1000x Gravity Snail - Evolved Virtual Creature

[fighterpiolt1992]

too bad this isnt for linux operating systems :(

[ArePenguinsAwesome2]

Am I the only 13 year old who likes these kinds of programs?

[cbrhawk1]

I'm trying a huge population and tiny tournament size to keep the gene pool diversified. Is that going to help or hurt?

My goal is to soften up the root regression that always shows up. With small populations, you have a huge root regression of a higher number as I've experienced multiple times, why some runs turn out good and bad IMO.

[vincentmagician]

... wouldlike to see what happens in a higher gavity for these little guys

[radonas]

im going to set gravity x100 and choos height

[pythor2]

0:44 he tipped the little guy over! what a big bully.

[danielkirk1]

Can some evolutions turn out 'junk'? I've made about 2 awesome evolutions about 5 *snore* evolutions :P

Could lowering the population help refine better creatures in some circumstances? Thanks.

[kjlg74]

@danielkirk1 Oh yeah, some runs just don't do well at all. So far on this channel I've posted pretty much every creature that I or others have evolved (at least the ones they've sent me) so the variety seen in the videos - the number of interesting vs dull creatures, it roughly on par with what one can expect from the program.

(continued)

[kjlg74]

As for small population size, I find it helps refine creatures more quickly as long as their performance doesn't vary too too much (i.e. same creature does terribly one time and fantastic the next). I sometimes start with a large population, for the variety that offers, and then once the population settles to a single body plan, I reduce the population size to quickly refine that body plan.

[danielkirk1]

@kjlg74 Ah thanks for that :) I tried this and I'm getting much nicer results.

You good sir deserve a nobel prize for this program. Cheers!

[kjlg74]

@danielkirk1 Thanks a lot, danielkirk1! Feel free to send my way any interesting creatures you might evolve, if you like. I'll be happy to record and post videos of them. Alternatively, if you post any creature vids of your own, I'll add a link on the web site :)

[TheReasonWhyGuy]

they're like little race cars XD

[kjlg74]

@TheReasonWhyGuy Bumper cars! :D

[saadhamza1]

i had tried to run mine almost all night.... when i woke up... my dad had turned off the computer.

[kjlg74]

Well, I can see how that might be a problem :P

If the program wasn't properly shut down, it's possible that your .evolution or one of your .creatures files has become corrupted. That might prevent you from continuing the evolution.

[saadhamza1]

@kjlg74 i continued it and it worked alraighht. so now worries :P. but i had to restart the thing T~T.

i hope i get something really cool like the end over end worm. now THAT was cool. im going to name my generations" EPIC"

[kjlg74]

The chances of getting something as interesting as the end-over-end worm are pretty small, but it could happen.

[saadhamza1]

@kjlg74 yea, but we can always hope :P.

[Draconic74]

What about setting gravity at x100? A lot less, but still quite a bit.

Would be interesting to see what would happen if you put a high gravity creature into normal gravity.

[kjlg74]

I know this one is pretty much immobile at regular gravity. I tried it. It's abuse of the physics engine stops happening, and it doesn't really move about anymore.

As for 100x gravity instead of 1000x, there is one such creature already on the channel - check out the "100x gravity snail". IMO it's more interesting than this one, though it also uses physics engine loopholes to move around.

[04carrolld]

should of called him goku snail

[belliebum12]

You should set gravity to .001 normal strength and see what you get.

[kjlg74]

Pretty much anything that moved would throw itself a good distance, I expect.

[Zetimenvec]

@kjlg74

maybe couple that with a preference to stay low to the ground?

[kjlg74]

Why the hell don't I think of these things!? :D That's actually a clever idea, and doable without me having to make any coding changes! Thanks!

My CPU is currently busy with more pressing computation tasks (also evolutionary computation, but not creatures) - but if anyone wants to try this low-gravity-ground-hugger idea, I can explain how to set it up. I have no idea what to expect.

[Zetimenvec]

I actually already started one, lol.

[kjlg74]

Excellent! :)

I was thinking - watch out for the balance between the "distance" weight slider and the "TOG (time on ground)" weight slider for the fitness function. I suspect if both are set to full, the creature will not be very strongly encouraged to stay on the ground. You might want distance slider reduced to, say, 60-70%.

This is just a hunch. Could be I'm totally wrong. You might want to do several independent runs with different settings to get a feel for it.

[Zetimenvec]

yeah, I let it go about 20 generations before it was obvious they were just hurling themselves into the air. I actually settled on putting the sliders at .10 for movement and 1 for tog

[kjlg74]

Sounds good.

[Zetimenvec]

okay, I feel like the TOG function isn't properly weighted or properly being measured. I currently have movement set to .01 and TOG set to 1, and while they don't get any points for hurling into the air, they do seem to be getting some points for touching the ground, and THEN they hurl themselves up, kinda like a spring loading itself against the ground, and then once a specific block touches the ground it hurls itself up. I dunno what do to anymore :\

[kjlg74]

That's frustrating! :(

If I remember correctly (I'm fair sure) all of the sliders are actual weights that contribute to a weighted sum (i.e. fitness = w1*v1 + w2*v2 +... where w's are weights and v's are values) except for TOG which is used like: fitness = TOG*(w1v1 + w2v2 + ...) where TOG is the fraction of time spent touching the ground, between 0 and 1.

..continued..

[kjlg74]

..It could be that jumping is just *way* more of an easy route toward increasing fitness early in a run. Letting them evolve longer might help - but I have no idea how likely that is. Next step might be to a) let it go longer just to see, or b) gravity is just too too low and might need increasing.

[Zetimenvec]

right, well, I'm on generation 412 right now. I'm guessing that if I were to increase the runtime it may introduce the variable of landing, because by the time they launch themselves there isn't enough time for them to land again. so I think I'm going to run the next criteria in the fitness schedual to be more of the same but just longer evaluation time. If that's not helping the issue then I'll decrease the movement slider by another factor of 10!

[kjlg74]

412 generations! that was fast - if it's stuck by then, it's probably stuck for good. Definitely time to try a change.

[videonfan]

it looks cute!