Different forms of the games in multiagent reinforcement learning: alternating vs. simultanous movements

Abstract

Multiagent systems are one of the most promising solutions in most of real life applications in which some kinds of social interactions or conventions are involved. Agent oriented applications are broadly explored among which learning in unknown environment is well developed based on Markov Decision Process (MDP). On the other hand, learning in multiagent systems has been recently introduced, basically in conjunction with game theory which is the science of investigating multiple interactive agents. During learning, self-interested agents are attempting to find the equilibrium policy based on the structure of the game, mostly considered as normal form games. In this paper, we focus on bringing into discussion game structures, addressed as normal form games and extensive form games, in learning process. This includes also some modifications and refinements in initially introduced concepts as well as a proposed approach in extensive form games.