Genetic Optimizing Method for Real-time Monte Carlo Tree Search Problem.

Monte Carlo Tree Search is one of the best algorithms for solving board game problems. However, Monte Carlo Tree Search is not suitable for real-time game problem because the problems have uncertainty of opponent’s action and a lot of simulation when determining behavior. We propose a Genetic Optimizing Method to solving the problems encountered when applying Monte Carlo Tree Search to real-time games. Our method helps solve the dilemma of Real-time Monte Carlo Tree Search between simulation and the number of branching factors by utilizing genetic algorithms. Finally, we applied our method to the Real-time Fighting Game to verify its performance.