MASPI - A Multi-Agent System Playing Imperial 2030

Abstract

Board games are a popular subject matter for development of artificial intelligence because of their discrete environment, because they are a proxy for intelligence, and because they can show quantifiable mastery from competition. Research in this area has commonly focused on mastering classic abstract games or working on communication and deal-making. Modern board games feature novel learning environments that challenge common forms of learning in board game artificial intelligence research today. In MARS - A Multi-Agent System Playing Risk the Multi-Agent system structure is capable of handling the nuances of the game Risk. In this paper, the same structure will be adapted to the game Imperial 2030. To ensure that the Multi-Agent System Playing Imperial, MASPI, is competitive against common bots, MASPI plays games against different configurations of bots. The results of these games show that MASPI’s strategy allows it to outperform the other examined bots in multiple types of games.