MiaoSuan Wargame: A Multi-Mode Integrated Platform for Imperfect Information Game

Abstract

Just like data plays a fundamental role in perceptual intelligence, simulated environments become the cornerstone of artificial agents in cognitive intelligence. Meanwhile, the evaluation of decision agents has put forward an urgent need for human-machine accessible game platforms, on which humans and agents can interact and confront, even the “Turing test” can be conducted. In this work, MiaoSuan wargame platform, a human-computer gaming platform for multi-agent imperfect information game is proposed. It provides multiple scenes and tasks, together with various confrontation forms (computer vs. computer, human vs. computer, and human/computer hybrid) and game modes (online confrontation, offline training, and Turing test) for agents’ training and evaluation. The MiaoSuan platform is constructed as a modular architecture with separable engine, AI interface, human interface and other functional modules, in which different game environments and tasks can be conveniently integrated. Two kinds of workflows, online confrontation and offline training, are designed to schedule different modules to achieve their corresponding game modes. Besides, based on this platform, a “Turing test” method for agent evaluation can be implemented automatically. It introduces a comprehensive evaluation of agents’ performance through subjective human assessment or objective data analysis, rather than only depending on their game scores. The MiaoSuan wargame platform is widely used in individual practice and organized events. It is available at http://wargame.ia.ac.cn/.