Transcript Evolving "elementary sight" strategies in predators via Genetic

Evolving "elementary sight" strategies
in predators
via
Genetic programming
ICBV Project
20.2.07
Lior Becker
Goals





Witness the evolution of the predator "strategy".
Imitate the evolution of the parts in the brain that
handle the visual informal interpretation .
Try to understand the development stages in the
strategy.
Try to analyze the usage of the photoreceptors as
part of the brain function .
Test if the development of sight strategy is a complex
process or can be emulated in a computer .
What is Genetic programming ?

Bio-Inspired

Inspired by Darwin’s evolutionary
principles

J.Koza style.
Charles Darwin
Principles
Competition
 Variation
 Overproduction
 Survival of the fittest

Population adaptation
Genetic programming
Main algorithm:
1. Generate the initial population.
2. Fitness evaluation.
3. Create new generation:
–
–
–
4.
Selection.
Cross Over.
Mutation.
Repeat until stop condition.
Genetic programming
Individual Representation


Individual is a Scheme-Like Function
Represented as a tree (AST).
Genetic programming
Recombination - cross over
Predator strategy through GP

World simulator
Predator
Prey

Process of work


Prey





GP.
Brain function.
Undeveloped eye
15 photoreceptors.
Moving ability.
Fitness: catching prey.
Tree components







Function
IFLTE , if less then.
PLUS , add 2 num.
PROGN2 , run r1 &
return r2.
TL, turn right, 5 Deg.
TR, turn left , 5 Deg.
MF, move forward.
MB, move backward.




Terminals
RP, resting potential.
AP, action potential.
P1 .. P15,
photoreceptors , 2 Deg.
MAXPP, max value of
the photoreceptors.
World simulator & Prey
WORLD



2D world.
100*100 Matrix.
Predator and prey can
be at any location.
PREY




Static prey.
Straight Line prey
Circle prey
Random prey.
Process of work





Evolving 51 generations, different preys.
Test cases: unlearned preys.
Plot fitness through time.
Recording movies.
Function analysis.
Results:
straight Line
prey
Results: test case


Test Case
Why is it important ?
Results: Fitness vs. generations


Improvement.
population
adaptation.
Results: Function
(IFLTE
(IFLTE P6 (PROGN2(IFLTE P3 P11 P13 P13 )(IFLTE P2 MAXPP MF P5 ))
(PROGN2 P4 P6 )(IFLTE AP MB P5 MB ))
(PLUS MAXPP P15 )
(PLUS(IFLTE P3 P1 MF P14 )(IFLTE TR MF P1 P12 ))
(PROGN2(PLUS P12 P10 )(PLUS P11 TL )))

Redundancy ? – Dead code.
(IFLTE
(IFLTE P6 (IFLTE P2 MAXPP MF P5) P6 (IFLTE AP MB P5 MB ))
(PLUS MAXPP P15 )
(PLUS(IFLTE P3 P1 MF P14 )(IFLTE TR MF P1 P12 ))
(PLUS P11 TL ))
Pi – photoreceptors; TL – turn left; TR – turn right; MF – move forward.
Results: photo receptors



External spreading.
Why ??
Human eye Diff.
Conclusions & discussion
1.
Predator strategy evolvement.
–
–
–
2.
3.
4.
Random strategy
Left/Right circle rotation strategy.
Combined (Left & Right) strategy.
External photoreceptors spared out.
Function redundancy, The key to new life.
None sophisticated strategies
“efficient chase”, why ?
Future work

More realistic 3D world.
–
–
–
–

Obstacles.
3D eye
3D world
Sophisticated preys.
Co-Evolution, prey and predator.
References








Darwin, Charles: On the origin of species by means of natural
selection. London, John Murray. (1859)
John R. Koza: Genetic Programming: On the programming of
computers by natural selection. MIT
Press, Cambridge, Mass. (1992)
John R. Koza: Genetic Programming II: Automatic Discovery of
Reusable Programs. MIT press,
Cambridge, Mass. (1994)
John R. Koza: Evolution of Subsumption Using Genetic Programming.
MIT press, Cambridge, Mass. (1993)
Holland, John H. Adaptation in Natural and Artificial Systems. Ann
Arbor, MI: University of Michigan Press (1975).
Haynes, Sen.: Evolving behavioral strategies in predators and prey,
University of Tulsa (1996).
Movie Data Base