Recent advances in Game Theory have applications in Economics, Smart Power Grids, Intelligent Transportation, Smart Cities, Self-Driving, Multi-agent Robotics, etc. As all of these systems would require a multi-agent rational decision making to derive optimal outcomes. These procedure of achieving maximum payoff, would often lead towards competing or cooperative strategy outcomes for an independent rational agent. One similar rational agent is the learning agent using reinforcement learning to explore and exploit the environment, in order to learn optimal policy. However, similar approach does not work as effectively for multi-agent setting, where since all the agents have shared state-space, one agent’s learned policy might over-fit to other agent’s policies or may fail to generalize during execution. One reason is, reinforcement learning agent is an independent agent, which interacts only with Markovian and static environment. But, These assumptions does not hold for Multi-agent setting, where agent has to react dynamically based on other agent’s behaviours. To accommodate such issues and effectively learn best responses for each agent, authors in [2] introduces new metric Joint-policy correlation and describes a meta-algorithm for general Multi-agent Reinforcement Learning. Based on economic reasoning, this algorithm uses deep reinforcement learning to compute best responses to a distribution over polices and uses empirical game-theoretic analysis to compute new meta-strategy from learned distributions. Approach presented, rather than computing exact best response strategy, computes approximate best response using reinforcement learning. Objective of project is to gain better understanding of the new Joint-policy metric, model the hierarchical approach defined and replicate the experimental results presented of Meta-strategy learning algorithm.

• D. Bloembergen, K. Tuyls, D. Hennes, and M. Kaisers. Evolutionary dynamics of multi-agent learning: A survey. Journal of Artificial Intelligence Research, 53:659–697, 08 2015.
• M. Lanctot, V. Zambaldi, A. Gruslys, A. Lazaridou, K. Tuyls, J. Perolat, D. Silver, and T. Graepel. A unified game-theoretic approach to multiagent reinforcement learning. In I. Guyon, U. V. Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vishwanathan, and R. Garnett, editors, Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc., 2017.
• R. Sharma and M. Gopal. Synergizing reinforcement learning and game theory—a new direction for control. Applied Soft Computing, 10(3):675–688, 2010.
• K. G. Vamvoudakis, H. Modares, B. Kiumarsi, and F. L. Lewis. Game theory-based control system algorithms with real-time reinforcement learning: How to solve multiplayer games online. IEEE Control Systems Magazine, 37(1):33–52, 2017.
• W. Walsh, R. Das, G. Tesauro, and J. Kephart. Analyzing complex strategic interactions in multi-agent systems. 01 2002.