Darwinbots Forum
General => Off Topic => Topic started by: Greven on May 19, 2005, 10:47:32 AM
-
Have you ever run a evo sim?
If you have, what have developed and evolved of interesting bots / behavior?
But then again what is interesting bots / behavior? Be critical and no lying :rolleyes:
Read this post first! (http://s9.invisionfree.com/DarwinBots_Forum/index.php?showtopic=344&st=30) (it is a post by me from the evo sim topic in hints, tips and tricks ....)
-
I kinda did , but not in db.
Here is a list:
[you]Make a robot fight me:[/you]
Robot Sneacks up and attacks from behind , moves faster then turning ratio , no way to win. (I would of done better if I was not too lazy to use the up and down keys, but it was calculating faster then me anyway, slim chance) (This is done with only one dna and retrys)
[you]Make a bunch of robot eat one tiny food source in the middle of the screen:[/you]
Robots line up in to a snake and eat the food in a line , like they are standing on line , they share energy but are not killing each other for food.
Finaly my favorete:
[you]Crank Up your AntiVirus and make an .exe file reproduce and mutate:[/you]
AntiVirus was closing illigal copys until it totaly messed up the antivirus window, I have this some were in another post, It did not close the window , it kinda made a "work of art" out of it.
-
I normally run evo sims to test the limits of the software so naturally they crash more than most people's do.
A couple of times I have just run a few million cycles for the hell of it. Almost invariably my robots end up losing the ability to move then dying a slow painful death.
They even do this while chasing an active veggie like my Rabbit or Shvarze's Alga Grexa.
I am not going to give up though.
I will figure out a way to make these darned things evolve properly even if I have to rewrite the whole evolution system.
I think the main problem is that there are never any ecological niches for evolution to exploit.
-
Almost invariably my robots end up losing the ability to move then dying a slow painful death.
Hint: Turn down the mutation rate :)
-
Hint: Turn down the mutation rate
The last set that did this (5 times consecutively) was run with a mutation rate set to 25,000 on the largest possible sim size, using moving veggies (Rabbit in this case)
At the beginning, my simple robots were able to chase the rabbits and feed on them. A robot would continue to feed this way until.
1) The veggie died.
2) The bost lost the veggie due to an abrupt course change on the part of the veggie.
In both these scenarios the bot typically has upward of 20,000 energy an so it goes into a spasm of multiple birthing. This results in a large group of robots in an otherwise empty space with no veggies anywhere in sight.
Sometimes a robot will lose its conspec recognition and eat all of its siblings while they just sit there doing nothing or spinning on the spot
The next time a veggie passes by, the surviving bot/s from the litter more often than not either can't follow it or just plain aren't interested in it.
Eventually after a few million cycles (usually less than 20 accumulated mutations) they all just sit there and ignore everything around them till they die.
I had this sim set on auto-restart and it did so 5 times over a period of well over a week. Each extinction was for the same reason. Lethargy!
-
I reckon my stsrter bot must be prone to natural selection which takes it up an evolutionary dead end.
The survivoors seem to be those that take out siblings with little or no movement.
:(
-
The next time a veggie passes by, the surviving bot/s from the litter more often than not either can't follow it or just plain aren't interested in it.
Most of the off-springs must be just as good at hunting rabbits as the parent. What you describe sounds exactly like "Muller's ratchet". Also, your rate of mutation accumulation is 5 times higher than mine - I get 4.3 mutation per million cycles (or about 6 mutations per 100,000 born bots). All of this tells me your mutation rate is too high.
How long is your bot's DNA?
How quickly do you see "average mutation" graph go up?
-
For several generations the offspring remained as good as (in fact mostly identical to) the parent.
It was only after mutations started to accumulate that the problems began.
Tell you what. I will post my start-bot and Rabbit into the beastiary and you can try it out for yourself.
-
Ok , The only explonation I see is that for some reason after reproduction somthing causes the parants to die off.
Look at the logical side of things: Ok, A robot gave burth and lost some energy in the prosses, but if the robot reproduces with twice the energy it started the simulation with, and the energy it started the simulation with is enough for it to find food and reproduce. Why the hell is it dieing of if the parants dna is supposed to remain unchanged?. Maybe they simply lose the abilaty to reproduce?
Anyway I beleave the problem is there somewere , once again somthing is causing healthy robots to die.
I mean really what kind of scifi is this: a robot generation is unable to turn and move but is able to kill off all its parants...
-
PY, I tried your start bot and it can't shoot its own shit. It just truns away from rabbit whenever it sees it. Could you post the settings file you use? And also check that you posted correct file.
-
Haven't ran the bot yet but I saw at least one problem with it in regards to mutation. That last gene uses *.eye5 != as a condition for avoiding conspecs I'm assuming that you expect it to be comparing to 0 but if any loose value is inserted and not gotten rid of the bot could start rotating for no reason. Normally insertion seems to be the most useful or neutral mutation but for this bot it'd be detrimental.
Endy B)
-
Eventually after a few million cycles (usually less than 20 accumulated mutations) they all just sit there and ignore everything around them till they die.
This and the spinning PY is describing is actually what I always get to!!!???
Why is that? I often have atleast around 500 bots, but suddenly, 495 of them
starts to spin on the spot, and they die out. The last 5 runs maddely around to find food and die! PING! The sim is over!!!
What I also wrote in the evo sim topic, I think the problem with DB is not lack of complex envoriment, but because the DNA structure is so tight that most mutations are bad, and it is so complex we have a very hardtime to find the problem/bugs, because so much can interfere with this: All the genes of a bot etc.
Few people have the time or opppertunity to run sims worth 10 millions cycles or more, and therefore then simple answer to that is setting the mutation rates high (meaning more mutations), this is not desirable, and a lot of mutations, as pointed out by shvarz, in a short time makes bad sims, because natural selection have a hardtime to pick out the best bots(you know what I mean).
Over the weekend I will try to stechout a new DNA system, and post it. I know this is a tabu, but I think if we need to get over the last edge we need a new DNA structure.
Taking into account that it will not make the older bots incompatible with the new system, and I know it will take 100 hours of coding and very much is need to change, but we need it! I would volentieer to help coding if needed!
-
Haven't ran the bot yet but I saw at least one problem with it in regards to mutation. That last gene uses *.eye5 != as a condition for avoiding conspecs I'm assuming that you expect it to be comparing to 0 but if any loose value is inserted and not gotten rid of the bot could start rotating for no reason. Normally insertion seems to be the most useful or neutral mutation but for this bot it'd be detrimental.
Endy B)
The whole point for inserting the "*.eye5 !=" condition is to stop it spinning. Remember *.eye5 is always renewed on every cycle so it can never hold a stray value.
This condition is designed to make sure that the robot can only rotate if it is actively seeing something at the time.
-
Over the weekend I will try to stechout a new DNA system, and post it. I know this is a tabu, but I think if we need to get over the last edge we need a new DNA structure.
Nothing is tabu here.
If you come up with a superior system then we use it. If you don't then we won't. Simple as that.
-
PY, I tried your start bot and it can't shoot its own shit. It just truns away from rabbit whenever it sees it. Could you post the settings file you use? And also check that you posted correct file.
I have just taken a look at the robots again and I did find one problem in the current settings setup that will stop the sim working most of the time.
You will notice that the start bot has a speed limit of 20 while the rabbit can currently go faster than that. Try adjusting the start bot such that it is a little faster than the rabbit. (originally the rabbit was slower but I sped it up) I hoped that faster bots would arise through mutations.
I don't see any problem with the bots turning away from the rabbits. They work just fine here. They can't keep up with the rabbits but they sure as hell chase them.
Here is my settings file.
I will double check my robots in the beastiary to make sure they are the right ones.
-
Halleluja!!!!!! Great Bot GOD :D
-
I will double check my robots in the beastiary to make sure they are the right ones.
Yup they are the right ones. Note that Simple Startbot2 has a top speed of 20 while the rabbit has a top speed of 30.
D'ya reckon they might mutate faster speeds?
Probably not.
Maybe we should level the playing field a little and try them both at the same speed.
Or maybe Simple Startbot 2 should be like a hunting cat and run like heck when it sees a rabbit but otherwise amble along slower.
Adding a gene such as
cond
*.eye5 0 !=
*.refeye *.myeye !=
start
35 *.vel sub .up store
stop
could do the trick. It will be just a bit faster than the rabbit but only for short bursts to conserve energy.
-
What I also wrote in the evo sim topic, I think the problem with DB is not lack of complex envoriment, but because the DNA structure is so tight that most mutations are bad, and it is so complex we have a very hardtime to find the problem/bugs, because so much can interfere with this: All the genes of a bot etc.
I am afraid that I disagreee with this statement Greven.
Evolution is basically a two part process.
1) Random mutations.
2) Selective pressures, survival of the fittest.
In DB we have the first part more or less sorted. Mutations are 100% random. They may be a little limited in scope by the DNA language itself but they are definitely Random
Part 2 is the problem as I see it. There is only one selective pressure in DB. find the food and eat it. In a bland world with no niches, changes in physical constants or any kind of variation whatsoever, this just doesn't cut it.
We should get stronger fighters eventually but we wil never get robots evolving in different directions to form a stable ecology.
-
You have a very good point there PY!
But you must agree on the DNA structure of DB is very very complex, compared to other AL sims. When we try to mimic the world through arbitary rules we should not expect 'real-life' behavior. Still DB is, as far as I know, the most ambious (cant spell it right sorry) project yet undertaken, with behavior in mind, and succes in many points when it comes to design of bot behavior. But what really lack, is 'WOW'-evolution, meaning things we couldnt design from scratch, or havent through of, suddenly appear (evolve).
Those things Shvarz is doing (or AZPaul by the way), is guided evolution! This is 'props' to the semi-creationist ;)
When I think about it a little, I can see that the enviroment isn't complex at all, so I will withdraw my statment about this. :redface:
But I still think the DNA structure need some re-work! :boing: :boing:
-
Think about it! When a tie-feeder suddenly evolve from a simple bot! What a true unbelivable [span style=\'font-size:14pt;line-height:100%\']FEAT/WONDER[/span]! or if the bots suddenly start to communicate, or a species evolve to the fishes or some kind of ant behavior appears!
OMG! It will be so wonderful! I know that it will be way into the future, but still...
Think about the feeling, all the emotions that go through you in THAT moment!
When some of that happens to me, I will give everyone of you a beer!!!!!!!!!!
-
And I see your point too Greven.
The entire DNA language of DB is completely artificial.
We have been through this kind of thing before and my main problem has always been that there is no way that any bot can ever develop something that cannot be manually programmed. Therefore the physical restrictions of the bot's DNA code also restrict things that it can evolve into.
This is one of the reasons why I support the addition of as many sysvars, commands and abilitie/interactions of abilities as possible. It will never make the system 100% open ended but it might be that evolution will come up with something that we haven't thought of first.
The only way that I see to make a true evolution system would be to change the DNA language beyond all recognition. Make it exactly (or equivalent of) like real DNA with almost infinite combinations of simple building blocks.
The restrictions then depend on the program to interpret what these building blocks can become and that is where the real problem lies.
How the heck can you code something to be able to do anything? :blink:
Let's say a bot mutates the ability to make sounds?
The program has to be able to cope with this. That means that it must already contain the code to create/modify and play .wav files.
If a bot becomes multi-cellular, the program has to be coded to deal with all the consequences of this.
Basically you have to hard wire every physical law in the universe into the program to make this possible. The concept just boggles my mind. :wacko:
As far as DB is concerned, I think we need to work on increasing the variation of the environment and the complexity of the DNA language. DBs need more capabilities to do more stuff and unfortunately we have no option other than to hard wire it into the program.
If you know another way then speak up before my brain explodes.
-
Those things Shvarz is doing (or AZPaul by the way), is guided evolution! This is 'props' to the semi-creationist
:shoot: :pokey: :angry: :burnup: :plzdie: :banghead: :bash: :cuss: :tantrum:
You have to defend that! How is what I am doing a guided evolution? I take bots, drop them into environment and let them reproduce. I don't guide them. Hell, I would not even know how to guide them! What are you talking about?
:pokey: :pokey: :pokey: :pokey: :pokey:
-
I think I understand what he means.
It isn't really guided evolution per se though.
It is just that your specific environment (over which you have absolute control) guides the bots toward a certain goal. That goal being the best DNA to survive within your environment.
Therefore any specific environment does tend to favor particular mutations so it is (in a very loose sense) guided toward a certain end.
Let's say you set up a high friction sim with moving veggies (like you did). It will favor bots that are able to chase down the veggies. You know it will do this and that non motile bots are likely to die. This could be construed as a kind of guiding although I wouldn't personally go that far.
The only way to get away from this is to mutate the environment also then there can be no accusations whatsoever of guiding.
-
Duh! This is stupid!
(by the way, feel free to move it to the rant section any time you like).
So the fact that organisms adapt to the environment is "guided evolution" now? This REALLY PISSES ME OFF! This is the most natural thing and this is how evolution proceeds in real life! Environment forces organisms to adapt. Take Ice Age - it's gotten colder and colder and most warm-loving animals died out and life on Earth adapted to cold environment. OFF COURSE life adapted to cold environment, because IT WAS GETTING COLDER every day! Is this guided evolution?
As I said:
REALLY PISSES ME OFF!!!
I decide to create an environment that is different from F1 league conditions and people start calling me creationist!
OK, I am gonna go meditate for a while
:bigginangel:
-
Sure it is how real evolution works and who is to say that isn't guided too?
I say that it is guided.... by the environment!
There is no inteligent guidence involved of course but prevailing conditions sure as hell influence which mutations survive and which don't.
It is obvious that your artificially created environments are a veiled attempt to play God with the helpless little DBs :P
Just kidding ;)
All evolutions is guided by prevailing conditions. Mutations aren't but the prevailing direction of the evolution has to be. That is what natural selection is all about.
You are obviously getting robots that are the most suited for their environment. That is exactly as the TOE predicts. All creatures are a direct product of their environment.
However, the environment is a complete fake as all DB environments are, so in that respect the evolution is guided. As long as we have control over the environment then it can't be any other way can it?
If God exists/ed and the ice age was deliberately created to force a new path for evolution then it would by definition be guided. Guiding requires an inteligent input.
Still a valid experiment though :D
-
Muller's Ratchet, as Schvarz mentioned before, is the overwhelming law of thumb in this case.
I myself didn't know about Muller's Ratchet until Schvarz and AZPaul mentioned it a few months ago.
The concept, basically, is that once an organism has a harmful mutation in a vital gene, neither it nor it's descendants have a chance to recieve a good copy of that gene. And if this organism is more successful than an organism without this mutation in the short run (like the creation of a cannibot (or a cousinonly-cannibot especially)) then this deletrious mutation will get spread to future generations.
Wiki link:
Muller's Ratchet (http://en.wikipedia.org/wiki/Muller%27s_ratchet).
DB is definately along the right tracks for realism. I don't think I've ever heard of Muller's Ratchet being so harmful to another AL sim before. BTW, once you get sexual reproduction, the negative effects of Muller's Ratchet are largely nullified.
-
Read my lips... err.. fingers: S-H-V-A-R-Z :pokey:
I am in extremely angry mood today :help: That's what whole week spent working on computer does to me. Otherwise I am just an :bigginangel:
-
uh... did I mention spelling isn't my strong point?
-
Sorry Shvarz, I had no intentions to insult you, but what PY is saying, is actual what I mean, I sometimes have great and clear thoughts :D , but I have a had time converting them into proper words! :(
We have been through this kind of thing before and my main problem has always been that there is no way that any bot can ever develop something that cannot be manually programmed. Therefore the physical restrictions of the bot's DNA code also restrict things that it can evolve into.
Again this is true! Humans has done the opposite through the entire recorded history. Looking at the nature and copying it! (To some extent)
But think about an eye! The complex structure is so ilogical. Think about trying to make an eye from scrath in the real world? Or making an camera with the same structure that an eye have!
Everthing in this world can be DNA-coded (again to some extent), and many things in DB might appear before our eyes, that we didnt think of!
I know that I am comparing DB with the real world, but then again some (major) adjustments will make more (fun) evolution simulations.
-
I promised to creat an entire new DNA structure in the weekend!
This project I have undertaken is much larger than I ever thought of!
There are so many things to think of, it is very difficult to have 1000 of things in mind.
It may take atleast a week more! :)
-
It may take atleast a week more!
If you can get a working version up by Christmas I will be impressed.
This is one heck of a huge job that you have taken on. :blink:
-
Evo-sims come in two flavors, chocolate and vanilla.
The vanilla kind involve creating a bot, some food, maybe an enemy, an environment and letting DB mutations act over x million cycles to see what happens. This is fun. Not really very productive, but fun in that it is interesting to see what can develop, which isn’t much, usually, but it's fun anyway.
The chocolate kind is a bit more complex. In a chocolate flavored evo-sim you are setting up a system to test some very specific evolutionary concept. You must take care to control everything but the test variable and be careful to NOT guide the test variable where you think you want it to go.
DBII mutations may be random but they are also overly extensive, coarse and extreme; well outside natural events.
Look at a mutation in nature. Except for muto-toxic poisoning which can change entire segments of genes (thalidomide) the vast majority of mutations involve changing one letter of the three-letter codon on a gene. In some cases this mutation does nothing since there are multiple codons coding for the same amino acid in the protein syntheses. In the other cases the codon change causes a different amino to be placed into the protein at that specific location. The resulting protein may fold differently and may have different electrochemical properties. It may or may not act differently within the system (cell). If the protein is a vital function agent, like hemoglobin, or is essential to other vital systems the mutant individual may not live. In other cases the individual may live but have a slightly longer arm, more muscle fibers in the biceps, truncated ganglia in the spine, you name it. The environment then determines if the individual prospers or not.
Most mutations in DBII are of the muto-toxin variety. No subtlety. Most DBII mutations radically change the gene function in the extreme or make the operation nonsensical.
cond
*.eye5 40 >
*.in1 *.out1 !=
*.refnrg 3000 <
*.attacked 0 =
*.horny 0 =
start
mult mult 625
stop
A series of DBII mutations is apt to take the above gene and do this:
cond
*.memval 40 >
*.in1 *.out1 !=
*.aimdx .shootval <
*.attacked 0 =
*.horny 0 =
start
mult mult *.hitsx
stop
or this:
cond
*.eye5 40 >
*.repro *.out1 !=
*.shot 3000 <
*.attacked 0 =
*.horny 0 =
start
store
dec sub inc 6 sub inc inc sub inc inc rnd sub inc store
store
6 6 store
inc store
5 rnd rnd 6 div sub mult .vel 625
stop
A more realistic mutation would take a constant and randomly apply a +- 10%.
cond
*.eye5 36 >
*.in1 *.out1 !=
*.refnrg 3300 <
*.attacked 0 =
*.horny -1 =
start
mult mult 688
stop
Or change a variable or operator to some other related type.
cond
*.eye1 40 >
*.in1 *.out2 =
*.nrg 3000 <
*.attacked 0 =
*.horny 0 =
start
mult rnd 625
stop
This will still change the function of a specific gene and may kill the individual but it may also lend more subtlety with less radical change thereby allowing, IMO, more realistic and more useful change to develop.
For chocolate sims the DBII mutations are too coarse. In my mind DBII mutations are too coarse in the extreme for any reasonable evo-sim, even vanilla flavored, except as a game. But, you gotta start somewhere and I can imagine the coding effort is already quite intense. Besides, I can turn off mutations.
I do chocolate-flavored un-guided non-creationist evo-sims. At least I would do them if [begin hint] .sexrepro was working [end hint] which, in my genius, I managed to work around except now I can’t get to the data since [begin hint] there is no way to dump memvals into a text file [end hint] for analysis.
Having said all this I hold out much hope for the future. DBII is by far the only system with the flexibility and capability to approach my needs. You folk have already done a superb job in putting this thing together and from what I’ve read in this forum your plans for the future are all stellar.
For now I lurk the updates with hope. And I’m learning that coarse vanilla can be quite fun.
Verbosely yours as always,
-P
-
When I'm done playing with the minor changes in DB and when I get tired of working on the egrid (so many problems! So little computer memory) I'd like to set up a comprehensive and hierarchial list of mutations and how they should work.
If anyone has any good references to how other ALife sims deal with their mutations (I was passively looking into Avida) I'd love to get a look and compare them all.
-
As far as values in memory locations is concerned, DB was recently altered so that mutations don't jump vast differences in value any more. (Change was applied for 2.36)
It now uses a gaussian function for mutations so that when a value is changed, the new value will probably be very similar to the old one.
It is now virtually impossible for something like..
854 .depth =
to mutate to.....
25687 .depth =
It is highly likely that the new value will be within 10% of the original value in the new code.
The same formula is (unfortunately) applied to memory locations too, as the DNA parser sees numbers and memory locations in very much the same way. This means that .depth (217) could easily mutate to any of the following.
shdn 211
shdx 212
shsx 213
edge 214
fixed 215
fixpos 216
depth 217
daytime 218
xpos 219
ypos 217
kills 220
In most cases the sysvars are reasonably well related to each other but it wouldn't be too hard for .....
854 .depth =
to mutate to.....
854 .fixpos store
That would have a pretty bad outcome since the robot will then lose all ability to move.
We are heading in the right direction but the mutations certainly do need to be tightened still further.
-
Perhaps we should go through the program and the sysvars file and drastically rearrange the order of all of them. They're about 60% related to each other by location in the file, but if we can increase that, mutations would be more likely to give good results.
-
That would probably be a good idea but the one problem I see with doing it is that some mutated bots that are using direct memory stores instead of sysvars, would then work completely different.
I have a mutated Hunter that has absolutely no repro commands in its DNA, yet it still reproduces.
Also Devincio Solo uses a similar repro technique. If .repro was changed from 300 to something like 500 then these robots would no longer be viable.
Just something to think about.
-
It's definately a trade off. But if we can devise a relatively logical and consistant sysvars pattern we'd only have to do this once.
There are only ~200 sysvars, so we should be able to scatter the sysvars around so that even future sysvars have a logical place they should go.