Two Multi-Unit Artificial Intelligence Decision Modes in Video Games：A Comparison

Using AI to manipulate game characters is already common and has been applied to various types of games. However, many current studies focus on AI operating a single role. In future online games, there will be a great demand to operate multiple units simultaneously for collaborative movement, and similar needs have already appeared in current MMORPG. In popular battle games nowadays, there is often a game mode where multiple players belonging to multiple factions engage in battles with each other. If AI needs to be deployed, multiple units need to be operated simultaneously to engage in combat. In response to this issue, two network architectures were proposed and trained and tested in a multiplayer game simulation environment. By comparing the training effects, one of the network structures known as “octopus- shape” is considered to be more effective and valuable for development. This result will contribute to the development of multi unit collaborative artificial intelligence in the future, helping to establish true AI games and real-world automated management systems.