Research software
Software and systems
My work includes the design and development of general game systems, AI research tools, and computational resources for the analysis of games.
Ludii General Game System
Ludii is a complete general game system, based on the ludemic approach and a class grammar mechanism, written in Java. It is designed to play, evaluate, and design a wide range of games, including board games, dice games, and many other combinatorial and traditional games.
Ludii makes it possible to model a broad range of traditional strategy games within a single playable database. Its game description language is typically shorter and easier to read than earlier general game description languages, while also enabling efficient execution.
The system includes built-in AI controllers and supports custom AI implementations. Instructions and examples for creating custom Ludii agents are available in its dedicated repositories and documentation.
This software is one of the major outcomes of the ERC-funded Digital Ludeme Project and has already supported an extensive body of publications in general game playing and game AI.
Resources:
Project website
Source code
WoodStock
WoodStock stands for With Our Own Developper STOchastic Constraint toolKit. It is a constraint-based general game player for General Game Playing (GGP).
It consists on two parts : (i) A network player spy-ggp communicating with the game manager from the GGP server ; (ii) A parser GDL to SCSP and of the MAC-UCB's algorithm (see). It's a new stochastic constraint-based approach combined with a Monte Carlo method for GGP.
When its first participation at a GGP's international competition organized by Stanford on ggp.org, WoodStock obtained the second place during the qualification day and finished third during the final day. (Open Tiltyard). Since March 2016, WoodStock is leader on the continuous play match on the Tiltyard GGP server. The August 25 2016, WoodStock won the International General Game Playing Competition (IGGPC'16) and became the GGP's world champion.
Resource:
WoodStock project page