戦略型カードゲームのための戦略獲得法 [in Japanese] Strategy-acquisition Scheme for Strategic Card Game [in Japanese]
Access this Article
Search this Article
Author(s)
Abstract
市販テレビゲームにおいては，CPU の人間らしさというリアリティに，プレイヤの意識が高まりつつある．従来研究では，将棋や仮想空間における CPU の人間らしさの検討はなされているものの，市販テレビゲームのような戦略型ゲームにおいて，人間らしい行動や戦略を AI 技術を用いて実現した例はほとんどない．本稿では，CPU の「人間らしさ」の実現を目標とし，戦略型カードゲームの戦略を自動的に獲得する機構を提案した．戦略学習における困難性として，部分観測に起因した巨大な状態空間が挙げられるが，サンプリング，相手の行動予測，ゲームの特徴を考慮した次元圧縮により克服した．また，戦略学習機構によって得た戦略を，ルールベースの戦略と対戦させ，有効性を評価した．The computer (CPU) like a human has lately attracted considerable attention in the video game. The past studies report strategy-acquisition scheme for the board game such as shogi. However, there are few studies about these scheme applied in the video game. We aim at the acquisition of strategy like a human, we present an automatic strategy-acquisition scheme for strategic card game. Because of this card game includes many unobservable variables in a large state space, we suggest a sampling technique, action predictor, and value function. To evaluate our method, we carried out computer simulations where our agent played against a rule-based agent.
The computer(CPU) like a human has lately attracted considerable attention in the video game. The past studies report strategy-acquisition scheme for the board game such as shogi. However, there are few studies about these scheme applied in the video game. We aim at the acquisition of strategy like a human, we present an automatic strategy-acquisition scheme for strategic card game. Because of this card game includes many unobservable variables in a large state space, we suggest a sampling technique, action predictor, and value function. To evaluate our method, we carried out computer simulations where our agent played against a rule-based agent.
Journal
-
- IPSJ SIG technical reports
-
IPSJ SIG technical reports 2008(26(2008-EC-009)), 9-16, 2008-03-07
Information Processing Society of Japan (IPSJ)