Framsticks

For an experiment I am running, I need a creature that is capable of moving
in more than one direction, where its direction is (indirectly) controlled
by the makeup of its neural network. I tried adapting the 'basic quadruped'
creature to make it have two 'heads', but this failed miserably. I also
tried evolving the 'bugsy' creature with the fitness criterium being the
direction moved in, but this also did not work.

Does anyone have any idea how to make such a creature? Ideally, the creature
would look something like the basic quadruped. This is my adaptation of the
basic quadruped (but it is probable I did something wrong):

MX[*][N,-1:0.592,1:2.45,2:3.279][N,-1:3.628,0:1.488,0:-2.022][N,-1:-4.188,0:2.763][@,-3:-2.231](RX(FX,
), MX[@,-3:-2.495][|,-3:2.468](rX(, X),
X[@,-5:2.495][|,-5:-2.468](X[@,-8:2.231]), RX(X, )), rX(, FX))

Thanks,
Eelke Spaak

I can't understand the reason of the swimming creatures are very similar to
the real ones and the terrestrial creatures are very different.

Any one can explain me?

Sorry, and Thanks,

Daniel Silva

Hi all,

Save everything below the --- in a file called 'framscript.syn' and you can
use it to enable syntax highlighting for Framscript in the TextPad text
editor.

Best regards,

Eelke Spaak
Nijmegen, NL

----

C=1
[Syntax]
Namespace1 = 6
IgnoreCase = No
KeyWordLength =
BracketChars = {[()]}~
OperatorChars = -+*/<>!%^&|=
SyntaxStart =
SyntaxEnd =
HexPrefix = 0x
CommentStart = /*
CommentEnd = */
CommentStartAlt =
CommentEndAlt =
SingleComment = //
SingleCommentCol =
SingleCommentAlt =
SingleCommentColAlt =
SingleCommentEsc =
StringsSpanLines = Yes
StringStart = "
StringEnd = "
StringAlt =
StringEsc = \
CharStart = '
CharEnd = '
CharEsc = \

[Keywords 1]
asm
break
class
case
continue
do
else
false
for
function
goto
if
new
null
return
switch
this
true
try
var
while
@include
\#include
label_
global

[Keywords 2]
Collision
Creature
CreaturesGroup
Dictionary
File
NeuroProperties
GenMan
GenManStats
Geno
Genotype
GenotypeGroup
GenotypeLibrary
Interface
Joint
LiveLibrary
Loader
Math
MechJoint
MechPart
Model
Neuro
NeuroClass
NeuroClassLibrary
NeuroDef
Part
sim_params
Simulator
stats
String
ShowProperties
Vector
World
WorldMap
ExpProperties
ExpState
Neuro
this
Creature
Element
GeomBuilder
GL
Loader
Material
MechJoint
MechPart
Scene
VertexBuilder
VisProperties
VisualModel
World
WorldMap
ClassBrowser
CLI
CreatureCharts
CreatureSelection
DynaChart
GLDisplay
NeuroDiagram
RowFormat
ShowManager
SimStatsCharts
TrackingCam
VisualStyle
Window

[Keywords 3]
script
expdef
show
class
prop
state

[Keywords 4]
name
info
code
id
type
flags
group
help
longname
prefinputs
prefoutput

[Keywords 5]
onExpDefLoad
onExpInit
onBorn
onStep
onKill
onExpLoad
onExpSave
onUserEvent
onLoad
onResize
onShowStep
onSelectionChange

I'm unclear on what the selection rules mean;
1. I understand random
2. Fitness proportional: Does this mean genotypes with the best fitness are
selected for mutation/crossover?
3. Tournaments: Don't understand this.

Thanks,

Will

I'd like guidance. I want the real fitness of a creature to be equal to the
Lifespan of its best offspring.

(I have experience making special fitness forumulas that are more involved
than the simple slider bar methodology. I don't know how to access
information about other genotypes in the gene pool though.)

William Sharkey

I'm a philosopher, not a programmer. I've been trying, in vain, to figure
out a way to get more than one gene pool into my simulations, so that I can
create a larger world with more complexities and multiple life forms, all
evolving and co-evolving together.

But all the resources for Framsticks seem to be disappearing rather quickly,
and I cant find much help other than this small forum.

Can anyone help me, or point me in the right, still active, direction?

Best,
ns

Hello,

My neuron has several fields, this works fine, however, problems occured, when I wanted to access (only read) them from onStep(). All the props have flags 0, that is fine. Then I'm recording the fields in every time step, for every creature and every neuron, in the onStep() function, like this (instead of saving I write it out):

Dears,
in Standard-like expdefs, where Genotypes have more instances (popsiz), do the GenotypeLibrary functions like roulette(), revroulette(),
tournament() or worst() work on the instances or on genotypes? For instance do all genotype instances have their slots on the roulette wheel,
according to their fitness, or do only genotypes have slots? Similarly, during tournaments, unique genotypes compete, or instances
(possibly all from one genotype?)? Thank you, Matej Hoffmann

Hello,

I am working on a project in framsticks and want to develop learning
creatures. The problem with this is that in any type of learning, especially
hebbian, there has to be an interaction between neurons to change the
weights. So my question is: is it possible to get a reference to an input
neuron AND change the weights from that neuron to your own. I tried looking
a the NeuroProperties object, but it has no documentation.

Regards,

Tom Schut

Hi,

I evolved a creature using the standard settings, evolving it for velocity.
I wanted to remove the "Junk" from the genome, so I wrote a python script. I
removed all the parts which I could without lowering the fitness. I used the
cli interface in order to determine the values.

I got something like this.
(, lFX[N][N][N]( lrLLLLX[N]LrLrrLLLrLLrrLLrrLLLLLLLLLLLX[|, 3:10, p:0.489]),
LLcXX[N][G][in:0.741074, -1:-1.737])
This entry gives the same fitness value as my original creature. However, I
see several Nuerons which were not being used. In particular it has three
Neurons in a row [N][N][N] in the beginning. It seems to me tha tI should be
able to remove them. There existence is not used and so it really shouldn't
effect the velocity/fitness of the whole. However, I get a massive drop in
velocity/fitness when I remove any of them.

Am I missing something? What is causing this?

Winston Ewert