Author Topic: Suggestion for randomness  (Read 3254 times)

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Suggestion for randomness
« on: August 02, 2011, 03:41:06 AM »
I watched a dokumentary on TV yesterday about chaos and order. They talked about nature generating randomness spontainlessly without any external input, caused by fractal geometry.

So I was wondering if this could be used in Darwinbots. When starting a sim the program could generate a fractal and then make a seed for the simulation from a snapshot of that fractal.

What do you think?
The internet is corrupt and controlled by criminally minded people.

Offline Botsareus

  • Society makes it all backwards - there is a good reason for that
  • Bot God
  • *****
  • Posts: 4483
    • View Profile
Re: Suggestion for randomness
« Reply #1 on: August 14, 2011, 01:04:44 PM »
vb already does this. Instead of fractals it uses "Timer" which is a long containing the system time.

What we need is to be able to set our own seeds or use the timer seed. And we need to be able to view the seed set by the timer for debug.

I already added all this mods to the evo system I am  working on for the past year...

P.S.

For some really crazy ways to improve randomization at the expense of consisitancy see the following post:

http://forum.darwinbots.com/index.php/topic,108.15.html

Quote
'COMBINE ONLY USING ATN() FUNCTION , I use getang as a custom atn using function

result = rnd(-(getang( rnd(-yourownrandomnumber) -0.5  , rnd(-thelist(t)) -0.5 )/360)
« Last Edit: August 14, 2011, 01:22:09 PM by Botsareus »

Offline Testlund

  • Bot God
  • *****
  • Posts: 1574
    • View Profile
Re: Suggestion for randomness
« Reply #2 on: August 15, 2011, 11:14:25 AM »
vb already does this. Instead of fractals it uses "Timer" which is a long containing the system time.

Not so well in the versions I've been running. From what I've seen so far the bots always evolve to do the same things depending on what they started with. I don't think I've ever seen a zerobot evolve to start shooting when it sees something or swimming close to something it sees for instance, or to try to feed from a tie.

They always evolve to shooting at intervals for no reason for a short time then stops, or they will shoot constantly.

The program seems to favor some mutations over others. The inc command appears too often while others never appears at all.


Quote
I already added all this mods to the evo system I am  working on for the past year...

If you have a version where you have improved randomness I would very much like to test it.  :)
The internet is corrupt and controlled by criminally minded people.

Offline Botsareus

  • Society makes it all backwards - there is a good reason for that
  • Bot God
  • *****
  • Posts: 4483
    • View Profile
Re: Suggestion for randomness
« Reply #3 on: August 15, 2011, 02:01:48 PM »
I made atleast 20 mods to the program and am waiting for the release of 2.46 so I won't have to recode all my mods twice... I made significant progress, but I also am not really ready to release because I may run into more bugs ( I found 5 bugs that are not related to my mods )

However, I can release  my work into a separate folder as is. This will not effect the main code but I need permission from Numsgil, a password , and a way to do that... Numsgil, you can pm me a password and a way to create a new side folder (was it called a branch?), thank you...

(I think I had a password for a user "bostareus" a long time ago, I think I lost that one)

err... on second thought, are you sure you want it? I gutted the heck out of it so it only does formula1 mode evolution for virsion 2.44.1. You also must have windows7 with spesfic resolution and settings for it to do generations properly (I am using external controlers that emulate mouse clicks) So, anyway you turn it you must know programming in order to extract anything usefull out of my code...
« Last Edit: August 15, 2011, 02:35:52 PM by Botsareus »