To Create Adaptive Game Opponent by Using UCT

Abstract

Adaptive Game AI improves adaptability of opponent AI as well as the challenge level of the gameplay; as a result the entertainment of game is augmented. Opponent game AI is usually implemented by scripted rules in video games, but the most updated algorithm of UCT (Upper Confidence bound for Trees) which perform excellent in computer go can also be used to achieve excellent result to control non-player characters (NPCs) in video games. In this paper, the prey and predator game genre of Dead End is used as a test-bed, the basic principle of UCT is presented, and the effectiveness of its application to game AI development is demonstrated. The experiment compares the performance of different NPCs control approaches: given a 300 milliseconds for each simulation step, the approach of UCT with recognized player's strategy pattern is better than the one of UCT without recognized player's strategy pattern, the worst one is Monte-Carlo approach.

A Survey of Monte Carlo Tree Search Methods

“…He et al [100] use UCT for the behaviour of the dogs in their artificial player. Their results show how the performance is better when the simulation time is higher and that UCT outperforms the flat Monte Carlo approach.…”

To Create Intelligent Adaptive Game Opponent by Using Monte-Carlo for the Game of Pac-Man

“…The application of Monte-Carlo for Pac-Man is based on the hypothesis that to control each Ghost in game board situation is actually determined by a choice of legal move [5]. The possible legal moves could be in the direction of North, South, East, and West.…”

To create neuro-controlled game opponent from UCT-created data

“…To use the Reinforcement Learning approaches of UCT or Monte Carlo [9,10] to generate game AI or to control non-player characters (NPCs) in video games not only requires no domain knowledge from the domain experts, but also saves burden of game AI programming.…”