Re: WS3.0 Philosophy and Status

Hi Stuart,

Thanks for your excellent summary.  In another project I started looking at
the slippy tile format (essentially what is used by open-street-map, google
maps, bing, and many many others.)  By merging on-the-fly downloaded sat
tiles (from .... you know ... wherever) and draping them over SRTM terrain,
I was able to get something that approximated what google earth looked like
15 years ago.  It dynamically downloads and builds as you fly (and caches
locally).  The slippy tile format enables quad-tree LOD schemes.  All of
what I've done is in python using the panda3d scene graph module ... these
days I've relegated C++ to mostly my arduino flight controller project and
do everything else in python when possible ...

https://youtu.be/d6G-cHO-uTU

I don't have nearly enough time to work on fun stuff these days, but
someday I was hoping to find ways to add buildings and trees.  This
approach does have all the expected downsides ... no lighting,
airports/taxiways get really blurry when you land/taxi, weird stuff happens
around bridges and tunnels, all the shading is cooked into the sat imagery,
etc. and if one of your online data sources goes away you have to scramble
to find something similar ...

It's not really in a form that anyone can just download and run without
getting a bunch of other pieces running at the same time, but if anyone
wants to poke around through my messes:

https://github.com/NorthStarUAS/Simulator

Curt.


On Thu, May 28, 2026 at 3:59 AM Stuart Buchanan <[email protected]> wrote:

> Hi Folks,
>
> Extracting one of the specific questions that came out of another thread:
>
> On Wed, May 27, 2026 at 12:20 PM James Turner wrote:
>
>> On 27 May 2026, at 11:21, Willie Fleming wrote:
>>
>> I think the best bang for the buck commits would be stuff I desperately
>> want to know myself like
>>
>> What is the philosophy behind WS3.0 and what is its status?
>> How do the various rendering options work?
>> How can I get good results from each aicraft with the new compsitor.
>>
>> If i knew that stuff Id be delighted to write it up and commit.
>>
>>
>> Well, the former, isn’t a docs topic, it’s a ‘ask here’ topic, and Stuart
>> will enlighten you :)
>>
>
> The philosophy behind WS3.0 was to replace the original terrain scenery
> (which I'll shorthand as WS2.0, though that's not really accurate) with
> something more maintainable and which would address many of the limitations
> that had become more and more apparent.
>
> For those not familiar, WS2.0 terrain is made up of tiles of approximately
> 30kmx30km.  Each tile is an irregular mesh of triangles, and each triangle
> is a flat surface of a particular landclass (forest, desert, water, road).
> Airports are generated as separate files and holes are created in them mesh
> for them to slot into.
>
> It's worth acknowledging how ground-breaking FlightGear's original scenery
> was:  It was designed from the ground-up to properly model the earth's
> curved surface using WGS84 at a time when contemporary simulators were
> building flat terrain, and support a diverse range of scenery sources.  It
> is a really impressive piece of work.
>
> However, it has a number of fundamental limitations:
> - As the visibility range increases, more and more tiles have to be
> loaded.  Doubling the visibility range from 50km to 100km requires 4 times
> as many tiles to be loaded.
> - Increasing the scenery detail level increases the number of triangles
> massively.  This is particularly apparent when adding more roads to the
> terrain itself.
> - The tile edges are irregular and vary depending on the triangles.  You
> cannot update just a small piece of scenery without creating
> discontinuities in the mesh.
> - Updating airport layouts requires regenerating the underlying scenery,
> - The mesh wasn't "graphics card friendly" as each landclass required a
> separate Effect.
>
> Furthermore, there were specific maintenance limitations:
> - WS2.0 scenery generation process required a very large custom database.
> Martin Spott did a fantastic job creating this database covering the entire
> world and using a variety of datasource at a time when satellite data was
> less ubiquitous than it is now.  However he retired from the project and
> the status of the database was unclear.
> - Even with the database intact, the process to update with new scenery
> data was difficult.
> - Because of the impact of airport layout changes on the underlying
> terrain, airports were not updated.
>
> WS3.0 seeks to address these limitations by:
> - Using LoD model so that scenery further away has less detail than
> scenery close by.  This means increasing visibility has far less
> performance impact.
> - Encoding landclass information as a texture rather than triangles so
> that a) increasing scenery detail has less performance impact and b) tile
> edges are regular and individual tiles can be regenerated. c) scenery
> rendering is more graphics card friendly
> - Draping line features such as roads over the top of the terrain at
> runtime
> - Fitting airports into the mesh at runtime so new airport layouts don't
> require new scenery.
> - Using more standard tools so that scenery generation is much easier for
> scenery developers.
>
> Current status:
> - WS3.0 is supported on "next", and will be part of Tech Preview 1.
> - The tooling for users to generate scenery themselves is available as a
> Docker image.  Wayne has been patiently testing this and finding plenty of
> bugs as he works on detailed scenery for the US.
> - We have detailed scenery for part of Europe, and Nia did a complete
> build of the world using Sentinel-2 data, which is available on Terrasync
> - We've still to work out the details on delivering WS2.0 and WS3.0
> scenery models in parallel.
> - There are still some rough edges.
>
> -Stuart
> _______________________________________________
> Flightgear-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/flightgear-devel
>


-- 
Curtis Olson

_______________________________________________
Flightgear-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/flightgear-devel