Non-Cooperative Game of Multi-Agent Countermeasure Systems Based on Cognitive Hierarchy Theory

Abstract

In order to research the non-cooperative and opposed problem in the real word, we establish the n-players n-strategies game model for multi-agent countermeasure network system with multi-tasking nature. Cognitive hierarchy (CH) theory is introduced to predict the game evolution, as game player is not the same reasoning in reality. A distributing virtual policy learning algorithm based on heterogeneous cognitive is proposed, which has the advantage of self-learning, self-organizing and self-optimizing. Simulation results indicate the proposed algorithm could make dynamic attack-defend mission planning effectively and give good predictions. Compared the experiment dates under different number of nodes and different parameter τ value, we find that the best τ value is between 1 and 2 and assuming appropriate τ value will optimize the whole performance of the countermeasure system.