1. “The History of Game AI in Japan”
Youichiro Miyake
(Board member of DiGRA JAPAN)
y.m.4160@gmail.com
Replaying JAPAN 2021
https://www.facebook.com/youichiro.miyake
http://www.slideshare.net/youichiromiyake
twitter: @miyayou
2. Youichiro Miyake
Career highlights
- AI Technical Advisor, “FINAL FANTASY XIV”
- Lead AI Architect, “FINAL FANTASY XV”
- AI Technical Director, “KINGDOM HEARTS III”
- QA Automation AI Technical Advisor, “FINAL FANTASY VII REMAKE”
- Visiting Researcher, Research Center for Advanced Science and Technology, The University of Tokyo
- Visiting Professor, Institute of Mathematics for Industry, Kyushu University
- Specially Appointed Professor, Graduate School of Artificial Intelligence and Science, Rikkyo University
- Director, Digital Games Research Association (DiGRA) JAPAN
- Director, The Society for Art and Science
- Director, The Japanese Society for Artificial Intelligence
Awards
- 2008 Funai Award, Digital Contents Symposium 8th
- 2012 Award for Programming/Development Environment Excellence, CEDEC
- 2012 Youth Encouragement Award, DiGRA JAPAN
- 2018 Book Award, CEDEC
5. Contents
Chapter 1. Oral History project in DiGRA JAPAN
Chapter 2. The Two types of Game Preservation
Chapter 3.
Overview of Oral History of Game Development technologies
Chapter 4 Youichiro Miyake’s Works
Chapter 5. Summary
6. Contents
Chapter 1. Oral History project in DiGRA JAPAN
Chapter 2. Two flow of Game Preservation
Chapter 3.
Overview of Oral History of Game Development technologies
Chapter 4 Youichiro Miyake’s Works
Chapter 5. Summary
8. Digital Game Research Association(DiGRA JAPAN)
President:Prof. Akinori Nakamura (Ritsumeikan Univ)
• Oral History Task Force
• Leader Prof. Toru Iwatani (Tokyo Polytechnique Univ, former-DiGRAJ
president)+10 members
• 2007-2012 Literal Record of veteran game developers’ lecture
• 2009- Interview with Indie-game developers(~20 developers)
• 2016- Interview with game developers about 80’s -90’s game
industry
9. Digital Game Research Association(DiGRA JAPAN)
President:Prof. Akinori Nakamura (Ritsumeikan Univ)
• Oral History Task Force
• Leader Prof. Toru Iwatani (Tokyo Polytechnique Univ 、former-
DiGRAJ president)+10 members
• 2007-2012 Literal Record of veteran game developers’ lecture
• 2009- Interview with Indie-game developers(~20 developers)
• 2016- Interview with game developers about 80’s -90’s game
industry
14. び
Mr. Takahashi has many game events in
the department store. Many children attended it.
He was called “Game Master. Takahashi”
(1985)
List of Department store
Where Mr.Takashi had
a game event
(DiGRA JAPAN 2009, Vol.3 No.2)
16. “The beginning of Japanese Digital Game”
(DiGRA JAPAN 2012, Vol.6 No.1)
17. Gee Bee (1978, Namco)
Rocking horse
Ele-mecha
=Combination of digital and
Analog techniques
“The beginning of Japanese Digital Game”
(DiGRA JAPAN 2012, Vol.6 No.1)
19. Digital Game Research Association(DiGRA JAPAN)
会長:Prof. Akinori Nakamura (Ritsumeikan Univ)
• Oral History Task Force
• Leader Prof. Toru Iwatani (Tokyo Polytechnique Univ 、former-
DiGRAJ president)+10 members
• 2007-2012 Literal Record of veteran game developers’ lecture
• 2009- Interview with Indie-game developers(~20 developers)
• 2016- Interview with game developers about 80’s -90’s game
industry
leader:Kobayashi(Hichibe)Nobushige
Memober:Yusuke Koyama, Akito Inoue, Youichiro Miyake
23. Digital Game Research Association(DiGRA JAPAN)
会長:Prof. Akinori Nakamura (Ritsumeikan Univ)
• Oral History Task Force
• Leader Prof. Toru Iwatani (Tokyo Polytechnique Univ 、former-
DiGRAJ president)+10 members
• 2007-2012 Literal Record of veteran game developers’ lecture
• 2009- Interview with Indie-game developers(~20 developers)
• 2016- Interview with game developers about 80’s -90’s game
industry
25. DiGRA JAPAN ORAL HISTORY PROJECT
Game Business Archive
Interview data Inerviewee
No.1 2017/5/10 Hideki Sato (Former SEGA president)
No.2 2017/7/11 Naoyoshi Aoki (Former SEGA director of game development and second party )
Yasuhide Kobayashi (Former SEGA director of second party)
No.3 2017/10/27 Masamitsu Niitani (Former president of COMPILE)
Hiroyuki Takahashi (Former president of SONIC, president of CAMELOT )、
Masato Maekawa (President of TREASURE)
No.4 2017/10/27 Masato Watanabe (Mega Drive Fan, first editor in chief )
Koji Aizawa (Mega Drive Fan, second editor in chief )
Yoji Kawaguchi (BEEP, first editor in chief )
No.5 2018/6/27 Yuji Naka (Formet SEGA main programmer, Game Producer)
https://kawag2017.jimdo.com
26. Third Party
Third Party
Management
Game Machine
Development
Game
Development
Sales and Management of
Game Machin and software
Advertisement and
Marketing
Second party
Game Shop
Game
Distribution
Book and magazine
publisher
TV
Yoichi Kawaguchi’s editiion
29. Contents
Chapter 1. Oral History project in DiGRA JAPAN
Chapter 2. The Two types of Game Preservation
Chapter 3.
Overview of Oral History of Game Development technologies
Chapter 4 Youichiro Miyake’s Works
Chapter 5. Summary
30. The Two types of Game Preservation
Game
Package
Game
Developmen
t
Materials
Game
Package
Game
Package
Game
Developmen
t
Materials
Game
Developmen
t
Materials
31. Game
Package
(Data for final build )
Final Program
Final Data(Asset)
Final Tool
(Design document)
Final Game design documents
Final Art Design
(true)Game Developers Credit
Game Package released and published
Binary data
Game Developers Credit
Intermediate data
Program
Data
(Intermediate Design
Document)
Game design documents
Rough Sketch
Conversation data (e-mail,
oral, slack…)
Some
millions
of line
Some
hundreds
of tera
byte
Game
Developmen
t
Materials
32. In “Ready Player One”, game development are recorded as video-
data, and anyone can see all movies in the library.
That is very ideal case !
http://wwws.warnerbros.co.jp/readyplayerone/
33. The Two types of Game Preservation
Game
Package
Game
Developmen
t
Materials
Game
Package
Game
Package
Game
Developmen
t
Materials
Game
Developmen
t
Materials
History of game
34. The Two types of Game Preservation
Game
Package
Game
Developmen
t
Materials
Game
Package
Game
Package
Game
Developmen
t
Materials
Game
Developmen
t
Materials
ゲームの歴史
History of game
Development
History of game
35. • Game History
A History as a commercial package and of users
• Game Development History
A History as game development and game developers
The Two types of Game Preservation
36. The Two types of Game Preservation
Game
Package
Game
Developmen
t
Materials
Game
Package
Game
Package
Game
Developmen
t
Materials
Game
Developmen
t
Materials
ゲームの歴史
History of game
History of game
Development
37. The Two types of Game Preservation
ゲーム
パッケージ
ゲーム開発
資料
ゲーム
パッケージ
ゲーム
パッケージ
ゲーム開発
資料
ゲーム開発
資料
ゲームの歴史
Game Package public in the world
Game Development Materials and
information public by interview, paper,
thesis, articles, and books.
History of game
History of game
Development
41. Open Sources
• John Carmack released
DOOM SOURCE CODE in 1997
• BLACK BOOK (Quake technologies)
Michael Abrush published in 1997 (free online in 2001)
https://github.com/id-Software/DOOM
http://www.drdobbs.com/parallel/graphics-
programming-black-book/184404919
46. Two flow of Game Preservation
Game
Package
Game
Developmen
t
Materials
Game
Package
Game
Package
Game
Developmen
t
Materials
Game
Developmen
t
Materials
ゲームの歴史
History of game
History of game
Development
47. Two flow of Game Preservation
ゲーム
パッケージ
ゲーム開発
資料
ゲーム
パッケージ
ゲーム
パッケージ
ゲーム開発
資料
ゲーム開発
資料
ゲームの歴史
Game Development Materials and
information public by interview, paper,
thesis, articles, and books.
History of game
History of game
Development
Intervirew
Book
Paper
48. Problem
• Some information are published to the occasional chance, but
they are not organized.
• Any game developer cannot understand a whole image of big
scale of modern game, and we must reconstruct the whole image
of game development from the game development materials.
• How we should ?
49. Contents
Chapter 1. Oral History project in DiGRA JAPAN
Chapter 2. Two flow of Game Preservation
Chapter 3.
Overview of Oral History of Game Development technologies
Chapter 4 Youichiro Miyake’s Works
Chapter 5. Summary
50. Miyake’s history
Joining Game Industry in 2004
There is no information of AI in game industry
History of Game Development ≠ History of Game
I keep reconstructing a history of game AI for technical design of a
new game title
It is very useful, and I shared the reconstructed game AI history
51. Important points to reconstruct a game
development history
What technologies are used in which game title should be
fixed.
• Reconstructing the history of game development should be
shared, because to gather information of game development
cooperation of game industry is required.
53. Reconstructing the history
• Connecting points makes a context of game development,
and it enables game developers to predict the next stage of
game development.
new
54. A large scale problem of game
development
• Anyone cannot understand all program (over some millions of
lines) and see all data assets (several hundreds of tera-byte)
• After game development, reconstructing to understand a
whole figure of game.
55. Current situation of game developer
Development
Publishing
PC broken and replacing(lost in personal folders)
Lost in Human brain
56. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
60. Japanese Society of Artificial Intelligence,No.1, 2019
Anyone can download the article:
https://jsai.ixsq.nii.ac.jp/ej/index.php?active_action=repository_view_main_item_detail&page_id=13&block_id=23&item_id=9670&item_no=1
66. PAC-MAN Game AI techniques
• Speed and spawning timing control for each level
67. • 4 different AI (personalities) for 4 characters, and it makes
indirect cooperation.
• Characters’ behaviors makes game dynamics such as 4
characters surround a player but make a space to escape for
a player.
• Game difficulty control by decreasing numbers of enemies
when a player lost one Pac-man.
③ Speed and spawning timing control for each level
PAC-MAN Game AI techniques
68. ① Monsters’ Wave attack two modes
② Monsters’ characteristic behaviors and indirect cooperation
④ Speed control for each level
③ Spawning timing control
PAC-MAN Game AI techniques
69. ① Monsters’ Wave attack two modes
② Monsters’ characteristic behaviors and indirect cooperation
④ Speed control for each level
③ Spawning timing control
PAC-MAN Game AI techniques
70. AI Monsters’ Behavior Policy: PAC-MAN
• Wave Attack … 4 monsters attack to a player with all
of them at once.
• Two mode: Surrounding mode and dispersion mode
74. ① Monsters’ Wave attack two modes
② Monsters’ characteristic behaviors and indirect cooperation
④ Speed control for each level
③ Spawning timing control
PAC-MAN Game AI techniques
75. PAC-MAN Gaor me AI techniques
• Characters’ behaviors makes game dynamics such as 4
characters surround a player but make a space to escape for
a player.
76. Monster Wave Attack
Surrounding Dispersion
AKA
Go to the Pac-man’s position
(8x8dot)
Go to the upper right of play
field and move around it .
PINK
Go to 3 cell forward position
to Pac-man’s position
Go to the upper left of play field
and move around it .
CYAN
Go to the symmetry positon
to AKA’s position around
Pac-man’s position.
Go to the lower right of play field
and move around it .
ORANGE
Within 130 dots from Pac-man,
it has the same character of
AKA, and Out of the 130 dots,
it moves randomly.
Go to the lower left of play field
and mover around it .
Wave Attack two modes and monsters’ behavior
77. PAC-MAN Gaor me AI techniques
• 4 different AI (personalities) for 4 characters, and it makes
indirect cooperation.
130dots
Go to p 3 glids forward to layer’s positi
Random moving
78. ① Monsters’ Wave attack two modes
② Monsters’ characteristic behaviors and indirect cooperation
④ Speed control for each level
③ Spawning timing control
PAC-MAN Game AI techniques
79. PAC-MAN Game AI techniques
• Game difficulty control by decreasing numbers of enemies
when a player lost one Pac-man.
2 monsters stays here
in some seconds.
80. 2 3 4 monsters
A
B
C
Number of Monsters to be spawn
244
30個 90個
3 4 monsters
50個
4monsters
(Count of eaten foods)
81. ① Monsters’ Wave attack two modes
② Monsters’ characteristic behaviors and indirect cooperation
④ Speed control for each level
③ Spawning timing control
PAC-MAN Game AI techniques
82. PAC-MAN Game AI techniques
• Speed and spawning timing control for each level
Speed
4 types of
Speed level
83. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
86. Dynamic Difficulty Adjustment in Xevious
• 『Masanobu Endo’s Xevious Seminar』(1987、FujiTV).
• Xevious has a simple AI. The AI judges a player’s skill, and AI
spawns strong enemies to a player, and spawns weak
enemies to a weak player.
87. • Auto-adjusting system od difficulty of game level = Meta-AI
Principle of Dynamic Difficulty Adjustment
in Xevious
Enemy 0
Enemy1
Enemy 2
Enemy3
Enemy4
Enemy5
Enemy spawning schedule table
When a pleyer is destoyed,
the spawning table is winded back.
88.
89. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
90. 80’s Namco Game Engine
“Job-con””Obj-con”
• Time to develop:Former of 1980
• Developer: Shoichi Fukatani (1954-1985)
• He developed common game engine in Namco to develop
many game titles effectively
• There becomes little bug by using the game engine for anyone.
• Many Namco arcade games used it.
• The engine is the origin of game engines in Japan.
91. 深谷正一 Shoichi Fukatani (1954-1985)
• He was a technical leader in Namco in 1975-1985.
• Many veteran Namco developers respect him.
• He made fundamental parts of game technologies in Namco.
•
https://www.mobygames.com/developer/shoichi-fukatani/credits/developerId,505419/
ENDING CREDIT of 源平討魔伝(NAMCO, 1986)
92. Interview with Kurosu Kazuo 黒須一雄
• I interviewed with Kurosu Kazuo.
• Kurosu Kazuo is a game programmer.
• He was Namco programmer, and he used “Job-con” “Obj-con”.
• [1]Youichiro Miyake, Interview with Kazuo Kurosu、DCAJ-report p.301-305.
http://www.dcaj.or.jp/project/report/pdf/2008/dc_08_03.pdf
• [2] Youichiro Miyake, Interview with Kazuo Kurosu、DCAJ-report p.306-330.
http://www.dcaj.or.jp/project/report/pdf/2008/dc_08_03.pdf
• [3] Youichiro Miyake , the origin of task system ,
Digital Games Research Association JAPAN Proceedings of 2016 Annual Conference
95. Easy chart of “Job-con””obj-con”
Object
Object
Object
Program Address
Stack
Program Address
Stack
Program Address
Stack
Task
Task
Task
Task
96. “Job-con””Obj-con” system
• Game Engine by Assemble language
• Any can make a game only by writing an object code
• No requirement to change a main routine
• New programmers learns “Job-con” and “Obj-con” when they
joined Namco.
• In Namco, technical texts are distributed as “all the mysteries
of video game program”: 4 texts called “天・地・人・虎の巻”.
97. Why is the Game Engine made ?
• 1980 is the turning point of Namco.
• New young developers joined Namco from University.
• Education was required.
• United game development environment was required.
98. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
99. Mahjong game in “Gambler-jiko-chushin-ha”
(GAME ARTS, 1987).
Game developer: Yoichi Miyaji(At the time:president of Game Arts)Interview(2017)
101. The design document is preserved in perfect form by Yoichi Miyaji.
Mahjong game in “Gambler-jiko-chushin-ha” (GAME ARTS, 1987).
102. He founded “GAME ARTS”
when he was a University student.
Yoichi Miyaji talked about his childhood
And Akihabara in 70’s
103. GAME ARTS (1985-)
1985 ->
->Company established in Setagaya-Ku Tokyo. Commenced planning and development of entertainment software for PC.
Released [Thexder]
Released [Cuby Panic]
1986 ->
Released [Silpheed]
1987 ->
Released [Gyuwambler Jikocyushinha]
->Moved headquarters to 2-9-9 Minami-Ikebukuro, Toshima-ku Tokyo.
Released [Solitaire Royale]
Released [Zeriard]
1988 ->
Released [Gyuwambler Jikocyushinha2 Jisyou! Kyougoujanshi-Hen]
Released [Professional Baseball Family Stadium]
Released [Veigues]
Released [Gyuwambler Jikocyushinha3 Boukyou Sasuraijanshi-Hen]
1989 ->
Released [Gyuwambler Jikocyushinha Mahjong Puzzle Collection]
Released [Fire Hawk]
Released [Jan-Oh Touryumon]
Released [Professional Baseball Family Stadium '89]
https://www.gamearts.co.jp/en/company/history_of_corp.html
104. • Utility-based AI : give a move to numerical value
• Estimating (how to decide a vaule) is difficult, but Yoichi Miyaji
finalized the Utiliy-based AI by keeping playing Mahjong for
two years.
“Gambler-jiko-chushin-ha” Game AI techniques
106. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
110. • Not exact speech recognition
• When SEAMAN cannot understand a user’s speech, he ignores
it, and begins to lead the conversation by himself.
• SEAMAN has some different pronunciations for one speech. It
makes difference of meaning by pronouciation.
SEAMAN Game AI techniques
111. Oral history of game technologies
(1) AI in PAC-MAN (Toru Iwatani interview,2018)
(2) Meta-AI in Xevious (Masanobu Endo, 1987)
(3) Game Engine of Namco in 80’s(Kazuo Kurosu interview,2008)
(4) AI in Mahjong (Yoichi Miyaji interview, 2017)
(5) AI in Seaman(Yutaka Saito interview, 2017)
(6) Ganbare Morikawa-kun 2go, Astronoka, Kuma-uta (Yukito
Morikawa, 2019)
113. 1997「GANBARE MORIKAWAKUN 2GO」 (PLAYSTATION)
Character Control by
Neural Network
三宅 陽一郎「ゲームAI技術20年の進化とこれから」 https://cedil.cesa.or.jp/cedil_sessions/view/1943 森川幸人「やってきたこと」より引用
Action
Player orders an action Player orders an action
Reaction from an item
Decision Making
Learning from the result
AI!
115. Color1
Color 2
Form
Smell
Sound
Behavior
…
Take an action
Knock
Kick
Take back
Eat
Jump
Push
おいしい
快い
不快
Impression from item
Nourishing
Enjoyable
Damaged
Fear
Fine
…
Table Data
…
…
…
…
Texture
Neural Network with
Backpropagation in game
ゲーム中にBP学習される
ニューラルネットワーク②
Put it on
the head
116. アイテム#1の属性
Impression from Item1
Impression from Item2
…
Impression from ItemN
State Impression
PiT’s condition
Policy at the stagae
Go to the next stage
Finding a food
Finding the best item
Go home
Searching the best item
Pre-trained Neural Network
Color1
Color 2
Form
Smell
Sound
Behavior
…
Texture
Neural Network with
Backpropagation in game
117. 1998「Astronoka」 (PlayStation)
Generating parameters of character specification
by genetic algorithm.
三宅 陽一郎「ゲームAI技術20年の進化とこれから」 https://cedil.cesa.or.jp/cedil_sessions/view/1943 森川幸人「やってきたこと」より引
Trap Jump
Break
Round
about
Creatures evolves to adapt the
environment
118. Genetic Algorithm
Make a group evolve in one direction
First generation New generation(100~)
…
Evolution by generation
One generation produces a next generation
119. Genetic Algorithm
遺伝子
Gene
Next
Generation
Parent ①
Parent②
Pick up two parents
from base group
Crossover two genes Produce new generation
(selection) (crossover) (production)
Present Generation
The iteration makes a desired generation in the environment
遺伝子
Gene
120. ②Simulation and Evaluation
In the environment, after all NPC take actions, they are evaluated.
Order Evaluation
1st 86.3
2nd 78.4
3rd 75.3
..…
…
100th 38.2
Evaluation function
(Example) If you want to make strong NPC
Evaluation function=0.7*Number of enemy it defeated+0.3* HP Remains
If you want to make survival NPC Evaluation function= time to survive
Game World
敵
Enemy
How an NPC adjusts to the world ?
Evaluate a gene by the result in the world of the NPC who has the gene.
Enemy
Enemy
121. From Evaluation Value to Fitness Value
Order Fitness
1st 9.32
2nd 8.83
3rd 7.81
..…
…
100th 0.02
Transform evaluation value to fitness value
Evaluation value is score.
Fitness value shows how the agent fit the world.
Order Evaluation
1st 86.3
2nd 78.4
3rd 75.3
..…
…
100th 38.2
122. (Example)④ Astronoka
Initially, they are trapped easily… They learn to avoid many traps.
MuuMuu, 「Astronoka」(Enix, 1998)
http://dlgames.square-enix.com/jp/psga/2008/astronoka/
http://www.muumuu.com/product.html
New Generation(5~)
First Generation
Want to
Eat
vegetables
The best AI game in the world that uses Generic Algorithm
( Classical and great work )
123. What is Astronoka?
Player grows vegetables in a field.
NPC comes to eat them.
Player makes traps around the field to protect vegetables.
Player gets a prize by growing good vegetables!
http://dlgames.square-enix.com/jp/psga/2008/astronoka/
http://www.muumuu.com/product.html
MuuMuu, 「Astronoka」(Enix, 1998)
124. Flow chart
of GA
A player makes traps
Beginning of a day
Trap battle begins
Trap battle ends
Evaluation
Fitness value
Order
Delete the last two NPC
Pick up two parents from NPCs with high fitness value
Generate two new NPCs
Generate new generation
Get enough number of new NPC
The day ends
Generation exchange
rate changes
Mutation rate changes
125. 4-① Gene mapping
Gene code and initial parameter set
56 x 8 =448 bit
遺伝子
Height power Run speed Patient
Weight Patient for Light
1.87 6.85 16.25 25.03 25.03 16.25 6.85 1.87
0 1 2 3 4 5 6 7
[Bit weight]
[NPC attribute (56 parameter)]
126. 4-② Simulation and evaluation
Evaluation score is proportional to how long NPC go into the field
avoiding and breaking traps.
G
Fitness=Score+Time*0.3+Enjoy param*0.5+Trap_Score+Safety+HP*0.5
Time for clear
G
C B A B C
C B C
C
S
G
A
B
C
Start
Bonus =100
Goal
Bonus =50
Bonus =30
Bonus =10
Bonus =0
127. GA improvement in Game system
Average of fitness
A curve of fitness a day
Generations
(1) To give a feeling of evolution to users,
the game system evolves 5 generations a day by GA.
(2) To make constant speed of evolution of one day,
the game system controls the number of iterations of GA evolution.
129. Knowledge
DB
User Data DB Phrase DB
Melody Parts DB
Phoneme DB
Keyword
Generation
Melody
Generation
Voice
Generation
Lyrics
Generation
User
Evaluation
User
Evaluation
130. Chapter 4 Youichiro Miyake’s Works
(7) Chome Hounds (2006) and Armored Core V (2012)
(8) FINAL FANTASY XV (2016)
132. FINAL FANTASY XV AI techniques book
• Book of AI for FINAL FANTASY XV
• In 2019
• 15 developers wrote the article. I
planned this book two years ago
• I asked SQUARE-ENIX to allow
publishing the book.
• The book was published in 2019.6
• …But in Japanese only
• Now I searching the Published which
can publish the book in English.
137. レベル
Character AI
Enemy
character
Player
情報獲得
Spatial AI
Meta-AI
Order
Ask &
Report
ゲーム全体をコントロール
Support
query
query
頭脳として機能
MCS-AI dynamic co-
operative model
Dynamic allocation of enemies
Observing level in real-time
Direction for agents
Making progress of game
Autonomous thinking
Cooperation
Team AI
Preparing data to make Meta-AI and
Character AI recognize the level
Managing object representation
Managing Navigation data
Path-finding / Tactical point analysis
138. Contents
Chapter 1. Oral History project in DiGRA JAPAN
Chapter 2. Two flow of Game Preservation
Chapter 3.
Overview of Oral History of Game Development technologies
Chapter 4 Youichiro Miyake’s Works
Chapter 5. Summary
139. Reconstructing 80’s game AI technologies
• Namco’s first game engine “Job-controller”
• Game AI (Character AI)begins from “PAC-MAN”
• Meta-AI begins “Xevious” and Iwatani
• Utility-based AI in “Gambler-jiko-chushin-ha”
Game industry has lead the game AI techniques, but
the technologies has not been shared by my interview.
140. Reconstructing 80’s game AI technologies
• Namco’s first game engine “Job-controller”
• Game AI (Character AI)begins from “PAC-MAN”
• Meta-AI begins “Xevious” and Iwatani
• Utility-based AI in “Gambler-jiko-chushin-ha”
80’s
141. Reconstructing 80’s game AI technologies
• Namco’s first game engine “Job-controller”
• Game AI (Character AI)begins from “PAC-MAN”
• Meta-AI begins “Xevious” and Iwatani
• Utility-based AI in “Gambler-jiko-chushin-ha”
• SEAMAN’s Ai
• Morikawa’s works (Neural-Net, GA, auto-generating)
80’s
90’s
142. Reconstructing 80’s game AI technologies
• Namco’s first game engine “Job-controller”
• Game AI (Character AI)begins from “PAC-MAN”
• Meta-AI begins “Xevious” and Iwatani
• Utility-based AI in “Gambler-jiko-chushin-ha”
• SEAMAN’s Ai
• Morikawa’s works (Neural-Net, GA, auto-generating)
80’s
90’s 90’s game industry did not succeed
to 80’s game AI technologies because
they didn’t share knowledge.
143. Problems and the Future
• Sharing game technologies is very important for game
industry.
• Ever game preservation is only preservation of published
game titles.
• Game development preservation should begin.