A study on Reinforcement Learning system for agents to acquire cooperative behavior in gap-widening situations

This article proposes an Interactive Hierarchical Reinforcement Learning system (IH-RL). The goal of our study is that the agents using the IH-RL acquire adequate behaviors to cooperative in “gap-widening” situations. Such situations are observed in a variety of real-world environments (e.g., economic gaps between humans or between companies in a community), and are thus important to solve. Computer simulations are carried out to evaluate the basic performance of our system. The results showed that the IH-RL resolves gap-widening situations through agents' cooperative behaviors.