A Multi-agent System for Autonomous Control of Game Parameters

Control of game parameters to reach domain-related objectives cannot be easily handled with classic control theory approaches. Given the dynamics and complexity of modern game engines, diversity of human players and their constantly changing nature, this paper advocates for means to tune game parameters in real time, with no use of game or users models. The proposed approach, based on a multi-agent system, is used to control two dynamic systems before analyzing the results.