On 1/16/2012 6:47 PM, Casey Ransberger wrote:
Top post. Heightmapping can go a really long way. Probably not news
though:)
I am still not certain, since a lot of this has a lot more to do with my
own project than with general issues in computing.
I had messed with a few technologies already.
height-maps (long ago, not much used since then, generally randomized).
the issue was mostly one of being "not terribly interesting", but it
makes sense if one wants terrain (and is "fairly cheap" in terms of
memory use and performance impact).
a more advanced variety would be to combine a height-map with a
tile-map, where the terrain generator would also vary the texture-map to
give a little more interest. I have considered this as a possibility.
also tried randomly generated voxel terrain (similar to Minecraft, using
perlin noise). issues were of being difficult to integrate well with my
existing technology, and being very expensive in terms of both rendering
and memory usage (particularly for storing intermediate meshes). one may
need to devote about 500MB-1GB of RAM to the problem to have a
moderately sized world with (with similar specifics to those in Minecraft).
I suspect that, apart from making something like Minecraft, the
technology is a bit too expensive and limited to really be all that
"generally useful" at this point in time and on current hardware (I
suspect, however, it will probably be much more relevant on future HW).
I also tried randomly generated grid-based areas (basically, stuff is
built from pre-made parts and randomly-chosen parts are put on a grid).
I had also tried combining this with maze-generation algorithms. the
results were "functional" but also "nothing to get excited about". the
big drawback was that I couldn't really think of any way to make the
results of such a grid based generator "particularly interesting" (this
is I think more so with a first-person viewpoint: such a structure is
far less visually interesting from the inside than with a top-down or
isometric view).
it could work if one were sufficiently desperate, but I doubt it would
be able to hold interest of players for all that long absent "something
else of redeeming value".
the "main maps" in my case mostly use a Quake/Doom3/... style maps,
composed mostly of entities (defined in terms of collections of
key/value pairs representing a given object), "brushes" (convex
polyhedra), "patches" (Bezier Surfaces), and "meshes" (mostly
unstructured polygonal meshes).
these would generally be created manually, by placing every object and
piece of geometry visible in the world, but this is fairly
effort-intensive, and simply running head first into it tends to quickly
drain my motivation (resulting in me producing worlds which look like
"big boxes with some random crap in them").
sadly, random generation not on a grid of some sort is a much more
complex problem (nor random generation directly in terms of unstructured
or loosely-structured geometry).
fractals exist and work well on things like rocks or trees or terrain,
but I haven't found a good way to apply them to "general" map generation
problem (such as generating an interesting place to run around in and
battle enemies, and get to the exit).
the "problem domain" is potentially best suited to some sort of maze
algorithm, but in my own tests, this fairly quickly stopped being all
that interesting. the "upper end" I think for this sort of thing was
likely the .Hack series games (which had a lot of apparently randomly
generated dungeons).
it is sad that I can't seem to pull off maps even half as interesting as
those (generally created by hand) in commercial games from well over a
decade ago. I can have a 3D engine which is technically much more
advanced (or, at least, runs considerably slower on much faster hardware
with moderately more features), but apart from reusing maps made by
other people for other games, I can't make it even a small amount nearly
as "interesting" or "inspiring".
On Jan 16, 2012, at 8:45 AM, David Barbour <[email protected]
<mailto:[email protected]>> wrote:
Consider offloading some of your creativity burden onto your
computer. The idea is:
It's easier to recognize and refine something interesting than to
create it.
So turn it into a search, recognition, and refinement problem, and
automate creation. There are various techniques, which certainly can
be combined:
* constraint programming
* generative grammar programming
* genetic programming
* seeded fractals
You might be surprised about how much of a world can be easily
written with code rather than mapping. A map can be simplified by
marking regions up with code and using libraries of procedures. Code
can sometimes be simplified by having it read a simple map or image.
Remember, the basic role of programming is to automate that which
bores you.
Regards,
Dave
On Sun, Jan 15, 2012 at 4:18 PM, BGB <[email protected]
<mailto:[email protected]>> wrote:
I am generally personally stuck on the issue of how to make
"interesting" 3D worlds for a game-style project while lacking in
both personal creativity and either artistic skill or a team of
artists to do it (creating decent-looking 3D worlds generally
requires a fair amount of effort, and is in-fact I suspect
somewhat bigger than the effort required to make a "passable" 3D
model of an object in a 3D modeling app, since at least generally
the model is smaller and well-defined).
it seems some that creativity (or what little of it exists) is
stifled by it requiring a large amount of effort (all at once)
for the activity needed to express said creativity (vs things
which are either easy to do all at once, or can be easily
decomposed into lots of incremental activities spread over a
large period of time).
trying to build a non-trivial scene (something which would be
"passable" in a modern 3D game) at the level of dragging around
and placing/resizing/... cubes and/or messing with individual
polygon-faces in a mapper-tool is sort of a motivation killer
(one can wish for some sort of "higher level" way to express the
scene).
meanwhile, writing code, despite (in the grand scale) requiring
far more time and effort, seems to be a lot more enjoyable (but,
one can't really build a world in code, as this is more the
mapper-tool's domain).
_______________________________________________
fonc mailing list
[email protected] <mailto:[email protected]>
http://vpri.org/mailman/listinfo/fonc
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc
