Darwinbots Forum
Bots and Simulations => Simulation Emporium => Topic started by: Drognan on February 21, 2010, 05:38:19 PM
-
Hi guys,
Don't know if anyone ever checked this, but are we sure that number of zeroes in zerobot DNA speeds up evolution?
To evaluate this, started new simulation with two types of zerobots.
1. type: zerobot with 910 zeroes in genome
2, type: zerobot with 3600 zeroes in genome
Both species are marked as veggies.
Bets anyone?
I'll keep you posted...
-
ok, new event, guy with 3600 zeroes first learn how to feed, two nearest bots got eaten...
No reproduction yet, it's 1,9K cycles...
So it's 1:0 for bots with lots of zeroes... So far theory is correct
Olmost forgot, there is one more category zerobots, one with random numbers instead zeroes so now I inserted 8 bots with 4000 numbers, zeroes will have 1,9 Kcycles advantage, but don't want to start all proces over again
-
My moneys on the random number bots.
-
My moneys on the random number bots.
Are You sure? Guys with zeroes are 1,9Kcycles ahead
-
If they can't reproduce yet, then their dna is just random. The only problem is the zero-bots already have other commands while the random bots are stuck with numbers.
-
Just to mention, didn't know how to make random nubers sequence so I simply wrote sequence from 1-9 and then copy-paste it till I got 1234567891234567891234... hope that this still counts as random numbers zerobot? I posted dna for inspection.
-
You can use this random number generator over here:
http://www.psychicscience.org/random.aspx (http://www.psychicscience.org/random.aspx)
-
Just to mention, didn't know how to make random nubers sequence so I simply wrote sequence from 1-9 and then copy-paste it till I got 1234567891234567891234... hope that this still counts as random numbers zerobot? I posted dna for inspection.
It doesn't, because there is a pattern in the dna it isn't random. Testlunds link should do the trick.
-
I made this. It generates lines of random numbers between 1 and 1,000. It asks you how many lines you want, 40 numbers on a line. I'm going to improve it to do both random numbers and dna commands like store, inc, add, etc. Sorry about its size, the program I made it in doesn't compile very well. Technically it doesn't complile at all, it just tokenizes it.
-
I made this. It generates lines of random numbers between 1 and 1,000. It asks you how many lines you want, 40 numbers on a line. I'm going to improve it to do both random numbers and dna commands like store, inc, add, etc. Sorry about its size, the program I made it in doesn't compile very well. Technically it doesn't complile at all, it just tokenizes it.
Great, thanx for the program :-)
Inserted 7 "zerobots" (can we still call them like that?) with random numbers...
Update on simulation: the shooting guy just lost it's ability for shooting so it's again 0:0
-
I was wondwring when is good time to introduce costs. What is best cost settings to start with?
-
I'm looking into this right now, trying to figure out the best cost settings. Apparently there are some oddities with it.
Just keep it low, below 0 and use Dynamic Costs once the population starts to slow down your sim.
-
By the way I recommend using Ageing Costs together with Dynamic Costs otherwise there will be selective pressure for bots to do nothing.
You can set it to:
Age Cost: 0
Begins upon reaching: 0
Increase by: 0.000000001 (will be displayed as 1E-09)
I did this calculation for movement and rotation costs:
32000 / 24 / 60 /60 / 32 = 0.011574074074074074074074074074074 (1.157407E-02)
Hehe.
-
By the way I recommend using Ageing Costs together with Dynamic Costs otherwise there will be selective pressure for bots to do nothing.
You can set it to:
Age Cost: 0
Begins upon reaching: 0
Increase by: 0.000000001 (will be displayed as 1E-09)
I did this calculation for movement and rotation costs:
32000 / 24 / 60 /60 / 32 = 0.011574074074074074074074074074074 (1.157407E-02)
Hehe.
Wow respect man! How did ou do that???
-
Just some funny calculation I came up with. I use time, beat patterns and bytes as rules for how I set up my sim.
175 zerobots with lowest field setting will run the sim at close to 32 c/s at normal speed setting and if the morphological cost is 1.157407E-02 and they start out with 32000 nrg they will live for about 24 hours if they constantly shoot, turn or swim without gaining energy (at least in theory).
An animation speed around 30 cycles looks most realistic imo.
-
Impressive. This means that after a day of running in the background, if anything is left then those are the ones to save.
I need to start doing that.
-
Yes, it should weed out old boring bots and favor reproduction. Once you have a stable population you can turn off ageing cost. It's a bit artificial. Unicellular organisms never ages in reality (they just keep splitting in two). I pretend it's a tear and wear cost or whatever you call it in English.
-
When is good time to introduce costs? when reproduction starts? or feeding?
-
Well, you could set some low morphological costs for realism, then set the cost multiplier to 0, then turn Dynamic Costs on. Set the target population to a level where your sim runs at a decent speed, then an upper and lower percentage to try and keep the numbers of bots within the target.
I think the first priority is to have bots that reproduce regularly before they do anything else. Use ageing cost to encourage that.
In my personal experience zerobots are just hopeless to get anything from nomatter if you have costs or not, so I've gone over to randombots instead.
Right now I have a stable population of randombots on my laptop. Sometimes they peak in population, but the faster they peak, the faster they drop too.
I think the best results are when the population graph looks like a small sawtooth pattern instead of peaks, drops or flatlines.
-
You are probably right, in my sim zerobots and bots with random numbers are mixed together, but so fan no strain has evolved reproduction, some bots from both sides evolved shooting and then lost that ability... Right now about 6Mcycles
-
OK just had two offspring, in simulation, of course and parents are zorobots with lots of zeroes so it's 2:0:0, guys with random numbers didn't jutsify expectations so far, but it's only begining. Also bot's with only 900 zeroes are behind in evolution from their 3600 zeroes collegaues (as expected)... Bots with only 30 zeroes so far only evolved random moving (156 Mcycles).
The show goes on...
-
Well its not really evolution untill they can reproduce. Any and all behaviour is really just random. So you could get the one with 3 zeros to reproduce first but my guess is the more chances at mutations, the faster they will evolve reproduction, at least statistically.
-
OK, finally some news, we have reproduction going and the winner is zerobot with 900 zeroes! It was a bit surprise for me, because I beleived that first bots to reproduce will be ones with more zeroes or with random nubers...
Is this good time to introduce costs? There is over 400 zerobots with 900 zeroes in sim, other species are stuck below 10. Sim is draging like a dead cow by now, 1,6 cycles per sec.
-
Are you judging "first" based on real world time or number of cycles? 3600 bps will make any sim chug.
-
Is this good time to introduce costs? There is over 400 zerobots with 900 zeroes in sim, other species are stuck below 10. Sim is draging like a dead cow by now, 1,6 cycles per sec.
Wow. At this point, I'd be getting 10 seconds per frame
-
Is this good time to introduce costs? There is over 400 zerobots with 900 zeroes in sim, other species are stuck below 10. Sim is draging like a dead cow by now, 1,6 cycles per sec.
Wow. At this point, I'd be getting 10 seconds per frame
It's a new laptop, dual core , now there are 500 bots, same speed
-
Are you judging "first" based on real world time or number of cycles? 3600 bps will make any sim chug.
real world time, but these zerobots, and zerobots with 3600 zeroes were introduced into the sim in the same time so I guess that it's the same.
Now about those costs.... anyone?
-
You can always try to introduce aging costs and the like. But basically the costs can have unpredictable results at times, so it's less a science and more a fiddle-with-the-knobs-until-you-get-it-right. I'd say morphological costs (costs for moving, shooting, etc.) are better than arbitrary costs like aging.
-
You can always try to introduce aging costs and the like. But basically the costs can have unpredictable results at times, so it's less a science and more a fiddle-with-the-knobs-until-you-get-it-right. I'd say morphological costs (costs for moving, shooting, etc.) are better than arbitrary costs like aging.
I usually start by adding ageing cost first. This puts a premium on shorter reproduction cycles and speeds up evo. Usually when introducing other cost I relax ageing cost temporarily.
-
OK, contest is over, zerobots with 900 zeroes started shooting and reproducing and devastated all other species. So, I guess more zeroes or random numbers doesn't always guarantee survival. Oh, btw, zerobots with 30 zeroes never developed even reproduction.