hi & request

[Fae]

hihi! im new! (no. really?)

anyway, ive been trying to run a evo sim starting with anima minimalis, but the animini keeps dying out, thus, im either extraordinarily thick, or im missing something, so I was wondering if I could maybe get a 'mentor' of sorts who could guide me to creating a harsh, but survivable simulation, centered on ani-mini, ideally through MSN

you can PM me your addy if you like ^^, or failing that, some suggestions for settings

[Moonfisher]

Hi and welcome

I don't have MSN... I use googletalk... not a big fan of MSN.
I attached a settings file for a sim with Alga_Minimalis and Animal_Minimalis. (Also attacked the 2 bots used, if the settings file won't load try adding the bots to your Robots folder aswell)
All you need to do is load the settings file and it should have some descent settings for evolving this bot.

But I figured you may still be interested in knowing whats different in these settings :

Since it's a fairly short bot it wasn't enough to set 32X global mutations IMO, so I increased the odds of certain mutation types for Animal_Minimalis (I increased point mutations, copy errors and insertions I think, by having a lower number in the field). (Global mutations are set to 1 now, so it should also be easier to regulate mutations rates as you please)

I think the settings are standard F1 settings with some modifications, to get standard F1 settings you can run a league or a match between 2 bots and all the settings should be set to F1 defaults.
Changed the amount of alge and added increasing age costs at 10K cycles, to balance population around 200-300 and push for some quick evolution.
Also set it to restart the sim if all bots die, this way if the first few bots die out it'll just give it another try till they succeed. (They should manage though, unless you're unlucky or something)
And the alge only start with 500 nrg, to create a high demand for evolving -6 shots.

I should also mention I'm running version 2.43.1L, it's the most stable that I know of so far.

I'll leave the sim running for a litle while and post if anything interesting happens, or if I realize something isn't going as planned...

[ikke]

To give some contradictory advice:
1) lower the mutation rate. This will make the effect of selection stronger and you will have evolution and not devolution.
2) increase the length of insertions. The current default settings are favouring deletions. Setting the insertion length to 1-3-5 makes it more even
3) Duplicate the bots dna. This will give your bot  basepairs to evolve. Alternatively, insert zero code between animalis genes.

[Fae]

okie dokie, undatey

after a few sims that failed (ironically, im having the opposite problem with them since I started using moon fishers settings... it stablises for a hour or so... then EXPLODES suddenly, adding hundreds more in a few mins and making the sim unworkable D:) I finally achieved stability, and have been slowly increasing the costs and the rate of aging, my hope, is that a high cost/low food environment will help filer out the junk-mutations from my high mutation rates, and encourage competitive evolution.

(and, hopefully, give rise to cannbals eventually)

starting with ani-mini and alga-mini, using moons settings as a base:
 
I have the rate set to 1x, but the chances for point, copy err and insertion are 1 in 3000, rather that 1 in 5000, I used ikkes suggestions for the 1-3-5 rates on insertion, and set major deletions to 1 in 10000.

vegs come with only 500 starting energy, with a max pop of 5, repop at 1, 2 per repop and 10 cyc delay

costs are I believe, F1 default with the following mods: rotation, voluntary movement = 2, shot and tie = 2, dna upkeep = 3, age cost begins at 1500 cycs old, starts at 1, and increases by 1 per cycle

thus far I achieved (what I felt was) a meaningful evolutionary breakthrough in the form of continuous, meaningful, voluntary movement. I saved that bot as Ani-Mini Modius, and restarted the sim with with it as the base.

I am currently around 60k cycles in, with three derived species of ani-mini modius, one doing well, one doing not so well.


current the sim is a swarm of around 300 small, low energy bots, I was wondering if there was anything I could do to promote the formation of fewer high-energy bots and/or cannibals?

[ikke]

The phenomenon you saw was a mutation in the condition of the reproductive gene resulting in a reproduce every turn command. If you want these bots to die you'll have to check the veggiy energy from per cycle to kilobody point or quadratically. Energy per veggie breeds small cancerous veggies, the others big veggies with possibly reproduction based on body.

Another suggestion is to check if your mutations are not set too high. There is an option to place bots in the sim. If you reintroduce the original bots they should not be able to outcompete the evolved versions.

[Fae]

I got two variations out of my sim, dubbed Modius II-I and II-II, I started a new no-mutations sim with 5 Animal Minimalis, Ani-Mini Modius, Modius II-I and Modius II-II to compare their fitness

the animal minimalis died out very rapidly, while the original Ani-Mini Modius held on for a few k cycles before dying, II-II seems to be doing better than II-I (which is odd, since II-I was dominating the original evo simulation, even surpassed Ani-Mini Modius in population)

*is pleased ^__^*

*edit*
if yous dont mind, could one of you explain in abit more detail what the different veggy energy types mean? per cycle, kilobody point, quadradic, etc

[Shasta]

if yous dont mind, could one of you explain in abit more detail what the different veggy energy types mean? per cycle, kilobody point, quadradic, etc

Per Cycle: Gives X amount of energy to each veggy
Per Kilobody point: Gives X energy per 1000 points of body the bot has (2000 body gives 2*X)
Quadratically: If the veggy has 32000 body (Max ammount) it will receive ~60 times the amount a bot at 1000 body would recive

[Fae]

if yous dont mind, could one of you explain in abit more detail what the different veggy energy types mean? per cycle, kilobody point, quadradic, etc

Per Cycle: Gives X amount of energy to each veggy
Per Kilobody point: Gives X energy per 1000 points of body the bot has (2000 body gives 2*X)
Quadratically: If the veggy has 32000 body (Max ammount) it will receive ~60 times the amount a bot at 1000 body would recive

kilobody seems to defeat the point if it stacks :S

(from what ive seen, kilobody tend to swell up hugely and never ever die when assaulted by a swarm of small bots)

[Numsgil]

Think of it in terms of how the sim is pressuring the veggies to evolve:

per cycle favors lots and lots of small bots.  Because if a veggie doesn't reproduce, it gets X nrg, but if it does, its two children get 2X nrg.

per kilobody is agnostic.  If a veggie doesn't reproduce, it gets X nrg.  If it does, its two children still get just X nrg.  This is probably the most realistic model compared to real life, since you can pretend that each body point a veggie has represents leaf surface area or something like that.  The bigger the veggie, the more surface area it can use for photosynthesis.  So in theory it shouldn't provide any fitness cost/benefit to cancerous veggies vs. big berthas that never reproduce.

quadratically based on body is a little more complicated.  Basically it's an arbitrary function which gives extra nrg to bots which are extra big.  A veggy with 30x the body of a normal bot will get about 2X the nrg that 30 average sized veggies would get.  So there's a fitness cost to reproduce, so in theory it should provide a counterbalancing force against veggies cancerously reproducing.  In practice I don't know if it really works like that, but there haven't been many sims run with this option.

It's all a little confusing for a new user, so if you don't take anything else away from this, just remember to set the feeding method to per kilobody by default.

[Fae]

oi oi! I have a request! (and no, im still here, still fiddling with darwin)

up till now I have been running evosims, and im quite happy with the results ive been getting. but.

ive gotten a itch to start making my own. and am planning to take my evolved bot and improve it. ive given it blinding poision, waste disposal and a return-fire gene, but my main goal is to make it into a multi bot.

my understanding of the dna is still foggy, so I was wondering if I might put in a request:

I need a (very well documented) gene (or, much more likely; several) that would:
-spawn children to the four directions of the parent (one above, below, behind and infront of. or something to that effect)
-perma-tie a child with a non-flexible tie (hard tie?) to keep them in place
-cause the tie to lengthen as the parents energy increases

the objective is to make a ever-expanding multi-bot

if anyone takes up this challenge, ill be very thankful ^.^

*edit*
on further research ive found the code to stiffen a tie
I need to know what to put into "1 .fixlen store" to make it depend on the parents energy

[bacillus]

I've done quite a bit of work on Multibots, so I might be able to help you. To get the idea, I think it's easiest to start by making a worm. Put simply, this is the sequence:

• Create a tie on birth. This should be done at robage = 0, so the tie is formed before the cells can move away.
• If the cell was single before reproduction, become a tail. If it was a head, become a body. The new cell starts off a head.
• Use something like *.in1 ++ .out1 store on birth to create a cell 'hierarchy'. This way, a cell with one connection knows if it should be a head or a tail, depending on whether its value is smaller or larger. This is critical for when a bot takes a chunk out the middle.
• Remember to store 'head' 'tail' and 'body' in a location like 991 or something. This way, by checking which of the three values the cell is, you can set the behavior accordingly.

[bacillus]

Sorry about the double post, my keyboard's playing havoc and I need to go through an annoying sequence to stop it deleting everything I wrote.  
Anyway, I actually read your post now and realize you want to create a branching bot. I made a fairly nice one a while back; try looking up hydra or emeris in the bestiary.

Btw. the sequence involves taking the keyboard apart and spraying lens cleaner on the contacts. Sounds stupid but it works.