Framsticks

Why does the humanoid below simply fall over ?

Originally it seemed to be a center of gravity problem so i made the
feet huge. Now it seems to have something to do with the inability of
the ankles to support the creatures wieght. I am very familiar with
neural nets but being new to framsticks , I am not sure how to fix the
problem.

# created Sun Jun 30 21:28:17 2002
# by Framsticks 2.0 (16 May 2002)
org:
name:stan
genotype:~
//0
p:-.2,.4,0,v=.1
p:-.2,0,0,v=.1
p:.2,.4,0,v=.1
p:.2,0,0,v=.1
p:-.14,.16,.4,v=.1
p:.14,.16,.4,v=.1
p:0,.2,.8,v=.1
p:0,.3,1.2,v=.1
p:-.14,.14,.95,v=.1
p:.14,.14,.95,v=.1
p:-.2,.2,.7,v=.1
p:.2,.2,.7,v=.1
p:0,.36,1.3,v=.1
p:-.1,.37,1.35,v=.1
p:.1,.37,1.35,v=.1
p:0,.38,1.44,v=.1
j:0,1
j:2,3
j:1,4
j:3,5
j:4,6
j:5,6
j:6,7
j:7,8
j:7,9
j:8,10
j:9,11
j:7,12
j:12,13
j:12,14
j:13,15
j:14,15~
info:(manually created)
num:1
popsiz:46
lifespan:4999.98
velocity:6.42579e-05
distance:0.321289
vertdist:2.67922e-05
vertpos:0.0507901
user1:
user2:
user3:

After some time experimenting with Framsticks, I have found that
evolving complex creatures beyond a certain limit is near impossible. I
believe that the problem is due to the population's fitness becoming
stuck in local maxima. I apologize if this conclusion is obvious, I am
still new to the field of ALife.
A solution might be to design an expdef file (in Framsticks v2) that
implements the rank-space method and keeps track of local maxima. I'm
not sure if this is feasable with the current version. It would involve
computing a diversity function for each individual (I believe a
similarity function is available, at least it was in v1), computing the
age of each individual, and its fitness. There would be two populations:
the normal "young" population, and one of "old" individuals that have
found local maxima (passed a certain age). The rank for each "young"
individual would depend on both its fitness and its similarity to the
"old" individuals, with a penalty for greater similarity. Only the
"young" population would reproduce, while the "old" population continued
to grow full of local maximum "failures". The simple "similarity"
function in v1 is different from this method in that I believe it causes
the population to oscillate around a local maximum, without ever forcing
it to leave the maximum, which is why the "old" population is necessary.
From my experiences with Framsticks, I believe that a challenge to
this method would be that there are very few fit intermediate
individuals (between maxima). Perhaps a very large, possibly variable
genepool size of "young" individuals would help. Also, it would be hard
to design a foolproof diversity function -- certain mutations might
allow an individual to remain at a local maximum while appearing to be
very different. Still, I think that this method might improve evolution.
Does anyone has any ideas about how to implement this in Framsticks,
or whether this method may or may not be relevant? I can't seem to
figure out how to write expdef files very well. Tell me if anything in
this post needs explaination, I didn't really explain what I meant by
"local maxima" etc. Thanks for your suggestions and ideas.

Craig

I was just wondering how you can limit mutations and evolution just to the
brain or just to the body. A lot of the posts have mentioned this, but I
can't find any settings for it in parameters. Thank you.
Joel Mathewson

I attempted to use the constant manipulation method in the brain window, as
has been mentioned in previous posts. It worked on the first simple creature
that I tried, but the sliding scales no longer pop up when I double click a
neuron or sensor of any sort. Even if I use the same initial creature it
worked on, the sliding scales will not appear. Do I have a setting wrong or
are the windows appearing behind something else or what is the problem?
Thank you for any advice.
Joel

MK,

I like Framsticks v2 immensely! I've put F2 on my web site
http://www.execpc.com/~kubos/interest.htm . Keep up the good work!

--
Cheers,
Uncle Ken

Hello.
Adam Rotaru have written this almoust two years ago:

"Currently this is not the case, beacause if creatures are better and eat
more, balls are introduced more frequently, so the number of available
balls is always the same. In other words, a perfect food finder cannot
starve other creatures to death by eating all the balls."

By following this idea we may gain a new quality of experiment.

The bottom line is:
the more food you will eat in the "spring, summer and autumn" (time where
everybody is trying to get as much energy as it can) , the bigger chance you
have to survive the winter (time when everybody eaten everything, and only
spend energy).

best
Borovik

Firs of all: How about something like that of the Golem project where a
creature from one gene pool could emigrate to that of another computer? Sort
of a distributed computing system. The idea is very simple: If the new
creature in the gene pool have superior qualities, it would "take over" the
population, which is good since then more computers will evolve it. Or maybe
it just have good qualities when cross-over with the current breeds excist.

Framstick could create a set of 2-3 "standard" parameters with different
Objectives, and emigration could then only occur between computers that used
the same parameter-set.

So to the real question:
Hi, I'm looking for a good set of parameters. Ideally I want all aspects
involved = death, food, collission (not water).

With food:
Problems are the same as ever - i.e. the creatures will get so big that they
can just lay there and reach all food balls because of their size. Velocity
is "on" so they have to move some, but...

Velocity:
Too often the gene pool will lose its diversication at once if one creature
picks up velocity, if it just goes round and round.

Also, I can't find the parameters that controlled how much energy the
muscles used. This is an important factor to prevent those huge creatures.

Anyone that has a set of simulator parameters that actually works? Should be
able to start from scracth (X) and run for 1 year.

Hi!
It tried to get a simple Geno that only move a little bit, but i dont get
the part with Organism Control:
[ control input : weight , input : weight , input...]

Conrol is the moving right?
But what exactly is input and weigth?
I really dont get it.
Can anybody give me an good example? Would be very great :)

Or is there a better Tutorial?

What criteria does Framsticks use to decide when to delete genotypes? You
can select random, worst, or inverse-fitness, but I mean how does it decide
/when/ do it? I had originally assumed that it would do so only to make
room for new genotypes, but I quickly discovered that was not the case.

thanks for answers in advance,
Will

I'm still somewhat new to FramSticks, and I'm working on a from-scratch
redesign of my first creature, Bugger. The design goal is to make it a have
6 legs in 3 pairs, and walk like an insect does, i.e. 3 legs at a time so
that there are always 3 on the ground.
This is my current basic structure...
Bugger mk2
(MMMCCrrX[@7:1](MMMX[|7:1]),MMMMMMMMMMX[0:1][@-1:1](MMMCCrrX[@5:1](MMMX[|5:1
]),MMMMMMMMMMX[@-3:1][*:0][*:0][*:0][*:0](MMMCCrrX[@-4:1](MMMX[|-4:1]),,MMMC
CrrX[@-4:-1](MMMX[|-4:1])),MMMMCCrrX[@-8:-1](MMMX[|-8:1])),MMMCCrrX[@-8:-1](
MMMX[|-8:1]))

Each leg has 2 segments. For convenience I think of the upper as the 'leg'
and the lower the 'foot'
Each foot has a bend, for lifting the foot. Each upper-leg has a rotator,
for pivoting the foot forwards/backwards.
Insects walk by moving the front/back legs on each side in unison with the
middle leg on the opposite side, leaving 3 legs on the ground at all times
for stability. Each muscle from each cooperative set of legs reads from a
neuron in the spine. There are 4 of these, 2 for each set of legs (one
rotation, one for bend). I've been testing the critter out by tweaking those
4 neurons manually in the brain viewer. However, no matter what I do I can't
seem to get them to be able to pick up the sets of 3 legs. The spine always
twists so that all the lifted legs are still on the ground, no matter how
much I strengthen the spine's muscles.

Anyone know what I'm doing wrong? I've been fiddling and tweaking for many
hours, and finally realised I was repeating myself and wasn't coming up with
any new ideas for how to fix this. : (
Thanks in advance...

Will