Service Communautaire d'Information sur la Recherche et le Développement - CORDIS

FP5

NEUROBIT Résumé de rapport

Project ID: IST-2001-33564
Financé au titre de: FP5-IST
Pays: Italy

Neuro-robotic real-time closed-loop system

This is an experimental apparatus and real-time software application for bi-directional interaction of cultured neurons and a mobile robot. More in general, it can be used for closed-loop electrophysiology experiments at population or organism level. It is based on off-the-shelf hardware (PCs and acquisition boards). Design tools for rapid prototyping of real-time control and hardware-in-the-loop applications have been used to implement the real-time software application.

The control architecture is made of two PCs. PC1 runs a real-time operating system (QNX) and is responsible for (i) acquisition and sampling of electrophysiological signals, (ii) on-line spike detection and artifact blanking, (iii) decoding of the spike trains, (iv) robot control, (v) coding of robot proximity sensors signals and (vi) production of the pattern of stimuli that trigger the electrical stimulator. A second computer, PC2, connected to PC1 through an Ethernet link, is the experiment front-end. We used Simulink/ Real-Time Workshop (The Mathworks) and the RT-Lab package (Opal-RT) as hardware-in-the-loop development environments.

The control loop runs at 10kHz, and robot control is updated at 10Hz. This system allows simultaneous acquisition of neural signals from up to 32 recording sites. A video-capture system, running on a third PC, is capable of monitoring in real time the robot trajectories. Such information can be used in learning experiments, which require on-line assessment of robot performance. The software part includes a library of general coding and decoding modules for bi-directional neural interfaces. These modules are the basis for the implementation of the neural interface. As for coding, we provide schemes for proportional (i.e., intensity) as well as event coding. Blocks for stimulus generation include Poisson and non-leaky integrate-and-fire.

As regards decoding, there are command generation blocks based on estimates of the instantaneous firing rate. In particular, the library involves (i) spike detection, (ii) artifact suppression, (iii) firing rate estimation, and (iv) motor command generation modules. The final version of such library makes available a general tool for design and implementation of prototype neural interfaces.

Informations connexes

Reported by

DEPARTMENT OF COMMUNICATION, COMPUTER AND SYSTEM SCIENCES, UNIVERSITY OF GENOVA
Via Opera Pia 13
16145 GENOVA
Italy
See on map