Abstract
Intelligent agents are critical components of the current game development state of the art. With advances in hardware, many games can simulate cities and ecosystems full of agents. These environments are known as multi-agent environments. In this domain, reinforcement learning has been explored to develop artificial agents in games. In reinforcement learning, the agent must discover which actions lead to greater rewards by experimenting with these actions and defining a search by trial and error. Specifying when to reward agents is not a simple task and requires knowledge about the environment and the problem to be solved. Furthermore, defining the elements of multi-agent reinforcement learning required for the learning environment can be challenging for developers who are not domain experts. This paper proposes a framework for developing multi-agent cooperative game environments to facilitate the process and improve agent performance during reinforcement learning. The framework consists of steps for modeling the learning environment and designing rewards and knowledge distribution, trying to achieve the best environment configuration for training. The framework was applied to the development of three multi-agent environments, and tests were conducted to analyze the techniques used in reward design. The results show that the use of frequent rewards favors the emergence of essential behaviors (necessary for the resolution of tasks), improving the learning of agents. Although the knowledge distribution can reduce task complexity, dependency between groups is a decisive factor in its implementation.
This work is supported by CAPES and FAPERJ.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
References
Bellemare, M.G., Naddaf, Y., Veness, J., Bowling, M.: The arcade learning environment: an evaluation platform for general agents. J. Artif. Intell. Res. 47(1), 253–279 (2013)
Berner, C., et al.: Dota 2 with large scale deep reinforcement learning (2019)
Cohen, A., et al.: On the use and misuse of absorbing states in multi-agent reinforcement learning (2021)
Foerster, J., Assael, Y., Freitas, N., Whiteson, S.: Learning to communicate with deep multi-agent reinforcement learning. In: Proceedings of the 30th International Conference on Neural Information Processing Systems, pp. 2145–2153. NIPS’16, Curran Associates Inc. (2016)
Foerster, J., Farquhar, G., Afouras, T., Nardelli, N., Whiteson, S.: Counterfactual multi-agent policy gradients. In: Proceedings of the AAAI Conference on Artificial Intelligence, vol. 32, pp. 2974–2982. AAAI (2018)
Hessel, M., et al.: Rainbow: Combining improvements in deep reinforcement learning. In: Proceedings of the 32nd AAAI Conference on Artificial Intelligence, vol. 32, pp. 3215–3222. PKP Publishing Services Network (2018)
Iqbal, S., Sha, F.: Actor-attention-critic for multi-agent reinforcement learning. In: Proceedings of the 36th International Conference on Machine Learning, pp. 2961–2970. PMLR 97, Long Beach, California (2019)
Johnson, M., Hofmann, K., Hutton, T., Bignell, D.: The malmo platform for artificial intelligence experimentation. In: Proceedings of the 25th International Joint Conference on Artificial Intelligence, pp. 4246–4247. IJCAI’16, AAAI Press (2016)
Jorge, E., Kågebäck, M., Johansson, F., Gustavsson, E.: Learning to play guess who? and inventing a grounded language as a consequence (2016)
Kaelbling, L., Littman, M., Moore, A.: Reinforcement learning: a survey. J. Artif. Intell. Res. 4, 237–285 (1996)
Kempka, M., Wydmuch, M., Runc, G., Toczek, J., Jaśkowski, W.: Vizdoom: A doom-based ai research platform for visual reinforcement learning. In: IEEE Conference on Computational Intelligence and Games (CIG), pp. 1–8 (2016)
Mnih, V., et al.: Playing atari with deep reinforcement learning (2013)
Mnih, V., et al.: Human-level control through deep reinforcement learning. Nature 518, 529–533 (2015)
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., Klimov, O.: Proximal policy optimization algorithms (2017)
Summerville, A., et al.: Procedural content generation via machine learning (pcgml). IEEE Trans. Games 10(3), 257–270 (2018)
Sunehag, P., et al.: Value-decomposition networks for cooperative multi-agent learning based on team reward. In: Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems. pp. 2085–2087. AAMAS ’18, International Foundation for Autonomous Agents and Multiagent Systems, Richland, SC (2016)
Sutton, R., Barto, A.: Reinf. Learn.: Introduction. MIT Press, London, England (2018)
Vidhate, D., Kulkarni, P.: Enhanced cooperative multi-agent learning algorithms (ecmla) using reinforcement learning. In: 2016 International Conference on Computing, Analytics and Security Trends (CAST), pp. 556–561. IEEE (2016)
Vinyals, O., et al.: Grandmaster level in Starcraft II using multi-agent reinforcement learning. Nature 575, 350–354 (2019)
Yannakakis, G., Togelius, J.: Artificial Intelligence and Games. Springer (2018). https://doi.org/10.1007/978-3-319-63519-4
Zhang, Q., Zhao, D., Lewis, F.: Model-free reinforcement learning for fully cooperative multi-agent graphical games. In: 2018 International Joint Conference on Neural Networks (IJCNN), pp. 1–6. IEEE (2018)
Zhao, Y., Borovikov, I., Rupert, J., Somers, C., Bierami, A.: On multi-agent learning in team sports games. In: Proceedings of the 36th International Conference on Machine Learning (ICML) (2019)
Acknowledgments
The authors would like to thank NVIDIA, CAPES and FAPERJ for the financial support.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2022 IFIP International Federation for Information Processing
About this paper
Cite this paper
Ferreira, T., Clua, E., Kohwalter, T.C., Santos, R. (2022). OptimizingMARL: Developing Cooperative Game Environments Based on Multi-agent Reinforcement Learning. In: Göbl, B., van der Spek, E., Baalsrud Hauge, J., McCall, R. (eds) Entertainment Computing – ICEC 2022. ICEC 2022. Lecture Notes in Computer Science, vol 13477. Springer, Cham. https://doi.org/10.1007/978-3-031-20212-4_7
Download citation
DOI: https://doi.org/10.1007/978-3-031-20212-4_7
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-031-20211-7
Online ISBN: 978-3-031-20212-4
eBook Packages: Computer ScienceComputer Science (R0)