Résumé des résultats (Abstract)
(Italien)
|
NEUROBIT intends to develop the tools and technologies for connecting portions of living nervous tissue (cultured, and kept alive in-vitro) bi-directionally with external devices (i.e., a robot) in order to teach the biological component of the hybrid system to process information in a goal-oriented way. This will open to new opportunities of using engineered neurobiological units to process information in a natural and therefore, more intelligent way, and to study how simple, natural but at the same time artificial (i.e., bioartificial) systems code the information. The somewhat revealed, neural code could then also be utilised as a source of inspiration for new artificial systems (i.e., new algorithms and new neuromorphic physical devices).
Objectives:
Functional plasticity seems incomparably greater in the brain than in any artificial system. Especially its capability to adapt its performance to previous experience is a major challenge to scientists for understanding the underlying principles and operations, a necessary step toward implementing these same principles in physical devices. The main goal of the project is to take advantage of these unique plastic properties to control the sensorimotor behaviour of an artificial body moving in a changing environment. As living highly adaptable systems, we will use networks of neurones kept alive, in-vitro, out of the brain of mammalian embryos. The living network will be real-time connected
Work description:
A system based on a cultured neuronal network chronically coupled to a robot will be set-up. The bioartficial neuronal device will be used to drive the autonomous robot, equipped with one or more sets of sensors (light, acoustic, range, etc). Sensors will drive the stimulation of the microtransducer inputs, whereas the neuronal population activity from the microtransducer outputs will be used to control the wheels of the robot. A CCD will be used to generate an electrical representation of the position of a light source. Dedicated algorithms and electronics will be devised to real-time generate adequate stimulation patterns in the channels identified as inputs of the neural nework. A recording system will sense neuronal electrical activity and identify distinct patterns. Conversion of signal recorded from channels that were previously recognised, as outputs will be used to control a robot on which a CCD camera will be mounted. Thus, a closed feedback loop encompassing a neuronal network will be established.
|
