Session:

UGW - Undergraduate Student Workshop

Title:

A Genetic Algorithms Approach to Learning Communication and Coordination in Simulated Robots

Authors:

Chris Sotzing

Abstract:

This project is motivated by an existing robot system for mapping unknown environments and attempts to improve its effectiveness through the use of genetic algorithms. Using the Webots simulator, the mapping system is created using simulated Khepera robots and a simulated environment. The robots are controlled by a supervisor agent that makes the high-level decisions about tasks for individual robots to complete to accomplish the mapping effort. This research investigates the ability of adding a GA learning component to the supervisor to improve its ability to coordinate the robotic agents.