TY - GEN
T1 - Enhancements for Real-Time Monte-Carlo Tree Search in General Video Game Playing
AU - Soemers, Dennis J. N. J.
AU - Sironi, Chiara F.
AU - Schuster, Torsten
AU - Winands, Mark H. M.
PY - 2016/9
Y1 - 2016/9
N2 - General Video Game Playing (GVGP) is a field of Artificial Intelligence where agents play a variety of real-time video games that are unknown in advance. This limits the use of domain-specific heuristics. Monte-Carlo Tree Search (MCTS) is a search technique for game playing that does not rely on domain-specific knowledge. This paper discusses eight enhancements for MCTS in GVGP; Progressive History, N-Gram Selection Technique, Tree Reuse, Breadth-First Tree Initialization, Loss Avoidance, Novelty-Based Pruning, Knowledge-Based Evaluations, and Deterministic Game Detection. Some of these are known from existing literature, and are either extended or introduced in the context of GVGP, and some are novel enhancements for MCTS. Most enhancements are shown to provide statistically significant increases in win percentages when applied individually. When combined, they increase the average win percentage over sixty different games from 31.0% to 48.4% in comparison to a vanilla MCTS implementation, approaching a level that is competitive with the best agents of the GVG-AI competition in 2015.
AB - General Video Game Playing (GVGP) is a field of Artificial Intelligence where agents play a variety of real-time video games that are unknown in advance. This limits the use of domain-specific heuristics. Monte-Carlo Tree Search (MCTS) is a search technique for game playing that does not rely on domain-specific knowledge. This paper discusses eight enhancements for MCTS in GVGP; Progressive History, N-Gram Selection Technique, Tree Reuse, Breadth-First Tree Initialization, Loss Avoidance, Novelty-Based Pruning, Knowledge-Based Evaluations, and Deterministic Game Detection. Some of these are known from existing literature, and are either extended or introduced in the context of GVGP, and some are novel enhancements for MCTS. Most enhancements are shown to provide statistically significant increases in win percentages when applied individually. When combined, they increase the average win percentage over sixty different games from 31.0% to 48.4% in comparison to a vanilla MCTS implementation, approaching a level that is competitive with the best agents of the GVG-AI competition in 2015.
U2 - 10.1109/CIG.2016.7860448
DO - 10.1109/CIG.2016.7860448
M3 - Conference article in proceeding
T3 - IEEE Conference on Computational Intelligence and Games
SP - 436
EP - 443
BT - 2016 IEEE Conference on Computational Intelligence and Games (CIG)
PB - IEEE
T2 - 2016 IEEE Conference on Computational Intelligence and Games (CIG)
Y2 - 20 September 2016 through 23 September 2016
ER -