Authors:
Frédéric Koriche, Sylvain Lagrue, Éric Piette, Sébastien Tabary

Venue:
Revue d’intelligence artificielle, numéro spécial « IA des jeux informatisés », Lavoisier, 2017

Volume and pages:
Volume 31, no 3, pp. 307–336

DOI:
10.3166/RIA.31.307-336

Topics:
General Game Playing, stochastic CSP, MAC-UCB, symmetry detection, competitive AI

Links: PDF · Proceedings

This article describes WoodStock, the first general game player modeling each game from the General Game Playing (GGP) by a stochastic constraint network (SCSP). Each action played is decided by the resolution of this last one by the algorithm MAC-UCB.

After the translation of an instance described in Game Description Language (GDL) in a network representative of the state of the game at any time, WoodStock solves each state by the maintening arc-consistency algorithm (MAC) iteratively guided by the bandit-based stochastic sampling (UCB) of the next states.

Thanks to this algorithm, WoodStock is since march 2016, the leader of the GGP Tiltyard continuous tournament. Moreover, in its last version exploiting the game symmetries finding by the constraint symmetry detection, the search space associated with a game is significatively reduced. With that, WoodStock is now the GGP champion after its victory at the International General Game Playing Competition 2016 (IGGPC 2016).

This article presents WoodStock, the first complete implementation of the constraint-based General Game Playing approach developed in this line of research.

It details the full pipeline, from the translation of GDL games into stochastic constraint networks to the use of MAC-UCB under real competition constraints, and shows how symmetry detection can further reduce the search space.

This paper is closely connected to the PhD thesis and to the competitive results that established WoodStock as a leading General Game Playing system.

Koriche, F., Lagrue, S., Piette, É., Tabary, S. (2017). WoodStock : un programme-joueur générique dirigé par les contraintes stochastiques. Revue d’intelligence artificielle, 31(3), 307–336. Numéro spécial « IA des jeux informatisés », Lavoisier. DOI: 10.3166/RIA.31.307-336

@article{koriche2017woodstock,
  author    = {Koriche, Fr{\'e}d{\'e}ric and Lagrue, Sylvain and Piette, {\'E}ric and Tabary, S{\'e}bastien},
  title     = {WoodStock : un programme-joueur g{\'e}n{\'e}rique dirig{\'e} par les contraintes stochastiques},
  journal   = {Revue d'intelligence artificielle},
  volume    = {31},
  number    = {3},
  pages     = {307--336},
  year      = {2017},
  publisher = {Lavoisier},
  doi       = {10.3166/RIA.31.307-336}
}