23 June 2005

Agent Directed Simulation

Interesting talk about agents and simulation given by Dr. Levent Yilmaz of Auburn University.

Reminds me of an idea I had a while back where a network of nodes, representing a map through a jungle, would be traversed by a bunch of agents simulating monkeys. Different nodes would have different benefits - such as places to sleep, places to eat, places to breed, or places to avoid like tiger clearings. The monkeys would move and while they were in a clearing (node) with each other, they would be able to share some limited information with each other. As the simulation operated over time, there would different goals that motivated the agents for different periods of time. A sleep time would drive the monkeys towards the good sleeping place; periodically there would be a "to breed" sleep time which would drive the monkeys to the boinky-place; etc. As the monkeys traversed and learned the map of nodes (the jungle) they would build up a simple (also incomplete and unreliable) knowledge and memory of paths.

The goal of all this monkey business would be to see if the agents would exhibit some behavior over time, such as learning optimum paths between various locations.

Hmmm. A project for the future.

Tags: ,


Post a Comment

<< Home