Distributed Multi-Agent Reinforcement Learning Based on Feudal Networks

Abstract

This paper deals with distributed multi-agent reinforcement learning based on feudal networks. The reinforcement learning, and the features of distributed reinforcement learning are presented. Architectures of distributed reinforcement learning are reviewed. Reinforcement learning, based on feudal networks, is presented. The architecture and hierarchy of the feudal neural network are shown, and the key elements of the network: worker and manager are explained. The components of the manager and worker, their interaction with each other and their interaction with the testing environment are described. The environment for testing the algorithm is selected. The results of testing are demonstrated and the success of the algorithm, the achievement of learning successes and the interaction of agents with the environment are determined. The work done proves that feudal networks are suitable for implementing distributed learning and are able to learn and achieve the desired results in complex multi-agent environments.