Extension of a Conscious Decision-Making System Using Deep Reinforcement Learning to Multi-agent Environments

The author is currently investigating conscious decision-making systems in environments where multiple types of rewards and penalties exist. In this research, we consider applying a method based on deep reinforcement learning to multi-agent environments. We have observed that varying the level of information exchange in a multi-agent environment can improve performance in a single-agent environment. We believe that this result opens up the possibility of extending conscious decision-making systems to multi-agent environments.