smelling food 
Submitted by zac elston on Mon, 2001-01-22 21:27

Has anyone got a good smeller? I have been playing with the frogs again
trying to let Smell sensors evolve on the legs but they never last past a
few million generations. Specifying longer life doesn't seem to encourage
smelling food over Distance hopping. I'd like to know if anyone has had
luck with evolving specific traits like this?

-zac

Forums: 
Users
Maciej Komosinski's picture

Submitted by Maciej Komosinski on Thu, 2001-03-22 14:02

> Has anyone got a good smeller? I have been playing with the frogs again
> trying to let Smell sensors evolve on the legs but they never last past a
> few million generations. Specifying longer life doesn't seem to encourage
> smelling food over Distance hopping. I'd like to know if anyone has had
> luck with evolving specific traits like this?

Seems that the S receptor requires too strong stimulus to work (and
that may be improved in v2).

But food finders' author managed somehow to evolve smelling creatures.
Perhaps it would be good to contact him and ask for his environment
parameters.

MacKo

Submitted by zac elston on Fri, 2001-03-23 21:33

what I have tried to do is add a food sensor to each leg
(x,x,x,x,x) > (x[s],x[s],x[s],x[s],x[s])

then wire the legs for each side with the middle leg (head?) as the tie breaker for which side will push less to turn to the food

something like

(
back leg
X[|1:1][@4:1][S:1],
front leg
X[S:1][|1:1],
then the brain + 2 junctions for potential of smell + 2 smell sensors
X[-2:.3,-3:.3,1:.3,2:1][S:1][!:2,=:.5,0:1][S:1][-2:-1,-1:-.3,3:-.3,2:-.3],
front leg
X[|-1:1][S:1],
back leg
X[S:1][@-4:1][|-1:1]
)

Once I get the brain oscillating at the right speed (don't have my notes handy now) then I should be able to tweak the smell potential to inhibit the legs on that side.

i think i need a better brain with more averaging (indexing?) of values to get a smoother motion, but that's the basic idea.

Written by: Zac Elston

Submitted by zac elston on Thu, 2001-04-19 18:15

well I found a basic design flaw here.

using inhibition to turn causes one obvious thing to happen. the fram stops when it gets close to the food. (DOH! I laughed very hard) so much for that idea. the reverse should be better, using excititory amplification on the side away from the food
should help, but Maciej is correct, the food stimuli is very powerful ..too powerful to use [S] directly as the sensors on all legs pick up the source with minimal discrimination between sides. I'm currently trying to use averaging to assist in this but
am having a difficult time.

Submitted by Tjalve Aarflot on Wed, 2001-03-21 18:38

"zac elston" wrote in message
news:94icd7$8k1$1@net.frams.poznan.pl...
> Has anyone got a good smeller? I have been playing with the frogs again
> trying to let Smell sensors evolve on the legs but they never last past a
> few million generations. Specifying longer life doesn't seem to encourage
> smelling food over Distance hopping. I'd like to know if anyone has had
> luck with evolving specific traits like this?
>
> -zac

Have been playing with your frogs also. I let Life span give a huge credit
to fitness, so they have to find food to live long enough. Though, the frogs
may detect food, but their mechanic structure doesn't allow them to turn
around in a specific direction....or at least it seems so for me. They are
just jumping around at free will.

Are there an effective way to make them more able to turn after the food???

Tjalve