MEIS 2015 : A Multilayered Model for Artificial Intelligence of Game Character as Agent Architecture
5 likes2,917 views
The document presents a multilayered model for artificial intelligence in game character architectures, focusing on various AI structures including blackboard and subsumption architectures. It discusses the evolution from classical to modern AI in digital games, emphasizing the role of AI in enhancing character behavior and interaction with environments. Furthermore, it highlights the significance of designing information flow within AI systems to achieve effective decision-making and character responsiveness.
MEIS 2015 : A Multilayered Model for Artificial Intelligence of Game Character as Agent Architecture
- 1. A Multilayered Model for Artificial Intelligence of Game Character as Agent Architecture* Youichiro Miyake @miyayou https://www.facebook.com/youichiro.miyake [email protected] 2016.9.26 (Sat) @Nishijin Plaza Kyushu University http://mcg.imi.kyushu-u.ac.jp/meis2015/
- 2. Summary • I am a game developer in game industry for more than 10 years. • My research is focused on AI in digital game. • There are some mathematical structures in AI. • I want to show it in talk…
- 3. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 4. INTRODUCTION
- 5. Introduction • From classical AI to modern AI. • It means from scripted AI to autonomous AI.
- 7. Level Navigation AI Meta-AI Character AI 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 Support Enemy character Player Information Acquisition Control Brain http://dear-croa.d.dooo.jp/download/illust.html http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 8. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 10. Technological Innovation Time Scale The Industrial Revolution Information Revolution Network Revolution Intelligence Revolution Automation Computerization Networked Intelligent The 2nd Industrial Revolution Motorization 1750 1860 1960 1990 Now… Into “Intelligent” Wolrd
- 11. Society and AI Robot Generation Poppulation AI ageing society with fewer children Robot and AI will support ageing society with fewer children © 2015 SQUARE ENIX CO., LTD. All rights reserved
- 12. Intelligent Game System Game System Intelligent Game System Digital Games also has become intelligent. So, what is “Intelligent Game System” ?
- 13. What is digital game ? • Interactive Digital Space • The space becomes structured. • AI becomes a module
- 14. The System of digital game AI • A role of AI in digital game become clear. • There are three roles of AI. AI to control a whole game system AI of character (Character’s Brain) AI to recognize an environment in game by analyzing a level, terrain, and space. Meta AI Character AI Navigation AI
- 15. Two Axis to categorize AI • The problem of “What is AI” is very difficult problem. • We had better know what problem AI is seeking to resolve. Space Scale Time Scale Local Short time Global timeless Global Long time Meta-AI Character AI Navigation AI
- 16. Level Navigation AI Meta-AI Character AI 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 Support Enemy character Player Brain Information Acquisition Control http://dear-croa.d.dooo.jp/download/illust.html http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 17. Environment and Character Environment Character Character has interactions with an environment. http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 18. Environment and Character Environment Character Character has interactions with an environment. IN PUT http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 19. Environment and Character Environment Character Character has interactions with an environment. IN PUT OUT PUT http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 20. Environment and Character Environment Character Intelligence Character has interactions with an environment. IN PUT OUT PUT http://www.kaiyodo.co.jp/revoltech/yamaguchi_2012.html
- 22. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Information Flow Memory Internal Cyclic Information Flow Sensor Effector Intelligence Agent Architecture
- 23. Information Flow • Information flow connects the outer (environment) and the inner(=intelligence). • This is caused by mental activity to organize the inner mind’s dynamics. Information Flow Internal Cyclic Information Flow Making AI is achieved by designing Information flow. In actually, there are many information flow inside a mind.
- 24. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 26. Blackboard KS KS KS KS KS KS Arbiter Bruce Blumberg , Damian Isla, "Blackboard Architectures", AI Game Programming Wisdom (Charles River Media) , 2002 Blackboard Architecture (Knowledge-based AI) Each KS (Knowledge Source) has special knowledge for a problem. KS read and write information on Blackboard. Arbiter controls all of KS. All KSs communicate with each other via Blackboard.
- 27. Hierarchical Blackboard Primitive data Analyzed data More analyzed data More analyzed data More analyzed data KS KS KS KS
- 28. Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816 Hierarchical Blackboard (Example)
- 29. Hearsay-II Voice Analysis Process Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816
- 30. Hearsay-II Voice Analysis Process Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816 Blackboard
- 31. Hearsay-II Voice Analysis Process Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816 Blackboard KS (Knowledge Source)
- 32. Hearsay-II Voice Analysis Process Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816 Blackboard Arbiter KS (Knowledge Source)
- 33. Hearsay-II Voice Analysis Process Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816
- 34. Hearsay-II Voice Analysis Process PARAMETER SYLLABLE SEGMENT Erman et Al. The Hearsay-II Speech-Understanding System: Integrating Knowledge to Resolve Uncertainty, ACM Computing Surveys (CSUR) Surveys Homepage archive Volume 12 Issue 2, June 1980, Pages 213-253 http://dl.acm.org/citation.cfm?doid=356810.356816
- 36. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 37. AGENT ARCHITECTURE WITH BLACKBOARD ARCHITECTURE
- 38. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Information Flow Memory Internal Cyclic Information Flow Sensor Effector Intelligence Agent Architecture
- 39. Agent Architecture applied Blackboard Architecture
- 40. Agent Architecture applied Blackboard Architecture
- 41. Agent Architecture applied Blackboard Architecture Blackboard
- 42. Agent Architecture applied Blackboard Architecture KS (Knowledge Source)
- 43. Agent Architecture applied Blackboard Architecture KS (Knowledge Source) Modularization has advantage of game development, because each module can be developed and improved independently. And it becomes easy to add a module in the development. More intelligent creature has more intelligent modules in its architecture.
- 44. Agent Architecture applied Blackboard Architecture KS (Knowledge Source) Blackboard in Agent Architecture gives scalability and customizability.
- 45. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 47. Subsumption Architecture INPUT OUTPUT Time Abstraction (verbalizing) Reactive More Abstract thinking Theoretical thinking Abstract thinking primitive stimuli becomes language Making a body action Making a behavior
- 48. Classical : Central domain All processes of intelligent modules are executed in sequence. Subsumption : parallel & layered All processes of intelligent modules are executed in parallel. Rodney Brooks, A robust layered control system for a mobile robot Robotics and Automation, IEEE Journal of (Volume:2 , Issue: 1 ) 1986
- 49. Roomba (iRobot) http://www.irobot.com/en/us/learn/home/roomba.aspx Roomba has a subsumption architecture.
- 50. Subsumption Architecture INPUT OUTPUT Time Reactive R When a robot sees an object, It turns around.
- 51. Subsumption Architecture INPUT OUTPUT Time Reactive R When a robot sees an object, and Sees a cliff on the side It stops turning and gets back .
- 52. Subsumption Architecture INPUT OUTPUT Time Reactive More Abstract thinking Abstract thinking R When a robot hears a sound, It stops anyway by stopping all motions.
- 53. Subsumption Architecture INPUT OUTPUT Reactive More Abstract thinking Theoretical thinking Abstract thinking R When a robot cleans all rooms, It goes to energy station.
- 54. Subsumption Architecture INPUT OUTPUT Time Abstraction (verbalizing) Reactive More Abstract thinking Theoretical thinking Abstract thinking primitive stimuli becomes language Making a body action Making a behavior Agent Architecture applied subsumption architecture.
- 55. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Information Flow Memory Internal Cyclic Information Flow Sensor Effector Intelligence Agent Architecture
- 56. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Memory Sensor Effector Agent Architecture Hierarchical Recognition Synthesis of motions Knowledge Making Decision Making Motion Making Knowledge Making Decision Making Motion Making
- 57. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Memory Sensor Effector Agent Architecture Hierarchical Recognition Synthesis of motions Knowledge Making Decision Making Motion Making Knowledge Making Decision Making Motion Making
- 58. Game World （Environment/Level） Five senses Body Language Knowledge Representation Knowledge Making Decision Making Body Motion Making Memory Sensor Effector Agent Architecture Hierarchical Recognition Synthesis of motions Knowledge Making Decision Making Motion Making Knowledge Making Decision Making Motion Making Distributed Layered Architecture ＝ Subsumption Architecture
- 60. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. Modern Agent Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 61. MODERN AGENT ARCHITECTURE FOR A GAME CHARACTER
- 64. 身体の反射レベル Abstraction Abstraction Reduction Reduction Reduction World Object Object image on the lowest layer (Umwelt) Object image on the second layer Object image on the third layer Object image on the top layer Sensor Effector Artificial Intelligence
- 65. 身体の反射レベル Abstraction Abstraction Reduction Reduction Reduction World Object Object image on the lowest layer (Umwelt) Object image on the second layer Object image on the third layer Object image on the top layer Sensor Effector Artificial Intelligence Information Flow
- 66. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt (Environment World) 8. Hierarchical structure / commutative diagram / Homomorphism in agent architecture 9. Summary
- 67. UMWELT
- 68. Umwelt • “Umwelt” means a subjective world which a living thing has. • “Umwelt” depends on its ecology.
- 69. Umwelt Jakob Johann Baron von Uexkül, Adolf Portmann, Streifzüge durch die Umwelten von Tieren und Menschen, ROWOHLT HAMBURG,1956 http://monoskop.org/Monoskop
- 70. WORLD BODY Umwelt A living thing grasps an objects in the world by two hands: functional effectors and perceptual sensors. Object Functional Organ (Body = Effector) Perceptual Organ (Sensor)
- 71. WORLD BODY Umwelt Object Functional Organ (Body = Effector) Perceptual Organ (Sensor) A subjective world is created by functional effectors and perceptual sensors.
- 72. WORLD BODY Umwelt Object Functional Organ (Body = Effector) Perceptual Organ (Sensor) Each living thing has its different body and sensors. A subjective world is created by functional effectors and perceptual sensors.
- 73. WORLD BODY Umwelt Object Functional Organ (Body = Effector) Perceptual Organ (Sensor) Each living thing has a different subjective world. = Umwelt
- 74. Bee’s UmweltObjective World Bee’s Umwelt Bees are interested in flowers. Jakob Johann Baron von Uexkül, Adolf Portmann, Streifzüge durch die Umwelten von Tieren und Menschen, ROWOHLT HAMBURG,1956 http://monoskop.org/Monoskop
- 75. Snail's Umwelt Research ①Put a snails on a ball. ②iterate to push and pull a bar in front of the snails. ③change a frequency of the iteratetion. Experiment Result In 1-3 /sec frequency, a snail did not go to the front, but Over 4 frequency, It went front. Conclusion A snails recognize the bar stops over 4 frequency. ＝In snails’ Umwelt, a frequency of sensors is under 4. (Human frequency is about 18.) Jakob Johann Baron von Uexkül, Adolf Portmann, Streifzüge durch die Umwelten von Tieren und Menschen, ROWOHLT HAMBURG,1956 http://monoskop.org/Monoskop
- 76. Snail's Umwelt Research A child chicken is tied to a bar with rope and closed in a transparent dome which prevent sound propagation. Experiment Result （Upper） Parent chicken ignored it. （Lawer） Parent chicken went to help a child. Conclusion In chicken’s Umwelt, voice is more important rather than visual sensor. ユクスキュル/クリサート、 「生物から見た世界」 （岩波文庫） Jakob Johann Baron von Uexkül, Adolf Portmann, Streifzüge durch die Umwelten von Tieren und Menschen, ROWOHLT HAMBURG,1956 http://monoskop.org/Monoskop
- 77. 身体の反射レベル Abstraction Abstraction Reduction Reduction Reduction World Object Object image on the lowest layer (Umwelt) Object image on the second layer Object image on the third layer Object image on the top layer Sensor Effector Artificial Intelligence Information Flow
- 78. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt (Environment World) 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 79. HIERARCHICAL STRUCTURE / COMMUTATIVE DIAGRAM / HOMOMORPHISM IN AGENT ARCHITECTURE
- 80. 身体の反射レベル Abstraction Abstraction Reduction Reduction Reduction World Object Object image on the lowest layer (Umwelt) Object image on the second layer Object image on the third layer Object image on the top layer Sensor Effector Artificial Intelligence Information Flow
- 81. Real World
- 82. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ Image of object Image of body (=self) R RRepresentation
- 83. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ Image of object Image of body (=self) R R R R
- 84. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ O’’S’’ Action: f’’ Sense: p’’ R R R R R R
- 85. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R
- 86. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R Sequence of Self Sequence of Object R R
- 87. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R Sequence of Self Sequence of Object
- 88. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R Sequence of Self Sequence of Object “Self” is a sequence of selfs. “Object” is a sequence of objects.
- 89. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R Sequence of Self “Self” is a sequence of selfs. “Object” is a sequence of objects.
- 90. Representation • “Representation” is essential concept of AI. • “Knowledge Representation” (KR) is representation of knowledge AI uses. • To transform “real things” to KR is the most basic techniques in AI.
- 91. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R Sequence of Self “Self” is a sequence of selfs. “Object” is a sequence of objects. Vector to move Vector to ent Dude, Where's My Warthog: From Pathfinding to General Spatial Competence, D. Isla, Invited talk, Artificial Intelligence and Interactive Digital Entertainment (AIIDE) 2005 http://naimadgames.com/publications.html
- 92. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R “Self” is a sequence of selfs. “Object” is a sequence of objects. Killzone 2 Multiplayer Bots Remco Straatman, Tim Verweij, Alex Champandard | Paris Game/AI Conference 2009, Paris, June 2009 http://www.guerrilla-games.com/publications.html
- 93. S (Body) O (Object) Action: f O’ Sense: p S’ Action: f’ Sense: p’ O’’S’’ Action: f’’ Sense: p’’ R R R R “Self” is a sequence of selfs. “Object” is a sequence of objects. Handling Complexity in the Halo 2 AI, D. Isla, GDC 2005 Dude, Where's My Warthog: From Pathfinding to General Spatial Competence, D. Isla, Invited talk, Artificial Intelligence and Interactive Digital Entertainment (AIIDE) 2005 http://naimadgames.com/publications.html
- 94. Contents 1. Introduction 2. Agent Architecture 3. Blackboard Architecture 4. Agent Architecture with Blackboard architecture 5. Subsumption Architecture 6. MODERN AGENT Architecture for A game Character 7. Umwelt 8. Hierarchical structure / commutative diagram / Homomorphism in Agent Architecture 9. Summary
- 95. SUMMARY
- 96. Summary • I am a game developer in game industry for more than 10 years. • My research is focused on AI in digital game. • There are some mathematical structures in AI. • I want to show it in talk…