Final
Project:
Solving
Mazes with Genetic
Programming
Forrest Sondahl
CS 460: Multi-Agent
Modeling (2005)
Northwestern University
UPDATE, 3/25/2010: The model files found on this site were written for an old version of NetLogo (3.0).
So here's an updated zip file where the models have been ported to NetLogo 4.1.
Summary:
The goal of this project was two-fold:
- First, to create a framework (called the GP Library) for doing genetic
programming in the NetLogo language.
- Second, to demonstrate the use of this framework through two
example models.
The two models both employ genetic programming to solve mazes. In
the first, "Genetic Programming Maze Marchers", we seek to develop
walking directions for agents to solve a particular maze. In the
second, "Genetic Programming Maze Rules", we seek to develop movement
rules which are applicable to solving mazes in general. More
information on each of these models is available on their web pages, or
if you download the models you can view it in the Information Tab.
- Genetic
Progamming Maze Marchers
- Genetic
Programming Maze
Rules
- Maze Files Required by the Models Above
- Genetic Programming
Maze Maker (download
link only)
- ( A simple model for designing/editing your own maze
files. )
- Genetic
Programming Demo
- ( A simple model demonstrating the use of the GP
Library framework. )
- ( A good jumping off point for creating your own
genetic programming model. )
For information regarding the development of this project, the design
decisions that were made, the various difficulties encountered
along the way, and the "evolution" of the project itself, the following
chronology should prove useful:
Back to Forrest's NetLogo Main
Page.
Created by: Forrest Sondahl
Email: forrest@northwestern.edu