nagle@well.sf.ca.us (John Nagle) (03/12/91)
The insect in Beer's "Intelligence as Adaptive Behavior" actually works as well as the book says it does. I'm running an implementation from Pat and Greg Williams (Rt. 1, Box 302, Gravel Switch, KY 40328, $10). I recommend this as a starting point for those interested in AI at the insect level. The program requires a PC-type machine with an EGA, and Turbo C/C++ if you want to modify it. You get source and the right to redistribute and modify. I'm using this as a starting point for work of my own. It's suprising how robust the insect nervous systems are. This isn't a PDP-type system; these are engineered control systems, and every neuron has a specific function. Yet you can disconnect most neuron connections and the insect still gets around, even though it may limp. Beer is operating at a nice level of abstraction. His insects are just complicated enough that problems like balance and stability start to appear, but simple enough that a big solid geometric modelling system isn't required to handle the simulation. And they're really fun to watch. They move like real bugs. The behavior when the insect hits an obstacle is very insect-like; it doesn't look mechanical at all. It even looks more insect-like than Rod Brooks' machines. Watching the six-legged creature struggle with a corner gives one the feeling that it's a living creature. This may be real progress. John Nagle
nagle@well.sf.ca.us (John Nagle) (03/15/91)
For everyone who asked about Beer's book, here's the information. The book is "Intelligence as Adaptive Behavior", by Randall D. Beer, Academic Press, 1990, ISBN 0-12-084730-2. John Nagle