General Video Game AI: A Multitrack Framework for Evaluating Agents, Games, and Content Generation Algorithms

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.

Rinascimento: Optimising Statistical Forward Planning Agents for Playing Splendor

“…A number of game-based frameworks have been designed and implemented for different research purposes. The most classic ones include, but not limited to: GVGAI [3], microRTS Framework [6], Mario AI Framework [7], AI Bird Framework (AIRBIRDS) [8]. Each one has one or more relative competitions periodically run at major Game AI conferences.…”

Project Thyia: A Forever Gameplayer

“…The core part of the system will be the actual game-player: the algorithm which is able to play unknown games. We choose to base the game-player on a planning method, due to their flexibility, adaptibility and lack of training necessary, as well as high performance across multiple games [3]. The downsides of these methods are two-fold: they do not usually learn between games (i.e.…”

“…Although a lot of important advances in AI have been made in games, and games are still actively used as testing environments for AI, algorithms are only able to solve (in this case, win or achieve a high score) a subset of existing games [3]. Planning and learning algorithms alike are unable to act in an intelligent manner in all given games, unless they use human-tailored heuristics or features (often game-specific).…”

Reasoning in Highly Reactive Environments

“…With respect to existing literature [23,26], our approach promises to be better scalable to real contexts with higher size mazes; experiments aimed at confirming that are currently ongoing. Finally we developed two plugins based on our generation technique, which were respectively deployed as an asset available in the Unity development and in the GVGAI [24] frameworks [8].…”