37: Friction 2

Here we remember old challenge discussions that got purged in the great forum cleanup of 2018.
penbeuz
Posts: 1
Joined: Tue Oct 04, 2016 9:55 am

37: Friction 2

Post by penbeuz » Wed Oct 05, 2016 3:53 pm

I'm stuck with this one for several days, I tried evolve from a crawler without success.
Is there a way to know what are the parameters of the substrate of the challenge? I'm not quite sure mine is nearly the same.
User avatar
Alast
Posts: 2182
Joined: Thu Jul 02, 2015 9:15 am

Re: 37: Friction 2

Post by Alast » Wed Oct 05, 2016 4:51 pm

You can only approximate the substrate. There is no way to copy it or see the settings directly.

What helped me on this one was creating more surface on both ends of the Myocyte for better grip/traction.
Perfection hasnt reached me yet, but its trying hard!
User avatar
wapcaplet
Posts: 306
Joined: Thu Dec 08, 2016 9:52 pm

Re: 37: Friction 2

Post by wapcaplet » Fri Dec 09, 2016 3:27 am

This is my first post here, though I have been lurking for a few days. Gotta say a quick thanks to Petter for such a great app, and Bwisialo, Alast and others for all the insights - they've really helped me get through some of the harder (and even "easy") challenges!

I have not solved Friction 2 yet, but I'm darn close. For Friction 1, I designed a 3-celled inchworm (Genome here):
  • M1 -> M2 + M3
  • M2 -> M4 + M5
  • M5 -> M5 / M1
  • M1: Lipo (though phago seems to work just as well)
  • M2: Neuro (S1/S2 on 1-hour cycle)
  • M3: Phago
  • M4: Myo (using S1/S2 to affect contraction and lift)
  • M5: Phago
The S1/S2 signal cycles are a bit off, so it doesn't crawl very efficiently, and it only crawls in a straight line. It just barely reaches 300 cells in Friction 1, with a bit of luck in placing the seed cells.

After watching Bwisialo's video on evolution, I spent a couple hours mutating this genome on the experimental plate, using a mix of natural and artificial selection. The first big change to emerge was some tweaked split angles that caused it to crawl in an arc. Just like with flagello swimmers, arcs or spirals work better than straight lines for maximizing food. The next improvement I noticed was that the S1/S2 signaling fixed itself so it crawls properly, without pushing its tail backwards at each step (though I still don't quite understand why). It looks downright bouncy compared to the original!

Once I had an evolved worm that could survive long-term in conditions similar to Friction 2, I let it run overnight to get my "final" (for now) version. I tried this genome in Friction 1 and it maxed out the plate at 600 cells in no time. But it still can't hack it in Friction 2; cell count is fairly stable around 300-400, and occasionally peaks in the mid-500s, but never quite reaches 600. Here is the evolved genome.

It looks like this design is as optimized as it's gonna get with evolution, but I'm sure there are other approaches I haven't thought of. All the manual tweaks I've tried on this evolved worm just make it worse. I briefly tried a 4-celled version with a Lipocyte, hoping that would keep it alive longer, but the extra weight really hampered movement, and there just don't seem to be enough nutrients to sustain a 4-celled worm for long.

Any tips or advice would be more than welcome. I'm sure I'll get it eventually, but I'd love some ideas.
User avatar
bwisialo
Posts: 2955
Joined: Sat Aug 15, 2015 10:22 pm

Re: 37: Friction 2

Post by bwisialo » Sat Dec 10, 2016 11:49 am

I'd like to help but your genome opens as txt via Drive for me.
amor fati
User avatar
Reaper Elite
Posts: 737
Joined: Tue Jul 19, 2016 10:55 pm
Location: In the Underworld eating bacon and watching anime with Marilyn Monroe

Re: 37: Friction 2

Post by Reaper Elite » Sat Dec 10, 2016 6:09 pm

Wap change its name with a .genome extension and reupload then it should work.
"A mercenary can always find a war to fight if they look hard enough"
"You become a changed person when you face me and deny me your soul."
User avatar
wapcaplet
Posts: 306
Joined: Thu Dec 08, 2016 9:52 pm

Re: 37: Friction 2

Post by wapcaplet » Sun Dec 11, 2016 12:47 am

Hmm, they have a .genome extension, but the link only works for me if I open it in Chrome. If I let the Drive app open it, all I get is garbled text (it thinks it's an email message for some reason). I'd use dropbox but my account was recently closed due to inactivity. Try opening in Chrome or another browser maybe? Sorry for the inconvenience. :oops:
User avatar
bwisialo
Posts: 2955
Joined: Sat Aug 15, 2015 10:22 pm

Re: 37: Friction 2

Post by bwisialo » Sun Dec 11, 2016 1:01 am

No worries. When I open in Chrome, my phone says "Opening in Google Drive." Probably a default setting. I can probably download to my computer then open it on my phone.
amor fati
User avatar
bwisialo
Posts: 2955
Joined: Sat Aug 15, 2015 10:22 pm

Re: 37: Friction 2

Post by bwisialo » Sun Dec 11, 2016 7:36 am

Hmm... That’s a tough one. It's hard to give advice without simply saying what I did on some hunches. Neurocyte adjustment is fairly technical and can take some guesswork.

Looking at your Neurocyte settings, it seemed like S1 and S2 were too high and out of balance for a good oscillation. I ticked Snap values, set the 2nd channel to -52.78; the 3rd to 34.82; the 4th to 10.

The other change is straightforward: too much fixed bend. Changed that to a= -1; b=-.19.

In good starting position, I got four initial cells to hit 630.

There are no doubt other ways to get to 600 with this genome, but those seemed like some necessary changes.
amor fati
User avatar
bwisialo
Posts: 2955
Joined: Sat Aug 15, 2015 10:22 pm

Re: 37: Friction 2

Post by bwisialo » Sun Dec 11, 2016 6:57 pm

For future use:

I guess the general advice for Neurocytes doesn't require anything technical. Just check the settings; try increasing and decreasing one channel at a time; and see if anything impoves you're results.

With Myocytes, muscle movement means energy use. So, if you have a good organism, try decreasing values a bit: you might get all the movement you need, save some energy, and get a bit more reproduction. For bending, tight curves usually won't help: larger curves (lower bend values) can help and use less energy.
amor fati
User avatar
wapcaplet
Posts: 306
Joined: Thu Dec 08, 2016 9:52 pm

Re: 37: Friction 2

Post by wapcaplet » Mon Dec 12, 2016 10:20 pm

Thanks for the tips! I still haven't gotten it to succeed in Friction 2, but you got me to take a closer look at the neurocyte settings--it looks like evolution introduced some weirdness there, like using S3 as the basis for S1, when nothing is emitting S3, plus some changes to nutrient priorities, and lower split mass on cells that shouldn't split (like the myocyte). I'm probably better off starting fresh so I can experiment with the design in a more controlled way.

Good point on myocyte energy use, too. I never thought about how much energy could be wasted that way, but now that I'm into the later challenges, I have more appreciation for the value of minimizing movement, and even sitting still when it's practical (like the orange critters in "Predator").
Locked