This is the preliminary version of an original cognitive model I developed to demonstrate or experiment with the very basics of an attract and avoid network "internal world model" navigation system. During human cognitive development this network is a central part of what provides children with the imagination to stretch out their arms and say "I can fly!" then make believe they are a bird or airplane.
The program starts with the untrained virtual critter chasing food (attractors) in a circle. When the food reaches the upper left of the arena it ends up in a place to avoid, which causes it to get very skittish and noticeably unsure what to do. The critter will then sometimes charge at it then chicken out before eating it all, behavior expected from an intelligent animal.
Since this is a one (time and place) frame model it is not expected to do overly well with the included hidden shock-zone test, expected to require a two frame system for better predicting and responding to future events. But the critter still does sometimes avoid getting zapped, and it's aware of places to stay away from even though it might then be too late to get out of the way. It's the next step towards a proper understanding of how real intelligence works, at all levels of biology.
The program was written in Visual Basic 6. Commented code is in the ".frm" files. Most controls have on screen information that appears when cursor is on top of it. The ReFeed button helps get it going again in situations where it needs some coaxing such as after finding a place to wait for the feeder to move on its own, which can happen when the location of bumps into the wall are saved in network memory. It is then aware of the arena wall and may find a place to confidently wait there, for hours or more.
Forming Sphere with Self-Learning Particle-Bots.
5 years ago