This paper proposes a method for solving the problem of avoiding pincer moves of the ghosts in the game of Ms. Pac-Man to enhance the chance of survival for Ms. Pac-Man. To achieve this, the approach of using the UCT (UCB applied to trees) algorithm, a Monte-Carlo tree search algorithm that proved to be successful with a computer game of Go, was employed. The UCT algorithm is one of the tree search algorithms based on simulations and performs many searches on potential moves weighing the mean reward based on random simulation and the number of searches for a node. Although it is difficult to build a complete game tree that considers all situations in a real-time game like Ms. Pac-Man, because of the time limitation and the extent of implicit information, a method was developed to predict the probability of being trapped from pincer moves and avoid such situations at a feasible cost by repeating simulations on similar discretized situations. As a result of a performance comparison between the proposed system and existing programs, significant improvement in the performance of proposed system over existing programs was observed in terms of its ability to survive, implying the effectiveness of proposed method.
[1]
Jonathan Schaeffer,et al.
Monte Carlo Planning in RTS Games
,
2005,
CIG.
[2]
Simon M. Lucas,et al.
A simple tree search method for playing Ms. Pac-Man
,
2009,
2009 IEEE Symposium on Computational Intelligence and Games.
[3]
Rémi Coulom,et al.
Efficient Selectivity and Backup Operators in Monte-Carlo Tree Search
,
2006,
Computers and Games.
[4]
Alan Fern,et al.
UCT for Tactical Assault Planning in Real-Time Strategy Games
,
2009,
IJCAI.
[5]
Csaba Szepesvári,et al.
Bandit Based Monte-Carlo Planning
,
2006,
ECML.