Darwinbots Forum
Bots and Simulations => Evolution and Internet Sharing Sims => Topic started by: Drognan on October 25, 2009, 08:33:42 AM
-
Started new zerbot simulation with "voluntary movement" cost set to 1 and "bang efficiency" set to 10% from the start.
The goal is to see new movement methods develop, using ties etc... So far 1,5M cycles, nothing interesting is happening...
-
OK, 3,5M cycles, still no reproduction, but some bots started to rotate
-
4M cycles, still no reproduction.
Maybe I'm doing something wrong. How do I know when bots are capable of reproduction?
-
4M cycles, still no reproduction.
Maybe I'm doing something wrong. How do I know when bots are capable of reproduction?
euh an increase in population??
If you have no reproduction the only mutation you have are pointmutations, make sure they are enabled. After that it is just waiting till some store hits .repro
-
4M cycles, still no reproduction.
Maybe I'm doing something wrong. How do I know when bots are capable of reproduction?
euh an increase in population??
If you have no reproduction the only mutation you have are pointmutations, make sure they are enabled. After that it is just waiting till some store hits .repro
Veggie is feeder/repro, as is described on wiki. Population oscilates between 90-150 due to dynamic costs enabled. Should I remove that veggie and introduce feeder only veggie?
Antoher question... How do I know when some store hits .repro? (Don't know about programing, I'm medical doctor, sorry... )
-
Removed feeder/repro veggie, introduced feeder only veggie and population still rising!!!!!!!! Yes!!!!
-
4M cycles, still no reproduction.
Maybe I'm doing something wrong. How do I know when bots are capable of reproduction?
It depends exactly how you've set up the settings. For instance, if you have it set so that new zerobots get spawned if the population falls below some threshold, then it's sort of difficult. You could look for the best bot and see how many children it has produced. If the zerobots don't get auto-reseeded, you could just look at the count of the total bots, and see if it ever rises.
-
OK, new approach
30-50 zerobots that survived so far with feeding only veggie and max mutation. I'll wait till population graph start rising. Then I'll lower mutation rates and wait till they learn to feed. BTW how can I see if they started to feed on their own? Then I'll swap feeder veggie with alga minimalis and wait till movement starts.
Settings are: Bang efficiency set to 10, movement cost 1, zero costs when population fall till 40, no respawn, no dynamic costs.
Thank You for answers so far
-
There isn't an easy way to check if the bots have started feeding beyond just looking for shots or ties. Once you get a basic bot that can eat and reproduce and move, it becomes difficult to exactly see what new behaviors they start to develop. But you can use the graphs to record things like population and average energy, and see if they increase/decrease over time.
-
Zerobot evosims are a total pain in the ass. Not only do you have to micromanage them for a few days, you have to run them for a few days to begin with. And THEN you have to keep track of autosaves, check for improvements every time something interesting happens, ect.
It's a painful balancing act. I never really got anything interesting to develop. What's more, I always ran mine on an old rig with an outdated P4 processor.... so the sims never really went over 15M cycles.
-
Zerobot evosims are a total pain in the ass. Not only do you have to micromanage them for a few days, you have to run them for a few days to begin with. And THEN you have to keep track of autosaves, check for improvements every time something interesting happens, ect.
It's a painful balancing act. I never really got anything interesting to develop. What's more, I always ran mine on an old rig with an outdated P4 processor.... so the sims never really went over 15M cycles.
You are right, lost patience, at 10M cycles no repro. Probably dead end
Restarted the whole thing and started again with same settings.... At the moment 7,5M cycles, no repro
I'll wait and see
-
I usually DID get reproduction EVENTUALLY, just no useful behaviors ever really developed. It always seemed like they would evolve to drift in a certain direction and shoot constantly.
-
When you're running the very start of a zerobot sim, it's useful to talk about bp-cycles.
That is, calculate:
Number of cycles * number of bots * number of 0s per bot.
A bot with 1000 0s running for 1 Million cycles has roughly the same chance of reproducing as 1000 bots with 10 0s running for 100K cycles. The cheapest way to make your bp-cycles larger is adding more 0s to your zero bots.
-
Passed 7,5Mcycles, no repro!
When you're running the very start of a zerobot sim, it's useful to talk about bp-cycles.
That is, calculate:
Number of cycles * number of bots * number of 0s per bot.
A bot with 1000 0s running for 1 Million cycles has roughly the same chance of reproducing as 1000 bots with 10 0s running for 100K cycles. The cheapest way to make your bp-cycles larger is adding more 0s to your zero bots.
Thank You... I'll insert some new bots with more 0s...
-
You might also want to check that the bots are indeed mutating. After a while (a few million cycles), double click on one and examine the DNA. If it's still all 0s, then the mutations are disabled somehow...
-
Yeah. I can't really remember how many zeroes I had in them, but I usually ran around 100 bots for the seeding rounds, then I think around 50 after that. I can't remember. But it takes an assload of patience and luck to get anything out of them.
-
I used 100 bots with 1000 BP of random numbers (excel generated) to start an evo sim. Reproduction occurred after 1 M cycles. I started with veggies, no ageing cost. Added ageing cost when the pop had doubled (mass extinction of course). I started a new sim with the evolved bots as predators when they developed feeding (shooting IIRC).
|EDIT: looked up the thread: memory is definitly faulty:
evo sim thread (http://www.darwinbots.com/Forum/index.php?showtopic=2857)
this post may also give some info w/r evo sims:
my first evo sim (http://www.darwinbots.com/Forum/index.php?showtopic=2819)
-
Inserted some 1000bp zerobots, enabled dynamic costs, target pop. around 140, and now will wait.... Eh, now It's 23M cycles, no repro...
-
save the sim and post it. I'd like to check the settings.
-
There is problem with uploading sim (web page not accesible), I'll try in few days
-
You mean you can't upload things on posts? Is it a large sim? Have you tried zipping it?
-
OK , posted two sims. First I ran, but lost patience and started another. Same settings, but no reproduction... I don't beleive that there was an evolutionary dead end in oth simulations...
-
I am unable to unzip these files with winrar. What program have you used?
-
winzip
-
OK started new sim, animal minimalis and alga minimalis. I'll gradually decrease bang efficiency. Maybe it's better starting point...
Thank you guys for all help... I'll keep you informed about progress
-
You mean you can't upload things on posts? Is it a large sim? Have you tried zipping it?
OK here are simulations. Compressed in Winrar.
-
LOL, I remember when winzip was the norm and noone knew what winrar was
-
arj anyone?
On topic: No obvious (to me) errors. Lot if settings I never used though. Nice trick to use the transporter to keep movement in the sim. Point mutations are set high. You'll want to bring that down once something is going, but to clear the zeroes it is OK. I never use oscillation for mutation rates, but at this stage it doesn't matter. There is a shooter you may want to re seed in a new sim, but you may consider that cheating. No clear reason why repro cannot evolve.
Warning: increase in ageing cost is way too high. If something ever gets going you'll probably wipe out all life in a heartbeat.
The sim I recently posted had 100 1000 BP zerobots. Once point mutations transformed the 0's into random I brought the point mutation rate to it's current setting (default x10) and set the mutation rate to 1/32. repro within 2M cycles.
Sorry I can't be of more help
-
Thank You for suggestions... Oscilations for mutation rates is more biological. If you look at our own past, there are periods of stable ecosystems that lasts for thousand (milions) of years and then happens some radiation, polution, or natural disaster that causes lots of mutations. Then again is period of stability that enables natural selection of newformed biodiversity. That cycle repeats on and on and will probably repeat again... With short period of 16x mutation rate I get lots of mutations and then with long period of 0.16x mutation rates I get natural selection.
However I gave up of my zerobots, now is new simulation in progress, animal minimalis and alga minimalis. I gradually decrease movement efficiency, right now bang efficiency is around 10%, zero momentum on. I'll post sim when bang efficiency drops to zero
-
Thank You for suggestions... Oscilations for mutation rates is more biological. If you look at our own past, there are periods of stable ecosystems that lasts for thousand (milions) of years and then happens some radiation, polution, or natural disaster that causes lots of mutations. Then again is period of stability that enables natural selection of newformed biodiversity. That cycle repeats on and on and will probably repeat again... With short period of 16x mutation rate I get lots of mutations and then with long period of 0.16x mutation rates I get natural selection.
There was some discussion about this sort of thing years ago. I had your reasoning, and I think shvarz favored constant low level mutations. I'm not sure if there's any strong evidence either way. Which one you favor probably goes more to which version of evolution you hold internally (stochastic vs. gradual).
However I gave up of my zerobots, now is new simulation in progress, animal minimalis and alga minimalis. I gradually decrease movement efficiency, right now bang efficiency is around 10%, zero momentum on. I'll post sim when bang efficiency drops to zero
You might also try something like enitor comesum (see: my old old sim (http://www.darwinbots.com/Forum/index.php?showtopic=525&hl=enitor+comesum)).
-
I completely forgot all about arj
I used to hate it (Had no idea what all those flags did, og even what a flag was, back then)
(I had some bat file for using it, with some long line that made no sence to me, just changed the names of source and destination files)