Optimal Player Information MCTS applied to Chinese Dark Chess

Alpha-beta and Monte-Carlo Tree Search (MCTS) are two powerful paradigms useful in computer games. When considering imperfect information, the tree that represents the game has to deal with chance. When facing such games that presents increasing branching factor, MCTS may consider, as alpha-beta do, pruning to keep efficiency. We present a modified version of MCTS-Solver algorithm, called OPI-MCTS as Optimal Player Information MCTS, that adds game state information to exploit logical reasoning during backpropagation and that influences selection and expansion. OPI-MCTS is experimented in Chinese Dark Chess, which is a imperfect information game. OPI-MCTS is compared with classical MCTS.

[1]  T.-S. Hsu,et al.  Chinese Dark Chess , 2010, J. Int. Comput. Games Assoc..

[2]  I-Chen Wu,et al.  The Art of the Chinese Dark Chess Program DIABLE , 2013 .

[3]  Mark H. M. Winands,et al.  Quality-based Rewards for Monte-Carlo Tree Search Simulations , 2014, ECAI.

[4]  Mark H. M. Winands,et al.  Monte-Carlo Tree Search Solver , 2008, Computers and Games.

[5]  Nicolas Jouandeau,et al.  Small and Large MCTS Playouts Applied to Chinese Dark Chess Stochastic Game , 2014, CGW@ECAI.

[6]  Nicolas Jouandeau,et al.  Material Symmetry to Partition Endgame Tables , 2013, Computers and Games.

[7]  I-Chen Wu,et al.  Strength Improvement and Analysis for an MCTS-Based Chinese Dark Chess Program , 2015, ACG.

[8]  Richard Lorentz.  Early Playout Termination in MCTS , 2015, ACG.

[9]  Mark H. M. Winands,et al.  Monte-Carlo Tree Search and minimax hybrids , 2013, 2013 IEEE Conference on Computational Inteligence in Games (CIG).

[10]  Graham.  Editorial: IEEE Transactions on Computational Intelligence and AI in Games , 2015 .

[11]  Peter Auer,et al.  Finite-time Analysis of the Multiarmed Bandit Problem , 2002, Machine Learning.

[12]  Tsan-sheng Hsu,et al.  A Quantitative Study of 2 × 4 Chinese Dark Chess , 2014 .

[13]  Nathan R. Sturtevant,et al.  Monte Carlo Tree Search with heuristic evaluations using implicit minimax backups , 2014, 2014 IEEE Conference on Computational Intelligence and Games.

[14]  Christopher Archibald,et al.  Monte Carlo *-Minimax Search , 2013, IJCAI.

[15]  I-Chen Wu,et al.  Design and Implementation of Chinese Dark Chess Programs , 2015, IEEE Transactions on Computational Intelligence and AI in Games.

[16]  Csaba Szepesvári,et al.  Bandit Based Monte-Carlo Planning , 2006, ECML.

[17]  Tsan-sheng Hsu,et al.  Equivalence Classes in Chinese Dark Chess Endgames , 2015, IEEE Transactions on Computational Intelligence and AI in Games.

[18]  Shin-Ming Cheng,et al.  Technologies and Applications of Artificial Intelligence , 2014, Lecture Notes in Computer Science.

[19]  Simon M. Lucas,et al.  A Survey of Monte Carlo Tree Search Methods , 2012, IEEE Transactions on Computational Intelligence and AI in Games.

[20]  Nicolas Jouandeau,et al.  Monte-Carlo Tree Reductions for Stochastic Games , 2014, TAAI.