A STUDY OF EMERGENT COOPERATIVE BEHAVIORS OF MULTI-AGENT SYSTEMS USING REINFORCEMENT LEARNING
Issues in Information Systems · 2023
Multi-Agent systems consist of multiple interacting agents that solve complex problems too difficult for monolithic systems to solve.Each agent has a different role and is not aware of the state of other agents.Multi-Agent Reinforcement Learning is a common approach to training agents in these systems.Many researchers tend to focus on the implementation of reinforcement learning for training agents towards a particular goal.This paper, in contrast, implements reinforcement learning for a different purpose.This paper is about the analysis of emergent cooperative strategies between agents while achieving their goals.The research will be done using simulations of a cooperative game and documenting the strategies the agents create over a set number of episodes.