Abstract
Curling is one of the most strategic winter sports. Recently, many computer scientists have studied curling strategies. The Digital Curling system is a framework used to compare curling strategies. Herein, we present a computer agent based on the Monte-Carlo Tree Search (MCTS) for the Digital Curling framework. We implemented a novel action decision method based on MCTS for Markov decision processes with continuous state space. The experimental results show that our search method is effective for agents with a simple simulation policy and agents with a handmade complex one.
Users
Please
log in to take part in the discussion (add own reviews or comments).