xiaoyuan tu and demetri terzopoulos, dept. of cs, university of toronto

18
Mar. 30, 2001 Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto Artificial Fishes: Physics, Locomotion, Perception, Behavior Presentation by Siddharth Dalal

Upload: fuller

Post on 05-Jan-2016

48 views

Category:

Documents


0 download

DESCRIPTION

Artificial Fishes: Physics, Locomotion, Perception, Behavior. Mar. 30, 2001. Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto. Presentation by Siddharth Dalal. Intro & Background. What do fish do? eat, survive, when compelled by their libidos…. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Mar. 30, 2001

Xiaoyuan Tu and Demetri Terzopoulos,

Dept. of CS, University of Toronto

Artificial Fishes: Physics, Locomotion, Perception, Behavior

Presentation by Siddharth Dalal

Page 2: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Intro & Background

• What do fish do?– eat, survive, when compelled by their libidos….

• Physics based graphic modeling

• Worm Dynamics, facial model– more sophisticated spring mass model

• advanced behavioral animation

Any fish is good if caught on the hook.

Page 3: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Overview

Intention focuses sensory data causing behavior

Page 4: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Fishics 1 - Mechanics• Spring Mass Model

wiiii fw

dt

dxq

dt

xdm

2

2

m = massx = positionq = damping factorw = net force due to springsf = external force

Page 5: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Fishics 2 - Hydrodynamics• Swimming - Muscles + Hydrodynamics

Page 6: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Fishics 3 - Motor Controllers• Swim MC• Left and right MC

• Anterior and Posterior of fish - r1, s1, r2, s2• Max params scaled from 0 - 1 to produce varying speeds

Page 7: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Sensory Perception

• Two on board environment sensors:– Vision Sensor - extracts information from scene

geometry, object database, physical simulation. Cyclopean(?) vision - 300o viewing angle.

– Temperature sensor - senses ambient temp. at center of body

Page 8: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Behavio(u)r 1

• Intention based on– Habits– Mental State– Incoming Sensory Information

• decides behavior routine

• incremental - needs memory

Page 9: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Behavior 2 - Habits and Mind

• Habits - does fish like brightness, schooling, male or female (yes this is in habits)

• Mental State– Three mental states - HLF - hunger, libido, fear– H= min[1-n(t)R(Δt)/α, 1]– L=min[s(Δt)(1-H(t)), 1]– F=min[Σf, 1], f=min[D/d(t), 1]

(Fish like sex after dinner )

Page 10: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Intentions 1• Intentions

– avoid,– escape– school– eat– mate– leave– wander

Page 11: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Intentions 2

• Features of Generator– Persistence in intentions - no dithering– focusser - focus on most important intention

• Create ‘abnormal fish’– warp intentions

Page 12: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Intentions 3

• Behavior routines:– eight - avoid static obstacle, avoid fish, eat,

mate, leave, wander, escape, school– chasing target subroutine– other subroutines - looping?, circling,

ascending?, nuzzling

Page 13: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Fish Type = Warped Intentions

• Artificial Fish Types– Predators

• don’t escape, mate or school

• always cruise, so don’t leave

Page 14: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Fish Type = Prey Fish Grey Fish

• Artificial Fish Types– Prey

• school

• evade predators

Page 15: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Pacifists

• Artificial Fish Types– Pacifist

• no school, no escape

• just mate

• complex mating behavior implemented…– fish i chooses partner j

– criteria if i is female/male

– looping, circling, chasing-target, nuzzling

– etc.

Page 16: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Result

• 10 fish, 15 food particles, 5 static obstacles at 4fps on SGI R4400 Indigo2

• Future:– reproduction– other work

Page 17: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Links

• http://www.dgp.toronto.edu/people/tu/tu.html• http://citeseer.nj.nec.com/tu94artificial.html• http://www.cs.toronto.edu/~dt/

Page 18: Xiaoyuan Tu and Demetri Terzopoulos, Dept. of CS, University of Toronto

Guests and fish start to stink after two days.