Procedural terrain in DCS? - Page 2 - ED Forums
 


Notices

Reply
 
Thread Tools Display Modes
Old 03-27-2017, 02:45 AM   #11
Weta43
ED Testers Team
 
Weta43's Avatar
 
Join Date: Feb 2005
Location: Aro Valley Wellington New Zealand
Posts: 5,881
Reputation power: 76
Weta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to behold
Default

Quote:
Procedural generation is not random. On the contrary, it is completely deterministic.
True, but the way the real world has been developed is not, so as long as the map is of areas where the terrain is largely the result of human activity, it will never put things in 'the right place', rather it will put them 'where they should logically be'...

It may be repeatable, logically consistent. beautiful, system light, and immersive, but for flight sims where players complain that about the number of rivets on an aircraft, or that the length of time a battery can remain in before being depleted, it's just not the right answer to the problem (though as mentioned, it can fill in some of the detail - grass, bushes etc - within defined areas)
__________________
Cheers.
Weta43 is offline   Reply With Quote
Old 03-27-2017, 03:57 AM   #12
jamie_c
Junior Member
 
Join Date: Nov 2016
Posts: 6
Reputation power: 0
jamie_c is on a distinguished road
Default

There are problems with procedural terrain, for one is the sheer size of the data if you populate the base data with alterations to the terrain base and expect to have buildings representative of specific regions, for the longest time OT didn't even have Biomes everything was green, there was 1 tree everywhere. Additionally the algorithms mean what you see at altitude is a procedural approximation of what you get as you get closer to the surface which is further procedurally defined. You might say "well if they overlay OSM data it'll be fine", not so, take a look at real roads, the earth around them is shaped almost always, we elevate them, slope them, cut into mountains. Defining a perfectly flat road segment takes x number of bytes in the octree, each segment has a defined start, end, texture, width, thickness, elevation and an enumeration for how it alters the terrain and probably many other things. Roads alone bloat the data set and we haven't even mentioned other natural and artificial features. The more you depart from the default dataset in the case of OT, the larger the storage requirement.

OT's representation of the earth is as a cube stored in an octree data structure (separated into a folder structure last I looked, +x, -x, +y, -y, +z, -z, each subdivided 4 times at the top level, as you get further down it's a lot like google maps, 1 big tile becomes 4 tiles per "zoom level", they then use an algorithm to "round out" the sides with an Earth Centred, Earth Fixed (ECEF) coordinate system. The base data they generate from is essentially NASA SRTM with 90m post spacing, it has to be translated into the OT native data structures for it to work much like roads and other features.

This is before we even consider the physics aspect required for high fidelity flight sim where the gravity vector points toward the origin instead of simply being a negative value on an (seem to remember y is vertical in DCS) axis. Other sims like FSX and X-Plane are more like DCS in that you're on a terrain treadmill. I used to work with VBS (as military) and even it had a curved earth rendering option from VBS 2 version 1.3 I think but that was back when we were only doing single large chunks of terrain up to 200km, maximum weapons employment range was generally 4km for any platform. Physics wise nothing changed, it employed flat earth physics as a pure shader implemntation.

I'm not saying ED can't achieve this in anyway, it's just a totally different beast that has been in development for years. With DCS you're talking changes to the underlying architecture, a significant workload and a large investment with less returns. You essentially have two choices, a handcrafted high detail area that has been checked extensively for quality (NTTR, vegas strip), or mass produced approximation and at times floating objects, mangled roads and so on.
jamie_c is offline   Reply With Quote
Old 03-27-2017, 06:21 AM   #13
sobek
Veteran
 
sobek's Avatar
 
Join Date: May 2008
Location: Austria
Posts: 11,761
Reputation power: 186
sobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud ofsobek has much to be proud of
Default

Quote:
Originally Posted by Weta43 View Post
True, but the way the real world has been developed is not, so as long as the map is of areas where the terrain is largely the result of human activity, it will never put things in 'the right place', rather it will put them 'where they should logically be'...

It may be repeatable, logically consistent. beautiful, system light, and immersive, but for flight sims where players complain that about the number of rivets on an aircraft, or that the length of time a battery can remain in before being depleted, it's just not the right answer to the problem (though as mentioned, it can fill in some of the detail - grass, bushes etc - within defined areas)
As everything, it is a compromise, in this case between asset size and the LOD where things start to divert from RL. Interestingly enough, nobody ever complained that a blob on the noise texture for the Caucasus is in the wrong spot. I'm being facetious, but you get the point, in order to keep the size of a terrain sane, i believe there's no alternative to creating a lot of things procedurally and managing one's expectation of what can realistically be achieved in a simulation with regard to RL detail.
__________________
Good, fast, cheap. Choose any two.
Come let's eat grandpa!
Use punctuation, save lives!
sobek is offline   Reply With Quote
Old 03-28-2017, 06:10 AM   #14
Weta43
ED Testers Team
 
Weta43's Avatar
 
Join Date: Feb 2005
Location: Aro Valley Wellington New Zealand
Posts: 5,881
Reputation power: 76
Weta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to beholdWeta43 is a splendid one to behold
Default

I agree there has to be an element of procedural generation (there already is in DCS.world), & if it were in addition that what are currently the 'dead' spaces around the existing maps (Turkey / Russia / Crimea / California etc) had a coarse mesh and procedurally generated vegetation / towns / cities (vegetation & hamlets totally P.G., small towns and cities boundaries manually defined then automatically 'populated'), that would be better than dead space.

But (& I know you're not suggesting it is) it's not going to be a case of throwing algorithms at an elevation map of Korea & expecting people to be happy with the pretty countryside it generates as a Korean map...

(Much as I used to love making Far Cry maps & using covering those with interesting vegetation).
__________________
Cheers.
Weta43 is offline   Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

All times are GMT. The time now is 11:44 AM. vBulletin Skin by ForumMonkeys. Powered by vBulletin®.
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.