Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

General Video Game AI: a Multi-Track Framework for Evaluating Agents, Games and Content Generation Algorithms

Diego Perez-Liebana, Jialin Liu, Ahmed Khalifa, Raluca D. Gaina, Julian Togelius, Simon M. Lucas

TL;DR

The paper surveys General Video Game AI (GVGAI), detailing the VGDL-based framework, the multi-track competition (game playing, PCG, and learning), and the array of methods—ranging from MCTS and rolling-horizon evolution to hyper-heuristics and learning-based agents—that have been applied to single- and two-player GVGAI tasks. It highlights how GVGAI facilitates evaluating general AI across unseen games, emphasizes the role of forward models in planning tracks, and delves into content generation and educational uses. Key contributions include a taxonomy of approaches, empirical observations about what works across diverse games, and a roadmap of open problems (notably generalization, opponent modeling, and better feature extraction). The work underscores GVGAI's value as a research and education platform and outlines future directions, such as automatic game design, multi-agent tracks, and broader platform integrations to extend its reach and realism.

Abstract

General Video Game Playing (GVGP) aims at designing an agent that is capable of playing multiple video games with no human intervention. In 2014, The General Video Game AI (GVGAI) competition framework was created and released with the purpose of providing researchers a common open-source and easy to use platform for testing their AI methods with potentially infinity of games created using Video Game Description Language (VGDL). The framework has been expanded into several tracks during the last few years to meet the demand of different research directions. The agents are required either to play multiple unknown games with or without access to game simulations, or to design new game levels or rules. This survey paper presents the VGDL, the GVGAI framework, existing tracks, and reviews the wide use of GVGAI framework in research, education and competitions five years after its birth. A future plan of framework improvements is also described.

General Video Game AI: a Multi-Track Framework for Evaluating Agents, Games and Content Generation Algorithms

TL;DR

The paper surveys General Video Game AI (GVGAI), detailing the VGDL-based framework, the multi-track competition (game playing, PCG, and learning), and the array of methods—ranging from MCTS and rolling-horizon evolution to hyper-heuristics and learning-based agents—that have been applied to single- and two-player GVGAI tasks. It highlights how GVGAI facilitates evaluating general AI across unseen games, emphasizes the role of forward models in planning tracks, and delves into content generation and educational uses. Key contributions include a taxonomy of approaches, empirical observations about what works across diverse games, and a roadmap of open problems (notably generalization, opponent modeling, and better feature extraction). The work underscores GVGAI's value as a research and education platform and outlines future directions, such as automatic game design, multi-agent tracks, and broader platform integrations to extend its reach and realism.

Abstract

General Video Game Playing (GVGP) aims at designing an agent that is capable of playing multiple video games with no human intervention. In 2014, The General Video Game AI (GVGAI) competition framework was created and released with the purpose of providing researchers a common open-source and easy to use platform for testing their AI methods with potentially infinity of games created using Video Game Description Language (VGDL). The framework has been expanded into several tracks during the last few years to meet the demand of different research directions. The agents are required either to play multiple unknown games with or without access to game simulations, or to design new game levels or rules. This survey paper presents the VGDL, the GVGAI framework, existing tracks, and reviews the wide use of GVGAI framework in research, education and competitions five years after its birth. A future plan of framework improvements is also described.

Paper Structure

This paper contains 69 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. The GVGAI Framework
  3. The GVGAI Competition
  4. Game Playing Tracks
  5. Planning tracks
  6. Learning track
  7. PCG Tracks
  8. Level Generation Track
  9. Rule Generation Track
  10. Methods for Single Player Planning
  11. Basic Methods
  12. Tree Search Methods
  13. Evolutionary Methods
  14. Hybrids
  15. Hyper-heuristics / Algorithm Selection
  16. ...and 54 more sections

Figures (1)

  • Figure 1: Examples of VGDL games. From top to bottom, left to right: Butterflies, Escape, Crossfire and Wait for Breakfast.