved by joining brain research cognitive psychology, neuroinformatics and robotics, to:
i) explore the neuropsychological, neurophysiological, functional mechanisms of imitation learning in man and monkey;
ii) design a bio-plausible computer model of imitational learning;
iii) realise he full visuo-motor system. Results are a paradigm of imitation in man, a generic computer model and robot implementation, applicable to arte-facts that learn from and cooperate with humans, service robots and novel prosthetic aides.
OBJECTIVES
Living artefacts will critically depend on mechanisms of action and task-level learning: for understanding and achieving missions, for cooperation with humans, for adapting to the environment. Imitation holds the potential to become a much more powerful means for learning complex cognition-action sequences than classical approaches because it much reduces the state-action space to be searched for motion generation.
We aim at:
- exploring the functional and neuropsychological mechanisms of imitation learning;
- revealing the neurophysiological structures for finger and hand movements in man and monkey;
- designing a formal dynamic model of imitation learning;
- constructing the fully functional artefact (visuo-motorsystem);
- implement imitational learning on the artefact;
- disseminate results back into neuroscience and infotech;
- suggest applications of the new methods and models to artefacts with many degrees of freedom that cooperate with humans as well as new generations of prostheses.
DESCRIPTION OF WORK
The demanding goals can only be achieved by joining researchers from cognitive psychology, neurophysiology, cognitive neurology, neuroinformatics and robotics. Accordingly, the workpackages fall into several categories: basic research, modelling, prototype development and technical implementation, dissemination, coordination.
Concerning the functional/behavioural aspects of object grasping we shall:
(1) test critical features of imitative object grasping in healthy persons and in patients with ideomotor praxia and cerebellar deficits;
(2) construct a dynamic and a functional model of imitation.
For exploring the neurophysiological structures for object grasping and imitation in monkey we shall:
(1) describe the functional and motor proper-ties of prefrontal neurons (PF) in response to object grasping movements;
(2) develop a neurophysiological model of the basic visuo-motor organization of the pre-motor cortex (F5) and PF.
For revealing the neurophysiological mechanisms of grasping and imitation in man, we shall:
(1) carry out MEG recordings in healthy subjects during observation and imitation of action;
(2) describe the time course and topography of MEG responses of these experiments using dipole reconstruction techniques.
To develop a computer model for structural learning of multi-fingered grasping we shall:
(1) define an integrating model architecture a close mutual coupling between the perceptual and motor level;
(2) implement local learning rules that ensure the correct mapping between the available sensory information, movement primitives and behavioural sequences.
To implement the artefact on the basis of a robot system with advanced sensor and motor skills we shall:
(1) build a perceptual system with emphasis on robustness and design atomic behaviours for the multi-fingered hand according to the neurophysiological findings;
(2) integrate the dynamic model and the learning algorithms into a fully operational artefact.
Fields of science (EuroSciVoc)
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
- natural sciencesbiological sciencesneurobiology
- natural sciencesphysical sciencesastronomyplanetary sciencesplanetary geology
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringrobotics
- medical and health sciencesbasic medicineneurology
- social sciencespsychologycognitive psychology
You need to log in or register to use this function
Call for proposal
Data not availableFunding Scheme
CSC - Cost-sharing contractsCoordinator
80333 MUENCHEN
Germany