breve-Monsters

Copyright (c) 2010 The University of Texas at Austin
All rights reserved.
Refer to LICENSE.txt for detailed license information.

WEBPAGE

http://nn.cs.utexas.edu/?brevemonsters

ABOUT

breve-Monsters is a collection of software for simulating several
3D video games featuring NPC monsters, as well as a collection
of algorithms for evolving neural network controllers for the monsters
in these games. The code was developed by Jacob Schrum (schrum2@cs.utexas.edu).
The primary method used for evolving controllers is NSGA-II, a 
multiobjective evolutionary algorithm, details about which can be found in:

K. Deb, S. Agrawal, A. Pratab, and T. Meyarivan, "A fast elitist nondominated
sorting genetic algorithm for multi-objective optimization: NSGA-II". 
PPSN VI, pp. 849-858, 2000.

NSGA-II is enhanced to improve performance in agent control problems, such
as the included video games, via methods described in the several papers
available at:

http://nn.cs.utexas.edu/?brevemonsters

This software package is called "breve-Monsters" because it runs using the
multi-agent simulation environment "breve", developed by Jon Klein, and described
in detail in:

Klein, J. 2003. BREVE: a 3D environment for the simulation
of decentralized systems and artificial life. In ALIFE VIII, 
329-334. Cambridge, MA, USA: MIT Press.

The breve simulation environment is required to run this source code,
and should therefore be downloaded from:

http://www.spiderland.org/

In particular, this software was developed for use with "breve CLI 2.7.2 for Linux".
While this code may work with other versions of breve, such functionality is not
guaranteed. For those new to breve, it is recommended that you attempt executing some
of the basic examples provided with breve before attempting to run this software package.

For further instructions on how to run this code, see TUTORIAL.txt.

DIRECTORY CONTENTS

Agent.tz		- Agent abstraction
BehaviorState.tz	- Used by the Behavioral Diversity method of fitness-based shaping
CommandLineHandler.tz	- Handles commandline parameters
Constants.tz		- Simulation constants
Environment.tz		- Model of environment
EvolutionMode.tz	- Stores bundles of commandline parameter settings to make simulation execution simpler
FreeNetGa.tz		- Handles genetic operations on neural networks
FreeNetwork.tz		- Neural networks with arbitrary topologies
GraphicsHandler.tz	- Handles graphics
HUD.tz			- Used for picture within picture functionality (see NetViewer.tz)
LICENSE.txt		- Licensing information for this software
Log.tz			- Logs data during simulation execution
Math3D.tz		- Helper functions
ModeArbitrator.tz	- Used by multi-modal networks
Monster.tz		- Monster/NPC abstraction
NSGA2.tz		- Multiobjective Evolutionary Algorithm
NetIO.tz		- Reads and writes networks from/to file
NetViewer.tz		- Allows network visualization
Player.tz		- Abstraction for human players and static strategy bots
README.txt		- This file
Selection.tz		- Abstraction for selection methods
Sensor.tz		- Monster input sensor
Simulation.tz		- Control class for simulation. Run this file with breve.
TUTORIAL.txt		- Help getting started as well as examples for recreating results from previous papers
Triangulate.tz		- Helper functions
UnitTest.tz		- Verifies correct functioning of certain key components
ZScoresGA.tz		- Weighted-sum-based selection method. Only works with Battle Domain.

FOR MORE INFORMATION CONTACT

schrum2@cs.utexas.edu