Darwinbots Forum

Bots and Simulations => DNA - General => Topic started by: Welwordion on May 02, 2006, 05:31:45 PM

Title: A lazy turner(turning gene)
Post by: Welwordion on May 02, 2006, 05:31:45 PM: Ok I wanted to create a gene that lets a twobot Mb turn 180 degree when he sees his partner. But strangely he does so only once after birth.

Code: [Select]
cond *.numties 0 = *.robage 1 > start 5 .repro store .function inc stop cond *.robage 1 = *.refeye *.myeye = start .tie inc stop cond *.eye5 0 > *.refxpos *.trefxpos = *.refypos *.trefypos = start 618 .aimdx store stop end
Title: A lazy turner(turning gene)
Post by: Numsgil on May 02, 2006, 07:18:12 PM: I think it has to do with the tienum resetting (tienum let's you designate which tie to use).

Which version are you running? Try running the bot in the latest 2.4 and see if it works. I think Eric fixed it so tienum doesn't reset anymore.
Title: A lazy turner(turning gene)
Post by: Welwordion on May 03, 2006, 01:31:47 AM: I am running the latest version.
Title: A lazy turner(turning gene)
Post by: Numsgil on May 03, 2006, 01:55:56 AM: Try adding something like:

cond
*.tienum 1 !=
start
1 .tienum store
stop

to your bot and tell me if that fixes it.
Title: A lazy turner(turning gene)
Post by: Welwordion on May 03, 2006, 11:07:45 AM: nope no effect
Title: A lazy turner(turning gene)
Post by: Numsgil on May 03, 2006, 11:16:20 AM: Dunno then.
Title: A lazy turner(turning gene)
Post by: EricL on May 03, 2006, 12:51:21 PM: It seems to do what I expect it to do. It reproduces on cycle 2 then turns away from it's offspring. The offspring creates a tie to it's parent on it's cycle 1 then turns away so they are back to back. None of the genes in either bot fire after that for obvious reasons.

What are you expecting it to do?

Note that the reason the parent turns away is different from the reason the offspring turns away.

The parent turns away because both .trefxpos and .refxpos are 0 (as are .trefypos and .refypos). Why are they both zero? Well, both trefxpos and refxpos come from the tied or viewed bot's own xpos. Since the offspring has a higher bot number than the parent, it has yet to execute a single cycle when the birth tie is formed or when the parent first sees it. Thus, it's xpos has yet to be populated and thus when the trefvars are copied for the bith tie or when the refvars are copied when the parent first sees the offspring, the parent's refxpos and trefxpos get 0 because the offsping's xpos is still 0.

The offsping turns away becuase when it sees the parent and creates the tie to the parent, trefxpos and refxpos get populated with real values, which equal.
Title: A lazy turner(turning gene)
Post by: Welwordion on May 03, 2006, 01:41:35 PM: Well in my situation they tend to rotate(with the tie) around each other while their eyes stay in the same direction, so when they made a 180 degree rotation their eyes face each other, in this case they should see each other and turn away, but they do not.( oh and could someone answer my birthtie topic in bugs and fixes?)

Simplified they should turn away every time they see each other.
Title: A lazy turner(turning gene)
Post by: EricL on May 03, 2006, 01:55:15 PM: Quote from: Welwordion
Simplified they should turn away every time they see each other.
Ah. It's a bug. The trefvars only last a cycle. I made them sticky in the Tie routines, but they are being erased each cycle in EraseSenses() in another part of the code. Nice find. On the list for the next release.

Note that only the offspring will turn away. The parent will not since the trefvars are only autopopulated for the bot which initiated the tie.
Title: A lazy turner(turning gene)
Post by: Henk on May 03, 2006, 04:55:57 PM: I thought .readtie was required to use the tref commands?
Title: A lazy turner(turning gene)
Post by: EricL on May 03, 2006, 06:23:47 PM: Quote from: Henk
I thought .readtie was required to use the tref commands?
The autopopulate beginning in 2.42.3. See #10 here (http://www.darwinbots.com/Forum/index.php?showtopic=1240).
Title: A lazy turner(turning gene)
Post by: Henk on May 04, 2006, 02:52:24 AM: Ah ok, thanks
Title: A lazy turner(turning gene)
Post by: PurpleYouko on May 04, 2006, 11:50:18 AM: Quote from: Welwordion
Well in my situation they tend to rotate(with the tie) around each other while their eyes stay in the same direction, so when they made a 180 degree rotation their eyes face each other, in this case they should see each other and turn away, but they do not.( oh and could someone answer my birthtie topic in bugs and fixes?)

Simplified they should turn away every time they see each other.
The reason for this is that once the tie has stiffened, you cannot rotate relative to it by using normal rotation commands

You gotta set .tieang or tiang1 if you want the robots to face away from each other. You can't rely on any rotations prior to the tie hardening, to set the angle correctly. Use something like..

Code: [Select]
cond *.multi 1 = start 628 .tieang1 store stop
That will set the angle from the eye to the tie at exactly half a circle as soon as the bot becomes a part of an MB (ie. when the tie hardens after 20 cycles)
You can even add a tag of some sort so that the gene doesn't get activated on every cycle.
Title: A lazy turner(turning gene)
Post by: Welwordion on May 08, 2006, 01:17:41 PM: hmm does not work, they do not turn away from each other and I found out they keep rotating and rotating on and on,so this effect is not by intiasl inertia.
Title: A lazy turner(turning gene)
Post by: EricL on May 08, 2006, 01:24:42 PM: Quote from: Welwordion
hmm does not work, they do not turn away from each other and I found out they keep rotating and rotating on and on,so this effect is not by intiasl inertia.
Ties don't stiffen and lock their angles (as they should) in the 2.4 fork prior to the as yet unreleased 2.42.4 which is why they keep rotating in the version you are using.

They don't turn away from each other becuase of the trefvar issue above.

All this will be fixed in 2.42.4 which hopefully I can release later today.