|
Design mechanisms for sensory-motor coordination and multimodal integration in physical and social interactions| old_uid | 13694 |
|---|
| title | Design mechanisms for sensory-motor coordination and multimodal integration in physical and social interactions |
|---|
| start_date | 2014/03/27 |
|---|
| schedule | 10h30-12h |
|---|
| online | no |
|---|
| location_info | EuroMov, salle EJM |
|---|
| summary | My main research interest is to discover the design principles of motor, cognitive and social development using synthetic approaches, such as robots or computational models. I put forward the key concepts of enaction and complex systems theory, which are that (1) action plays a key role for structuring perception and cognitive development (i.e., the body “shapes” the mind) and that (2) the complex systems framework is capable to grasp its dynamical nature. Among others, I propose that the mechanisms that give rise to emergent processes in dynamical systems, such as synchronization and self-organization are essential for modelling sensorimotor control and neural development in robots and among these mechanisms, the detection of timing is one essential mechanism for integrating the micro-scale of neurons' dynamics to the macro-scale of motion coordination. I will present three researchs themes. First, how central pattern generators in the spinal cords can be modeled as chaotic controllers that synchronize to the body dynamics and decrease to the overall dimensionality of the control problem ; the so-called Bernstein problem. This framework is used in robots simulation and in a real robot in which rhythmical patterns emerge from the physics of the mechanical system. Chaos control has some implications on the design of robots with high-degrees of freedom and on the comprehension of the controllability of the human body. Second, I will present some models of the parieto-motor circuits responsible for the development of self-perception, the spatial representation of the body, and inter-subjecity. I will show how associative maps can emerge from sensory-motor coordination through the hebbian-like mechanism of spike-timing dependent plasticity. We will show how these units can confer to the system some properties similar to the mirror neurons system : the simulation of one modality from the other. In a visuo-tactile simulation of a infant face, I will show how the emergence of mimicry in neonates may be based on such feature. Third, the spatial representation of the body is not static but plastic. In situation of social interactions, of tool-use, or of motion, our receptive fields dynamically adapt to the new context ; I will present how the phenomenon known as the gain-field effect in parieto-motor neurons can confer to neural networks multimodal integration irrespective to its normal reference frame. Such feature permits to combine visuo-audio signals into one reference frame in a robotic head, and to solve the correspondence problem during tool-use and social interactions. |
|---|
| responsibles | Hoffmann, Marin |
|---|
| |
|