Presentation is loading. Please wait.

Presentation is loading. Please wait.

DECOI: Social Simulation - August 20071 Social Simulation Branimir Cace, Carlos Grilo, Arne Handt, Pablo Rabanal & Scott Stensland.

Similar presentations


Presentation on theme: "DECOI: Social Simulation - August 20071 Social Simulation Branimir Cace, Carlos Grilo, Arne Handt, Pablo Rabanal & Scott Stensland."— Presentation transcript:

1 DECOI: Social Simulation - August 20071 Social Simulation Branimir Cace, Carlos Grilo, Arne Handt, Pablo Rabanal & Scott Stensland

2 DECOI: Social Simulation - August 20072 Outline ► The Scenario and How We Modeled It ► The Hypothesis and Our Observations ► Experimental Results ► Conclusions

3 DECOI: Social Simulation - August 20073 The Scenario                     

4 DECOI: Social Simulation - August 20074 Plants ► Plants are either nutritious or poisonous (binary) ► Each plant P has an attribute r P („redness“):  r P  [ 0 ; 1 ]  set randomly upon creation of a plant ► Global parameter t (poison threshold):  t  [ 0 ; 1 ]  A plant P is nutritious  r p < t (to be explained below)  determines probability of plants being nutritious ► Plants are created at random  sprouting at a random place in the simulation area  no inheritance of redness poisonous nutritious rPrP 1 0 t

5 DECOI: Social Simulation - August 20075 Agents - Basics ► Each agent A has an energy level e A  e A ≤ 0  A dies ► Each agent A has an age a A  a A  a A + 1 at each step(aging)  a A > s  A dies(max. lifespan)  s is a global lifespan parameter ► Agents move around randomly. ► Agents are either male or female. ► Agents can see the redness of plants.

6 DECOI: Social Simulation - August 20076 Agents - The Filter Gene ► Each agent A has a gene f A (filter):  f A  [ 0 ; 1 ]  If agent A encounters plant P: ► A will eat P  r P ≤ f A  Initially, f = 1 for all agents, i.e. in the beginning, every agent will eat everything ► Interpretation:  Threshold for the amount of redness of a plant an agent accepts, i.e.  If a plant is more red than f A, A deems it poisonous.

7 DECOI: Social Simulation - August 20077 Agents - Eating ► If agent A eats plant P: ► if r P < t : e A  e A + d(nutritious) ► if r P  t : e A  e A - d(poisonous) ► d is the „energy delta“ for nutrition and poison

8 DECOI: Social Simulation - August 20078 Agents - Mating (1) ► Two agents A and B will mate only if  both agents are old enough ► a A  m p and a B  m p (m p : puberty)  and both have enough energy ► e A  m e and e B  m e (m e : energy minimum for mating)  and they are sufficiently close to each other ► m d  distance(A, B) (m d : distance maximum for mating) ► Each simulation step, each female agent will select the male agent with the highest energy satisfying the above conditions for mating.

9 DECOI: Social Simulation - August 20079 Agents - Mating (2) ► Each parent spends a fixed portion of its energy  parents: e A  e A - m e and e B  e B - m e  child: e C  2 * m e ► Inheritance: child randomly chooses a parent‘s gene  f C  f A or f C  f B ► Mutation (slightly changing f c )

10 DECOI: Social Simulation - August 200710 Adaptation ► Purpose: Adapt the filter gene to the global threshold t, so that agents eat only nutritious plants. poisonous nutritious 1 0 t time filter gene f

11 DECOI: Social Simulation - August 200711 Task, Hypothesis & Observation ► The main task was to build and test a theory on an optimal age of puberty  But: is there actually an optimal age of puberty? ► Our preliminary observation from running a few simulations:  Not having a puberty leads to the fastest adaptation  Adaptation slows down with longer puberty ► So we conducted series of simulations  Using a poison threshold of t = 0.8  Testing puberty ages of 0, 5, 10, 20, 40 and 80 steps  Averaging over 5 simulation runs  Using Entorama and NetLogo as environments

12 DECOI: Social Simulation - August 200712 Implementing the Model EntoramaNetLogo Continuous space, fixed boundaries Discrete space for plants (patches), torus structure (wrap- around edges) unlimited space for plants limited space for plants Plant gets eaten by the closest agent Agent eating a plant is chosen randomly from set of present agents at a patch Children are placed randomly in the world Children start at the mother‘s location Starts off with an entire population, distributed randomly over the area Population is faded in during the first steps of the simulation

13 DECOI: Social Simulation - August 200713 Results from Entorama

14 DECOI: Social Simulation - August 200714 Results from NetLogo 5 of 5 died out 1 of 5 died out

15 DECOI: Social Simulation - August 200715 Conclusions ► In this scenario  Evolutionary adaptation works ► Agents evolve so that they distinguish nutritious from poisonous food  Our experiments do not support the existence of an optimal (wrt. adaptation speed) age of puberty ► The shorter the puberty, the faster adaptation ► But short puberty may create the risk of populations dying out ► To be investigated  Artifacts created by the different simulation environments (e.g. populations dying out) ...?


Download ppt "DECOI: Social Simulation - August 20071 Social Simulation Branimir Cace, Carlos Grilo, Arne Handt, Pablo Rabanal & Scott Stensland."

Similar presentations


Ads by Google