![]() ![]() Human-level control through deep reinforcement learning.Īndrew Y Ng, Daishi Harada, and Stuart Russell. Marc G Bellemare, Alex Graves, Martin Riedmiller, Andreas K Fidjeland, Georg ![]() Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A Rusu, Joel Veness, Revisiting the master-slave architecture in multi-agent deep Xiangyu Kong, Bo Xin, Fangchen Liu, and Yizhou Wang.Įffective master-slave communication on a multi-agent deep In International Conference on Cooperative Design, VisualizationĪnd Engineering, pages 47–51. Jaekwang Kim, Kwang Ho Yoon, Taebok Yoon, and Jee-Hyong Lee.Ĭooperative learning by replay files in real-time strategy game. IEEE International Joint Conference on, pages IJCNN 2008.(IEEE World Congress onĬomputational Intelligence). Inferring strategies from limited reconnaissance in real-timeīuilding a player strategy model by analyzing replays of real-time Jesse Hostetler, Ethan W Dereszynski, Thomas G Dietterich, and Alan Fern. Neural computation, 9(8):1735–1780, 1997.ĭan Horgan, John Quan, David Budden, Gabriel Barth-Maron, Matteo Hessel, Hadoĭistributed prioritized experience replay. Tensorflow: a system for large-scale machine learning.ĭeep recurrent q-learning for partially observable mdps. Martín Abadi, Paul Barham, Jianmin Chen, Zhifeng Chen, Andy Davis, Jeffreyĭean, Matthieu Devin, Sanjay Ghemawat, Geoffrey Irving, Michael Isard, et al. ![]() UAlbertaBot for StarCraft AI Competitions. Tournament Manager Software for StarCraft AI Competitions. Official AIIDE 2018 StarCraft Competition Results. Official AIIDE 2017 StarCraft Competition Results. ReferencesīWAPI: Brood war api, an api for interacting with starcraft: Broodwar But how to train this multi-agent model and react in real time at relatively large scale is still an open question. It is difficult for rule-controlled units to behave properly in different situations and cooperate with each other. Micro management in StarCraft is a multi-agent system. Our bot achieves overall 83% win-rate, outperforming 26 bots in total 28 entrants. We evaluate our bot, named as LastOrder 1 1 1LastOrder is from a character in Toaru Majutsu no Index, on the AIIDE 2017 StarCraft AI competition bots set. Our framework combines the reinforcement learning approach Ape-X DQN horgan2018distributed with Long-Short-Term-Memory (LSTM) hochreiter1997long to improve the macro action selection in bot. In this paper, we propose a DRL based framework to do macro action selection. These rules are not scalable and efficient enough to cope with the large but partially observed macro state space in SC. One vital reason is that current bots mainly rely on predefined rules to perform macro actions. However, a big gap still remains between the top bot and the professional human players. Thanks to the annual AIIDE aiidecompetition and CIG cigcompetition competitions, a growing number of bots are proposed and being continuously improved. In recent years, SC is also considered as a testbed for AI research, due to its enormous state space, hidden information, multi-agent collaboration and so on. StarCraft (SC) is one of the most popular and successful Real Time Strategy (RTS) games. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |