Multi-Agent Cooperation in Games with Goal Oriented Action Planner: Use Case in WONDER Prototype Project

Abstract

Multi-agent cooperation systems often rely on an external commander-like entity that will plan for all the agents it manages. When it comes to games development, this entity can be hidden from the player, which makes the player believe the characters are actually taking their own decisions. However, it has limitations. We believe that to achieve truly believable character-interactions between non-playable characters, real(non-scripted) communication between the agents is the key. We will introduce our multi-agent cooperation system where each AI-agent thinks for itself and communicates with the other agents to cooperate and achieves complex goals. The system will be demonstrated with a use case in our WONDER prototype project.