Setings for making a set body (structure unchanged) walk? 
Submitted by framkenstein on Sun, 2014-03-02 03:53

Hi, I have designed a creature structure and muscles which I would like to make walk.I have played about with parameters but constantly run into issues, mainly very slow progression and the simulations seem to stop changing enough (or do I need to be ultra patient,400mil steps so far?). I have managed to set it so the body does not change but I think Iam doing something wrong with the other param's. If anyone could advise on the correct set up for my needs? Any help much appreciated!

I wanted to design an insect like creature, make it walk and then, when its mobile, teach it how to find food and finally add reproductive evolution when it can sustain itself in the world.
This is it at its current state :

((lllllX[N,in:0.921548,16:-0.195,8:0.233,8:1,6:0.148,9:1,si:2]lllllX[@,9:-6.222,p:1,p:1]LLFX[@, 24:-4.889,p:0.935][|,18:1.126], lllllllllllllllllllllllllllllllllllX[N,s:-0.063,2:1,0:0.919,0:1,24:1.803]llllllllllllllllllllllllllllllllllllllllllllllllllllllllllllX[|,29:1][S][@](llllllllIX[T], llllllllIX[T, r:0.999,r:0.802]), lllllX[N,16:4.454,-10:2.195,si:0.844331,24:1.676,-6:1,si:2]lllllX[|,-2:1][@,23:0.674]LLFX[@, -4:8.161, p:0.846][|,21:0.391,r:0.957,r:1], , ), X[N,6:3.564,13:-0.981,6:1,6:0.334](llllllllX[N,-7:-3.658,-1:-0.157,s:-0.086,si:3.60048,10:-3.828,-12:0.195,0:1,-8:1]lllllX[|,18:2.084][@,17:0.377]LLFX[@,-4:1][|,-12:23.737,p:0.617,r:0.986], llllllllX[N,-15:1,-12:1.729,7:0.287,0:1,-21:1]lllllX[|,-14:3.355][@,-14:10.633,p:1]LLFX[@, p:0.59,p:0.956,-3:0.479,p:1][|,-16:35.508,p:0.25], X[G]llX[N,0:4.698,7:-1.007,-11:2.88,si:-1.66946,-12:2.271,-6:-0.987,-12:1](llllllllllX[N,0:6.588,s:0.14,-22:-1.805,-13:0.124,6:1,-12:1.438,-12:1], , , , , , , , , , llllllllllX[N,fo:0.00681694,-29:0.598,-3:1.438,-23:1.022]), lllllXlllllX[|,-26:-3.13,p:0.622,r:1][@,-5:-3.02]LLFX[@][|,p:0.047,-17:25.789,p:0.25,p:0.215], llllllllX[N,-13:9.839,-8:6.289,s:-0.316,-8:2.153,-13:1,-34:-2.623,si:999,1:1]lllllX[*][|,-10:14.975][@,p:0.998,-21:-0.356]LLFX[@,p:0.963,-22:1,p:1][|,r:0.776,p:0.331]))

I initially coded the stick and rotational muscles then had it evolve neurons and connections (because I am still unfamiliar with how they work and need to be set out) could this be causing the problems?
Would it be better if the neurons were placed in any specific pattern?
Is it possible to have one neuron controlling all joints?

Forums: 
Users
Maciej Komosinski's picture

Submitted by Maciej Komosinski on Sat, 2014-03-08 04:10

Here are some suggestions:

  • I would recommend using the most recent version from http://www.framsticks.com/apps-devel
  • When speaking about evolutionary time, it is better to mention the number of evaluated creatures, not the number of simulation steps.
  • When designing evolutionary process, you first have to think about the size of the gene pool and the evolutionary pressure (the negative and the positive selection). These are the main forces that drive progress. Of course, the fitness function must be properly defined.

Now your questions...

I initially coded the stick and rotational muscles then had it evolve neurons and connections (because I am still unfamiliar with how they work and need to be set out). Could this be causing the problems?

This should not cause any problems.

Would it be better if the neurons were placed in any specific pattern?

The neurons that are strongly linked should be located nearby in the genetic code, but this is not a must, it will only decrease chances of crossing over destroying neural "modules".

Is it possible to have one neuron controlling all joints?

Yes, you just have to connect many muscles to a single output.