Impact of neighboring agent’s characteristics with Q-learning in network multi-agent system

Abstract

In a Network Multiagent System (NMAS), agent learning is one of the most investigated attribute and vital issues in terms of enhancing the global productivity of the system. The work in this paper implements a new approach that uses neighboring agent characteristics to improve the learning performance of an agent in NMAS considering social networks. The main contributions of this work are: First, it presents the basic principles of multi-agent learning while considering various learning issues. Second, it reviews the main research developments in the field of multi-agent learning. Third, it introduces a network multi-agent learning framework by considering the neighboring agent characteristics, i.e., the relative degree of neighboring nodes and past interaction histories of neighboring agents in a social network. Furthermore, the proposed framework is experimentally assessed by implementing various social networks (scale-free and small-world) in the form of a NMAS.