Cell Assembly Robots: Vision and Cognitive Modelling
Chris Huyck
Middlesex University (www.cwa.mdx.ac.uk/CABot/CABot.html)
April 2009


We have dubbed our current EPSRC project the CABot project. The
project is to develop an agent based on simulated neurons that behaves
in a simulated environment.  We have already built two prototype Cell
Assembly roBots, CABot1 and CABot2.  CABot3 is currently being
developed.  These agents receive dynamic visual input from the
environment.  They are directed by natural language commands from a
user who directly controls another agent in the environment.  The
CABots maintain their own plans.  In CABot1, and hopefully CABot3,
this is all done solely with simulated neurons.  
In the talk, I will describe the agents and how they were developed.

The agents are large, so the talk will focus on the vision subsystems.
This includes a simulated retina of on-off receptors, a primary
visual cortex of line, angle and edge detectors, and texture
detection via grating cells.  These are used as input to the
object recognition subnet.

The natural language parser is a reasonable psycholinguistic 
model.  The use of the agents and the underlying neural technology
as the basis of a cognitive model will also be explored.