nonverbal cooperation artificial game player applied tree search algorithm

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K17872
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 62040:Entertainment and game informatics-related
• Research Institution: Sasebo National College of Technology
• Principal Investigator: Sato Naoyuki 佐世保工業高等専門学校, 電子制御工学科, 准教授 (30826889)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: ゲーム情報学 / エンターテイメント工学 / 人工知能 / 探索

Outline of Final Research Achievements

By applying the primal Monte Carlo method to simulate random choice, including “the situation of the field as seen by other players,” we observed the generation of moves that propose collusion to another player in a game where their assistance is advantageous.
We also succeeded in enhancing computational efficiency of the proposed method. Using statistical data obtained by artificial players with naiive rule-based algorithm, we achieved fast search by substituting statistical observations for exact random simulations. We have confirmed that this improvement allows us to generate moves in practical computation time even in situations where the previous method was not able to perform sufficient search.

Free Research Field

ゲーム情報学

Academic Significance and Societal Importance of the Research Achievements

本手法で実現した「他のプレイヤから見える場の状況を考慮に含めたモンテカルロシミュレーション」は，四人対戦ゲームでは計算が極めて重くなるものの，他プレイヤとの暗黙の結託を施行した着手を生成できる潜在能力のある手法である．更にはこの探索を軽量化するために，単純なルールベースド型プレイヤを何度も対戦させて得た統計値でシミュレーションの一部を代替するアプローチは，シミュレーションの正確さをわずかに損なうものの，計算を十分に実用的な時間で実行させられる改良点である．
これらの知見は，『多人数ゲームにおける暗黙の結託を人工プレイヤ』の実現や更なる学術的探求に貢献すると考えられる．