QUAKE2AI

Quake2AI is an API created by Matt Parker to simplify creating a bot (an AI player) for the game Quake II. He used this API in his work on neurovisual control, which you can read about here [PDF].

Note:

• Quake II is a registered trademark of Id Software, Inc., of Mesquite, Texas.
• Although Id Software released the Quake II source code under an open-source license, they did not release the game media, such as the models used for the opponents. In order to use this API, you will need to acquire a copy of Quake II. (We found multiple copies on e-Bay when we started this work.)

The software components are linked in the relevant sections below.

The instructions are for installation on an Ubuntu Linux system.

---

Documentation

***This section is under construction. (Many links will not work until I finish it.)
***You can temporarily access the documentation here.

Quake2AI is a library and API for launching client-side Artificial Intelligence-controlled bots. It features the ability to read color information from the rendered screen (for visual input to the AI), as well as directly read location information for nearby entities (for more traditional approaches to game AI), and check for wall collisions between any two points in the map. A bot programmed to this interface has access to the same movement and control capabilities as a human playing the game would.

Quake2AI is written in C. There is also a Chicken Scheme interface and a Python interface, which rely on the C library, so be sure to install it even if you prefer one of these interfaces.

General Quake2AI

• Running a Quake II game server
• Quake II console commands
  Note: This is an external link. If it breaks, search for "Quake II" "console commands" and you will probably find another copy.

Quake2AI C Library

• quake2ai-0.22.tar.gz - The source code for the C library.
• Quake2AI Installation Guide
• Quake2AI API Reference

Chicken Scheme Interfaces (q2ai)

• chicken-2.6.tar.gz - Chicken Scheme version 2.6 (other versions might not work)
• q2ai.ss - The source code for the Scheme API.
• q2ai Installation Guide
• q2ai API Reference
• Tutorials:
  • Simple Bot
  • Neural Network

Python Interfaces (pyquake2ai)

• pyquake2ai.tar.gz - The source code for the Python API.
• Installation Guide
• pyquake2ai API Reference
• Tutorial:
  • Neuroevolution (Neuroevolution is the use of a genetic algorithm to train an artificial neural network.)

Running Quake2AI on the Cluster at UNR

Note: This is only possible for UNR students.

• Hints

Experiments: Neurovisual Control

Neurovisual control is the use of an artificial neural network (ANN) to control an autonomous intelligent agent, with inputs from a visual source such as a camera or a game screen, and outputs that directly control the agent's actions.

• Source code
  • Temporarily Unavailable; check back next week.
• Descriptions of the experiments
  • Retinal Layout
  • Lamarckian Neuroevolution
  • Backpropagation without Human Supervision
  • Cyclic Controller (not strictly neurovisual control, since the cyclic controller is not a neural network)