Neuro-robotics for intelligent information processing
The brain displays a unique functional plasticity, which can not be imitated by any existing artificial system, in particular its ability to adapt its performance to its experience. Challenged by this, the NEUROBIT project work concentrated on exploiting these extraordinary brain properties for controlling the sensorimotor behaviour of an artificial body that moves within a changing environment. This is expected to open new routes to use neurobiologically-engineered units for smarter information processing. Additionally, the study of the way that artificial systems code the information may also contribute in the development of new ones in the future. New tools and technologies have been realised to connect live, in vitro neurones' networks coming from mammalian embryos with external devices, such as robots in a bi-directional way. The living network is suitably chronically-coupled to a mobile robot resulting into a real-time connection. The experimental apparatus may be employed in closed-loop electrophysiology experiments at population or organism level. Built on off-the-shelf hardware tools for rapid prototyping of real-time control and hardware-in-the-loop, applications were implemented. It comprises two PCs, one running a real-time operating system and another one connected to the other through an Ethernet link used as the experiment front-end. The control loop runs at 10kHz and the same frequency is used for updating the robot control. The system is capable of collecting neural signals from up to 32 recording sites simultaneously. A third PC could also be used for video-capturing to monitor the robot trajectories in real-time, offering on-line assessment of robot performance. Implemented software involves a library of general coding and decoding modules for bi-directional neural interfaces serving as a general tool for design and realisation of prototype interfaces. For more information click at: http://www.bio.dibe.unige.it/neurobit.htm
