Author Topic: Maximizing numbers in EvoSim  (Read 3379 times)

Offline rwill128

  • Bot Builder
  • **
  • Posts: 67
    • View Profile
Maximizing numbers in EvoSim
« on: December 20, 2013, 12:51:59 PM »
What are the largest numbers most of you guys can achieve in your sims?

I used to keep the size relatively small because it makes for more interesting interactions, and my sims would run great on the computer I'm currently using (an i5 2500K at 3.3Ghz -- my work computer), but only if I kept the population somewhere between a few hundred and a thousand creatures (including vegs). If I let the numbers get higher than that the sim would slow to a crawl and get less than .1 - .2 frames per second.

But I was reading about a grid optimization that was implemented a while back and I realized it might be beneficial to make a huge environment and let the creatures spread out. I was able to get 5,000 veggies running in the same sime with 3,900 animal minimalis at about 5-6 FPS.  Big improvement!

So I tried doing the same thing in 8 sims at once to take advantage of the quad cores, and I was able to get 3-4 FPS on each. So almost 4,000 * 8 is about 30,000 animal minimalis and about 50,000 veggies all running at the same time at about 3-4 FPS. I'm hoping to get a new computer for home use in just a couple weeks, so imagine the possibilities then!

I wanted to point this out because with so many creatures running at once, it might be possible to get much more interesting results in an evosim. Anybody have suggestions on what to use / where to start?

Offline Botsareus

  • Society makes it all backwards - there is a good reason for that
  • Bot God
  • *****
  • Posts: 4483
    • View Profile
Re: Maximizing numbers in EvoSim
« Reply #1 on: December 20, 2013, 04:47:06 PM »
Very interesting; Reminds me of the days when Internet Mode was actually working.

You will be however surprised that you don't need ultra large simulations to achieve results. Then again my methods are very unconventional, I like to throw the book out the window sometimes and go with instinct.

That being said, you should check out http://forum.darwinbots.com/index.php/topic,6220.msg1388348.html#new



The reason I am physically limiting the simulation size with shapes is to have more selective pressure.
Also, I select "disable asexual reproduction for non repopulating species" to force the robots to reproduce sexualy.



Anyway I am going to try to take a mini vacation away from the forum for the remainder of the year. Happy New Year and Marry Christmas to Everyone!
« Last Edit: December 20, 2013, 05:06:27 PM by Botsareus »

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
Re: Maximizing numbers in EvoSim
« Reply #2 on: December 20, 2013, 09:06:50 PM »
Actually I'm surprised you managed to get as high as 3-4 FPS with those sorts of numbers, even spread out.  You might be the record holder :)

Connecting sims together locally on a single machine is probably the best way to go to really grab modern computing power.  I don't know of anyone that's tried more than two or three sims together for any length of time.  It would be interesting to see a long term experiment done with that many connected instances.

Offline Shadowgod2

  • Bot Destroyer
  • ***
  • Posts: 387
    • View Profile
Re: Maximizing numbers in EvoSim
« Reply #3 on: December 20, 2013, 11:59:07 PM »
i would like to see if anyone can get a sim going with 2 different windows at least with the costs the same one would be an underwater sim and the other would be a land sim. there would be shapes in such a way that they would simulate real life evolution in a since, form water to land to see if anything could come up from it

Offline rwill128

  • Bot Builder
  • **
  • Posts: 67
    • View Profile
Re: Maximizing numbers in EvoSim
« Reply #4 on: December 21, 2013, 02:31:14 PM »
That's a cool idea Shadowgod2, I'll definitely think about implementing that.

And to Numsgil, that's definitely what I'm planning on doing. I left it on last night, and I might drop by the office today to check it. I'll definitely check it on Monday and then leaving it running until Friday over my Christmas break.

Each sim (out of 8) should be near a million cycles by the time I get there today. I know that's not a lot, but last night I was browsing around and I've already found veggies that seem to be running away from animals, and I've sen a couple interesting mutations in animals as well.

What's more is that the average number of mutations is at about 5 (and only slowly increasing, which is a good sign) and (this is a first for me) the average DNA length of both animal minimalis and alga minamalis is actually going up slightly.

One of the really cool things is that in a sim of this size, and with this ratio of animals to vegs, there's actually a chance for vegs to escape from close encounters with animals, and I'm actually seeing evolution on both sides.

Offline Numsgil

  • Administrator
  • Bot God
  • *****
  • Posts: 7742
    • View Profile
Re: Maximizing numbers in EvoSim
« Reply #5 on: December 21, 2013, 08:29:09 PM »
Yeah, I think that's part of the key to a good sim: you need the veggies to actually survive instead of being repopulated.  One of the best sims I ever ran used Enitor Comesum, a cannibalistic robot that's not very efficient, and stock alga minimalis veggies getting fed per instance instead of per kilobody and with fixed radii.  The veggies when cancerous, as they always do, and were able to form a "foam" of veggies with only 1-2 nrg each and generally reproduce faster than they were eaten.

The result was an actual red queen race between the veggies slowly learning to evade getting eaten or reproduce faster than they could get eaten and the animals trying to survive.  I had to give it up when the veggies learned to wiggle just enough to break the line of sight the animals use to feed.  The animal population crashed too far to form a viable sim.