A Real-Time Multiagent Strategy Learning Environment and Experimental Framework
Many problems in the real world can be attributed to the problem of multiagent. The study on the issue of multiagent is of great significance to solve these social problems. This paper reviews the research on multiagent based real-time strategy game environments, and introduces the multiagent learning environment and related resources. We choose a deep learning environment based on the StarCraft game as a research environment for multiagent collaboration and decision-making, and form a research mentality focusing mainly on reinforcement learning. On this basis, we design a verification platform for the related theoretical research results and finally form a set of multiagent research system from the theoretical method to the actual platform verification. Our research system has reference value for multiagent related research.
KeywordsMultiagent Reinforcement learning Real-time strategy
The authors acknowledge the support of the National Natural Science Foundation of China (grant U1608253, grant 61473282), Natural Science Foundation of Guangdong Province (2017B010116002) and this work was supported by the Youth Innovation Promotion Association, CAS. Any opinions, findings, conclusions, or recommendations expressed in this paper are those of the authors, and do not necessarily reflect the views of the funding organizations.
- 2.Marc, G.B., Yavar, N., Joel, V., Michael, B.: The arcade learning environment: an evaluation platform for general agents. In: 24th International Joint Conference on Artificial Intelligence, pp. 4148–4152 (2015)Google Scholar
- 3.Stefan, W., Ian, W.: Applying reinforcement learning to small scale combat in the real-time strategy game StarCraft: Broodwar. In: 2012 IEEE Conference on Computational Intelligence and Games (CIG 2012), pp. 402–408 (2012)Google Scholar
- 5.Sainbayar, S., Arthur, S., Gabriel, S., Soumith, C., Rob, F.: Mazebase: a sandbox for learning from games. https://arxiv.org/abs/1511.07401
- 6.Nicolas, U., Gabriel, S., Zeming, L., Soumith, C.: Episodic exploration for deep deterministic policies: an application to StarCraft micromanagement tasks. https://arxiv.org/abs/1609.02993
- 7.Peng, P., Ying, W., Yaodong, Y.: Multiagent bidirectionally-coordinated nets emergence of human-level coordination in learning to play StarCraft combat games. https://arxiv.org/abs/1703.10069
- 8.Jakob, N.F., Gregory, F., Triantafyllos, A.: Counterfactual multi-agent policy gradients. In: The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI 2018), New Orleans (2018)Google Scholar
- 9.Oriol, V., Timo, E., Kevin, C.: StarCraft II: a new challenge for reinforcement learning. https://arxiv.org/abs/1708.04782
- 10.Marc, L., Vinicius, Z., Audrunas, G.: A unified game-theoretic approach to multiagent reinforcement learning. https://arxiv.org/abs/1711.00832