Problems with Custom Polymodels on Descent 2

Pumo · Post by **Pumo** » Sat Dec 03, 2005 10:44 pm

I have some problems with custom polymodels i'm making for Descent 2.

Some work just right, but some others didn't.
The problem is this:
In some models, there are some segments that doesn't show at all when they are divided in submodels.
When the entire robot is made of only one submodel, it works right.
And another issue is that, even on divided polymodels, some submodels can be seen from the oposite side of the robot in some angles (it looks something like a semi-transparent model, where you can see what's behind it).
I've tryied various submodel divisions, with more or less divisions, diferent sorting in Polytron32, but except on one model, it doesn't work for the others.
What can be the problem? Someone can help me?

PS: BTW, the models are converted from 3DS to COB to POL, and all of the segments are composed of only triangles, retouched on Polytron. Can this be a problem?

jakee308 · Post by **jakee308** » Fri Dec 09, 2005 5:24 am

i probably know less than you about the polymodels. my work with polytron32 has shown me that seeing through the submodel may mean the polygon normal is facing the wrong way. this causes the texture to not be drawn on the correct side. submodels disappearing may be because they penetrate other submodels. polytron32 was a work in progress and did not have complete functionality. if you got the robot construction kit off planetdescent.com then the help manuals may tell you something. i wish they hadn't stopped developing the programs for creating robots.

Sirius · Post by **Sirius** » Fri Dec 09, 2005 10:42 pm

Um... getting this right is quite complicated, and simply putting in more submodel divisions isn't necessarily going to help. You need to be certain where you want them and why - and if you haven't seen the manual describing the action of the BSP tree, you really should as otherwise you won't have much idea what you're doing.

Now, touching up in Polytron - this can work depending on the polymodel. However, IF your submodels are not simple, convex shapes - touching them in Polytron at all is going to totally nuke them as far as Descent is concerned. Anything cob2pm did to make it render properly will be erased, and you're pretty much SOL getting it back.

Unfortunately this means you have to make sure that the model is usable in the Descent engine as soon as it reaches .pol format. You can, of course, do texturing using POLVIEW - but you can't texture MAP anything efficiently using it, and you should probably do this, again, before it reaches .pol format.

Personally, my more complicated robots are started, built, and almost finished using Truespace 3.2. I only leave the texture mapping of the eyes and guns for POLVIEW - and, of course, the inevitable reassignment of textures.

It would be much nicer if either TS3.2's texture mapping wasn't gloriously imprecise (there is no workaround to make it otherwise), or POLVIEW actually had some texture mapping, but I guess you make do with what you have. The source code doesn't appear to be getting released any time soon.

jakee308 · Post by **jakee308** » Fri Dec 09, 2005 11:00 pm

what manual? where?

edit: hah! thank you sirius! just downloaded free version of truespace 3.2 from caligari. also thanks to igda forum.
here is link for anyone interested.
http://forms.caligari.com/forms/ts3all_free.html

So a new toy to play with! mmmmmmm! thanks again for the clues and btw i mean it, jot stuff down about robot editing or if you have it on file but it needs pruning or if you just need it typed or proofread. send it to me and i'll fix it up or whatever. you may be one of few left who understands how that stuff works.

Sirius · Post by **Sirius** » Mon Dec 12, 2005 12:17 am

The manual I mentioned is this:
http://www.descent2.com/descman/polytro ... /index.htm

You should especially read the sections on chunks, submodels and BSP trees. They explain in quite easy-to-understand terms how the whole system works.

Another document that will undoubtedly help you is this one:
http://mywebpages.comcast.net/knudsong/ ... index.html

That is mainly applicable if you are using Truespace, and generally tells you what you can do with it that you can't with Polytron.

Just for reference, models such as the level 3 bosses of Project: Mandrill, the Mandrill player ship (Stingray I think), the Obsidian player ship and the level 13, 14 and 15 Apocalyptic Factor bosses were all built using Truespace. If you look at them you will see that the complexity is quite a long way beyond what can be done with Polytron (that is, if you know how much can be done with Polytron; it is quite limiting in fact, and few Descent 1 or 2 robots could have been built with it).

Pumo · Post by **Pumo** » Tue Dec 20, 2005 8:49 pm

Well, first of all thanks for the answers!

At this time i already fixed most of my robots by selecting some normals in the submodels for the sorting tree (after reading again the manual you say, Siruis).

However, there is a robot that i cannot divide into submodels (and it's a simple one), because some of the vertex are duplicated and i don't know how, Polytron shares them with all the other submodels (including submodel 0 ).
I must divide it, because it's concave and to select a normal to face to the other submodels for the sorting tree. So the problem are those duplicated shared vertex. I'm not so familiarized with Truespace, that's why i use D3Edit. Maybe the problem is not when converting from COB to POL or ORF to 3DS, but when converting from 3DS to COB (i use a program named Crossroads).

But anyway, thanks for your help!

Kyouryuu · Post by **Kyouryuu** » Fri Dec 23, 2005 12:48 pm

Note though that some robot designs have those weird clipping issues, even in the production robots. The Spider in the Vertigo missions often has its back legs clip in front of its front ones. The Boarshead and Secondary Lifters have appendages that twist and contort strangely.

It's just a weird science. I remember the convex/concave thing. But then you have models like the Guidebot that are all one submodel, but certainly aren't convex.

Sirius · Post by **Sirius** » Fri Dec 23, 2005 7:12 pm

I have no idea what they used to build the robots in Descent 1 through Descent 2 Vertigo, but it certainly wasn't Polytron.

It wouldn't surprise me that much if they split the geometry by hand, actually.

Pumo · Post by **Pumo** » Sat Dec 24, 2005 7:00 pm

I haven't noticed the Secondary Lifter and Boarshead issue, but i think i've seen the one of the Vertigo Spider. However, now that most of my robots are fixed, it's not so much of a problem to make again the one i can't divide (it's only one).
I will try to make it again or see a way in Polytron to separate the vertex. Maybe with a cut and paste