1 / 31

Beyond Behavior

An Introduction to Knowledge Representation. Beyond Behavior. Damián Isla, Naimad Games Peter Gorniak, Rockstar. Knowledge Representation. We spend a lot of time on what our AIs do but very little time on what they know One of the great neglected problems of [game] AI This talk:

winslow
Download Presentation

Beyond Behavior

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. An Introduction to Knowledge Representation Beyond Behavior Damián Isla, Naimad Games Peter Gorniak, Rockstar

  2. Knowledge Representation We spend a lot of time on what our AIs do but very little time on what they know One of the great neglected problems of [game] AI This talk: • Introduce techniques • Agitate

  3. “Behavioral” Knowledge? • Behavioral knowledge • When to run away, when to shoot, when to flank left, etc. • Does an ant “know” where the anthill is? • State Knowledge • What is true about the world

  4. The Point of KR Perception of a thing != the thing itself

  5. The Point of KR Agent Object 1 Pathfinding Behavior Object 2 Animation Object 3

  6. The Point of KR Agent Object 1 KR Pathfinding Behavior Object 2 Animation Object 3 Decisions about action Decisions about perception and interpretation

  7. Why is KR Interesting? • Fun • exploit mistakes / limited perception • new modes of interaction • Lifelike • reason about AI as thinking perceiving creatures • emotional reactions • We’re doing it already anyway • Search for better representations == Search for more expressive power • build behavior out of better primitives

  8. Timescales “Guy X is behind the crate” “I have three bullets left” “That car is coming towards me” “dogs are animals” “birds have wings” “pushing the button calls the elevator” ∞ “Bobby is 5 years old” “Jane is spending the semester in France.” This instant

  9. 3 Key Concepts • Confidence • How sure am I in the knowledge I have? • Salience • How important is the sensory data I’m getting? • Prediction • What do believe will happen given what I’ve seen and what I know?

  10. Demo

  11. Behavior update void s_agent::behavior_update() { if (!confused()) { s_pos2d pos; omap.get_target_position(&pos); move_to(pos); } }

  12. Behavior update void s_agent::behavior_update() { if (!confused()) { s_pos2d pos; omap.get_target_position(&pos); move_to(pos); } } +

  13. Expectation-related Emotions • Confusion • Surprise <Something I was confident in is confirmed FALSE> <Something I thought unlikely is confirmed TRUE>

  14. Expectation-related Emotions • Confusion • Surprise

  15. Target Lists

  16. Target Lists Agent Object 1 KR Pathfinding Behavior Object 2 Animation Object 3

  17. Target Lists Agent Object 1 Target 1 Pathfinding Behavior Object 2 Target 2 Animation Object 3 Target 3

  18. Target Lists Target Perceived data location (x,y,z) action shoot hitpoints 44 Derived data Threat 0.8 Target weight 0.9 “Intentions” hurt_me Allows AI to make mistakes 0.99 0.99 0.99 0.8 0.95 0.98 0.6 0.9 0.98 Shared computation + expressive power

  19. shoot_at_target shoot_at_target Example switch_to_knife switch_to_knife search_for_target search_for_target !

  20. Phil Carlisle Memorial Memory Slide

  21. PCMMS Volatile behavior state • Working memory • Short-term • Episodic • ??? Remember that Target Perceived data location (x,y,z) Target Perceived data location

  22. Challenge #1:Representational Versatility Solution: Polymorphism Nazi Truck Wheel Fence Grass My hand

  23. Polymorphism Percept DAG (Synthetic Characters, MIT Media Lab, circa 2002)

  24. Challenge #2: Performance

  25. Challenge #2: Performance Object 1 KR Agent Agent Object 2 Agent Object 3 Shared KR

  26. Challenge #2: Performance X:location: <x,y,x> crates Object 1 Agent KR KR Agent Object 2 KR Agent Object 3 X:weapon: “pistol” KR enemies Hybrid KR

  27. Challenge #2: Performance Enemy 1 Agent Enemy 2 O A Agent Grass Salience Threshold O x A

  28. Benefits of Target Lists • Reasonable mistakes / limited perception • Shared computation • Expressive power

  29. Limitations of Target Lists • Relational information • Where does the notion of “behind” live? • Wholes and parts • Does a car’s wheel deserve it’s own representation? • A guy’s arm? • What about a mob of guys?

  30. Head Enemy has-a next-to Arm Enemy next-to Arm Representational Wankery behind holding Car has-a Hood Gun Wheel Wheel Wheel Wheel

  31. Wild Speculation Lazy Representation??? • Perception is active • Behavioral / emotional / motivational state changes the way you see the world • And WHAT you see in the world.

More Related