JSAI2019

Presentation information

Interactive Session

[4Rin1] Interactive Session 2

Fri. Jun 7, 2019 9:00 AM - 10:40 AM Room R (Center area of 1F Exhibition hall)

9:00 AM - 10:40 AM

[4Rin1-09] AI Trainning Platform and AI Implementation with DQN for Realtime Strategy Game

〇Yikun Zhang1, Tomonori Hashiyama1, Shun'ichi Tano1 (1. The University of Electro-Communications)

Keywords:Reinforcement Learning, Game AI

Research on video game AI has been done for a long time. In recent years, due to the rapid development of neural networks, the game AI which was considered difficult to build has appeared. After these AI beat humans in turn based games with complete information, such as GO, the next research direction has been focused on simultaneous games with incomplete information. The RTS game is one of the simultaneous-move games with incomplete information; therefore, in this paper we report the design of a platform for AI learning in the RTS game whose mechanism has been simplified. In addition, we implemented an AI for this type of game, and carried out some experiments.