Game AI techniques from algorithmic approach to machine learning

  • Full Conference Pass (FC) Full Conference Pass (FC)
  • Full Conference One-Day Pass (1D) Full Conference One-Day Pass (1D)

Date: Tuesday, December 4th
Time: 11:00am - 12:45pm
Venue: Hall D5 (5F, D Block)

Summary: The goal of this course is explaining modern game AI techniques which evolved in these 20 years. While, before 2000, game AI was the field that gathered very tricky programming techniques, after 2001, game AI has become to a outstanding theory by using academic AI techniques such as robotics AI. Furthermore, recently, machine learning techniques have been introduced to game development. The course is divided into five parts. The first part is showing the overview of game AI. The modern AI system consists of distributed three AI such as character AI, Navigation AI, and Meta-AI. Character AI means a decision making system for a character in game, Navigation AI is path navigation system and environment recognition system by analyzing static and dynamic terrain in game. Meta-AI controls a flow-of-game by generating enemy agents at best points and increases/decreases number of monsters to adjust difficulty level for each user. The second part is explaining character AI technologies such as agent architecture, seven popular decision making algorithm, and memory structure. The third part is teaching some techniques of Navigation AI such as Navigation mesh, way points, A* algorithm, Dijkstra algorithm and so on. The forth part is introducing meta-AI. Meta-AI is new and original technology in game AI. It controls game contents dynamically by generating terrains, populating enemies and changing stories. All these chapters are explained by using FINAL FANTASY XV examples. The fifth part is a new topic that shows a theory and examples of game titles of machine learning techniques and evolutionary algorithms. Recently some titles use them in game and game development process. Especially AI for quality assurance developed in these 5 years. One of the goals of QA-AI is auto-balancing of many parameters in a game, and the other is replacing human by AI for game play testing.

Moderator: Youichiro Miyake, Square Enix Co., Ltd., Japan
Lecturer(s): Kazuko Manabe, Square Enix Co., Ltd.; Advanced Technology Division, Japan

Author(s)/Speaker(s) Bio:
Youichiro Miyake has been involved in the development of video game titles while researching game AI technologies as the lead AI researcher at Square Enix. He has developed and designed AI for numerous game titles. He has given many lectures in the universities and game developer conferences and has published academic papers on game AI and procedural techniques for digital games. He is the chair of the SIG-AI in International Game Developers Association Japan Chapter, and is also the board of Digital Game Research Association Japan and the Society of Art and Science.

After receiving Master's degree with research topic as Supervised Learning for game of "Go", Kazuko Manabe started her career in Sony Interactive Entertainment as a programmer on game development in PlayStation4, where she also created tools to support process of asset creation. Presently she is working as an AI engineer/researcher at Square Enix, where she creates system for automatic QA and game balancing using AI.



/en/attendees/doctoral-consortium /en/attendees/production-gallery