Session:

UGW - Undergraduate Student Workshop

Title:

Evolving a Vision-Based Predator-and-Prey System for Two Robots with a Learning Classifier System

Authors:

Noah W. Smith

Abstract:

This paper describes an experiment to evolve a predator-and-prey system where the primary input for each robot is a linear camera. The method of learning is a Learning Classifier System. It builds on similar work implemented with an evolved neural network, in which both predator and prey behaviors were learned based mostly upon camera input. It is designed for implementation on two Khepera model robots with standard K213 Vision Turrets from K-Team, but the actual learning will be done in simulation.