一种基于强化学习的三国杀多智能体博弈方法

A 2v2 Three-Country Killing Multi-Agent Game Method Based on Reinforcement Learning

深度强化学习在处理序列决策与策略探索问题上取得了很大的成功,大多从游戏中展开研究获得启发,其应用领域从单智能体场景扩展到多智能体场景中。基于纸牌的多人对战策略游戏是一种多智能体系统,但现有研究较少,且大多都来自于斗地主、德州扑克。为拓展基于纸牌的多智能体策略游戏的研究,提出了一种基于强化学习的三国杀多智能体博弈方法(SGS-MAPG),自建了以三国杀游戏为背景的2v2对战游戏场景作为实验环境,基于策略梯度的思想对合作的多个智能体建模,在其决策过程中包含了多智能体系统的团队合作与对抗,解决了多个智能体环境下的不稳定性问题。经计算机模拟对战过程,上述方法使智能体经过训练具有良好的学习决策能力,并且能够尝试获得多于基础算法的最终团队奖励,并得到高出至少12%胜率。

Deep reinforcement learning has ac hieved great success in dealing with sequent ial decision-making and strategy exploration,and most of them ar e inspired by in-game research,and its appli cation field has expanded from single-agent scenarios to multi-agent s cenarios.Solitaire-based multiplayer strat egy games are a multi-agent system,but there are few existing studies,an d most of them come from Doudi Landlord and Te xas Hold'em.In order to expand the research of multi-agent strategy games based on cards,this paper proposes a 2v 2 three-country killing multi-agent game method(SGS-MAPG)based on r einforcement learning,which builds a 2v2 bat tle game scene with the background of three-kingdom killing game as the experimental environment,models coop erative multiple agents based on the idea of strategy gradient,and in cludes teamwork and confrontation of multi-a gent systems in its decision-making process,which solves the problem of instability in multiple agent environments.Through computer simulation of the battle process,this method enables th e agent to be trained to have good learning an d decision-making ability,and can try to obtain more final team rewards than the basic algorithm,and get at least 12%higher win rate.