Welcome To Darwinbots > Newbie

Hiya and wiki

<< < (3/5) > >>

EricL:

--- Quote from: Junior ---Oh, they do, but it takes some time and slows down the entire simulation until they're dead. Can I do something to stop this from happening or make them die as soon as it does?
--- End quote ---
Don't start from hand-authorred bots.  Start from a population of zero bots or at a minumum, a hand-authorred bot that is already cancerous.  

Cancerous bots (bots that attempt to reproduce every cycle) are the norm in all long running evo sims but you usually only notice them early in an evo sim that begins with an artifical starting point I.e. hand authorred bots instead of zerobots.  Hand authorred bots with hand-authorred conditional reproduction genes represent a lot of artifically stored nrg (body actually, since that is the key resource necessary for reproduction) that can suddenly be released by a mutation like a damn bursting and trigger a flurry of explosive reproduction.  Like a damn, this hand-authorred code is totally artificial and ineffecient from selection's perspective.   Not reproducing when you could gets selected against eventually.   Naturally evolved bots will not store nrg needlessly waiting to reproduce.  They will reproduce whenever they can (whenever they aquire enough body that the simulator allows them to).

Given the right environmental conditions such as preditor - prey co-evolution, we might eventually see selection favor larger bodies at the expense of more frequent reproduction.  One can imagine conditions where having larger offspring is critical to survival and thus the storing of nrg (as well as the conversion of nrg into body) before reproducing might be heavily selected for in such conditions but that hasn't been seen yet, at least not outside of internet mode.  It is possible the competition between different species in IM demands larger bots with more nrg reserved from combat and thus serves to provide enough selection pressure to preserve conditional reproduction logic, but this is largly conjecture on my part at this point.  To my knowledge, we have yet to see a bot evolved from a zerobot ancestor develop any sort of conditional reproduction logic.  (Doing so would qualify for winning the Conditional Logic Prize.)

If you want to stick with a hand-authorred starting point, you can use autocosts to jack CostX up when population spikes occur to keep it under control.  You might also try the latest drop 2.43.1f.  It's a lot faster for larger, sparser sims, but the benifits will be limited when you get dense clusters of cancerous growth.  It will help, but you will still see the slowdown when bots reproduce explosivly.  Eventually your hand authorred bots will all devolve and 10 million cycles down the road, your sim will be full of tiny bots that try to reproduce every cycle no matter where you start from.  They will all be cancerous, but with no artifically stored reserve to exploit.

Moonfisher:
Doesn't that depend on the sim ? If it's a not a veggy and it devolves to break it's repro gene it'll become too small to gather energy from shots and ties won't it ? So eventualy anything with a broken repro gene should die out if it's not an alge...
Either way I had the same problem, and I had huge bots, so when one of them mutated it could spawn lik 5000 litle bots...
And since I'm very impatient I used the following gene :

cond
*.body 20 <
*.robage 0 >
start
0 .repro store
-2 shoot store
31999 .shootval store
stop

This should clear out those growths rather fast, so they don't slow down the sim for too long...

Testlund:

--- Quote from: EricL ---Naturally evolved bots will not store nrg needlessly waiting to reproduce.  They will reproduce whenever they can (whenever they aquire enough body that the simulator allows them to).
--- End quote ---

I find this fascinating. The program has something to teach us here. For instance, when programming bots everyone uses '.repro', but in an evosim the program choses 'inc' instead. When I ran my long evosim last year, more than half a year around the clock, I don't think I ever saw .repro in the dna. The program kind of shows us what it has been designed for. I'm thinking here that in the computer world there is no such thing as absolut randomness. I'm thinking that a program makes choices even though it has been designed to be random.
When I found that I couldn't load my saves because they got corrupted, I started over by loading a sim from when I first started this sim, which contained bots wirth only zeros in it. After between 100000-200000 cycles I got reproduction going just like last time. I also discovered a bot that had the same position and offspring with the same colors after this time, as when I ran it the first time. So the mutations must have occured exactly the same as the first time. First I thought that DB must bee seeded instead of random, but that doesn't make sense either. Here's a screenshot of the bot with it's offspring:

EricL:

--- Quote from: Testlund ---First I thought that DB must bee seeded instead of random...
--- End quote ---
If you started from a saved sim as opposed to beginning a brand new sim, then the seed is the same and the sim should play out the same subject to version differences and external influences beyond the local simulator's control such as internet mode and teleporters.

Testlund:
You mean the mutations are pre-set when to occure from the beginning and in what order? What if I run a sim for several years, or forever? I don't see how it could be possible for the program to calculate that much mutations before starting the sim, or maybe the mutations goes through a limited list and then start from the beginning, making the same mutations repeat themselves. These bots got the same colors and chosed to reproduce and place themselves exactly like the first time.

Navigation

[0] Message Index

[#] Next page

[*] Previous page