Page 9 of 10

SpaceEngine TODO

Posted: 24 Jul 2019 05:32
by FastFourierTransform
Why is the simulation of the evolution of galaxies and stars labeled uncertain if possible?
Because stellar and galactic evolution is an extremely complex thing to simulate (requiring supercomputers in general) and the variety of results (many kinds of stars and galaxies out there) is so large that is difficult to implement that number of animations and renderings (even if they are not the result of an actual physics simulation).
But Vladimir has shown time and time again that what looks impossible to code can eventually be coded, so be patient, we might see this one day.

SpaceEngine TODO

Posted: 03 Aug 2019 04:16
by FastFourierTransform
Now that we are having texture updates for the Solar System I would like to remind about an old discussion we had on the possibility to extend the level of detail further and further. If I recall correctly SpaceEngineer thought of a solution, similar to that of Google Earth, involving the centralized storage of extreme resolution textures in a server and the ability to download them from it interactively and in real time. As the user points the camera to different parts of a planetary body the program would send queries to those servers and load the required portions of the surface. This would in principle allow for huge improvements in the user's experience.

I've just found an open-soruce software that does just that. It is called OpenSpace, and as you can see here it has this kind of aproach for Earth, Mars and many other objects


[youtube]VPIvZMqEbBk[/youtube]

[youtube]NWZAg6qpMlE[/youtube]

For Earth and the Moon they have also solved a problem associated with the representation on programs like Google Earth: where satellite and aerial imagery looks patchy at best with different light conditions, colors associated with different seasonal conditions and different quality of the image, all added side by side. In the Mars video you can see this problem. It looks weird from far distances. But they seem to have solved this quite well for Earth as I said, their textures are very nice and continuous.

Since the project is open-source (here is the code) I was wondering if this could be implemented easily on SE. Having LRO and MRO resolution textures for the Moon and Mars would be a great feature in the future.

SpaceEngine TODO

Posted: 03 Aug 2019 05:39
by N0B0DY
If I recall correctly SpaceEngineer thought of a solution, similar to that of Google Earth
If I recall correctly (and since I happened to be around in the old forums from the early days of SE development) the initial idea was to implement a procedural augmentation technique using a special refined algorithm to fill in the gaps of the catalog mesh data. This idea at some point was abandoned maybe / probably because of its complexity and difficulty (?). If this had been possible, no centralized server and thousands of GB of data would be required. Of course the zoomed in image of these bodies might have been a bit unrealistic compared to the real data but who would care. I wouldn't. But again I am probably an exception :) ..

SpaceEngine TODO

Posted: 03 Aug 2019 06:13
by FastFourierTransform
the initial idea was to implement a procedural augmentation
Yes that was also a proposal. Both ideas are compatible as long as they operate in different resolution regimes. From longer distances to closer ones:
  1. Default SE planet textures (currently on SE) [~ 5 km/pixel - 0.5 km/pixel resolutions]
  2. Server data gathering (like in OpenSpace and Google Earth) [~ 0.5 km/pixel - 10 m/pixel resolutions]
  3. Procedural reconstruction for lower level details (like in Outerra) [~ 10 m/pixel - 10 cm/pixel resolutions]
  4. Surface texture packs (currently on SE) [~ 10 cm/pixel - 1 mm/pixel resolutions].
This is hard work for sure. 4 different rendering algorithms should operate at the same time on different levels and the transitions might be the most difficult part of it all. But two of those things are implemented by now (the planet textures and small scale surface texture packs) and the algorithm for gathering real data from the servers is public because of OpenSpace. The procedural intelligent generation of the terrain in between those regimes is probably way more difficult than taking real data from the server (even if it requires internet connection which might be a small inconvenience for some users) because of the diversity and complexity of geologic features to be inferred from lower level details. If we get real data for higher resolutions we could potentially overcome the need for procedural augmentation since it would operate in such a narrow range of resolutions that it would be no longer needed; we could jumpt from server gathered real data to the small scale textures we have right now.

This at least could be reasonable for Earth, Moon and Mars. But for other object like Io, Umbriel, Triton etc... we don't have enought resolution in real data to fill the gap between server loaded real terrains and centimiter scale textures. In those cases procedural augmentation might be needed. Outerra does some of that and the algorithms are somehow explained, but here would have to develope a different algorithm to generate those extraterrestrial terrains from scratch (not an easy task).


Outerra has a 150 m detail texture for Earth that looks like this:
Image

The rest of details are procedurally generated to this level of detail:
Image

At least this is how it was 10 years ago!!!

SpaceEngine TODO

Posted: 04 Sep 2019 16:06
by Kinematic
I would love to see more detailed textures on earth
Possibly by implementing a procedural generation similar to outerra which procedurally generates landscapes in high res while using real map data for large geographical areas
Although with everything else in SE it would not be used to its full potential as to not compromise performance

SpaceEngine TODO

Posted: 04 Sep 2019 17:43
by A-L-E-X
Now that we are having texture updates for the Solar System I would like to remind about an old discussion we had on the possibility to extend the level of detail further and further. If I recall correctly SpaceEngineer thought of a solution, similar to that of Google Earth, involving the centralized storage of extreme resolution textures in a server and the ability to download them from it interactively and in real time. As the user points the camera to different parts of a planetary body the program would send queries to those servers and load the required portions of the surface. This would in principle allow for huge improvements in the user's experience.

I've just found an open-soruce software that does just that. It is called OpenSpace, and as you can see here it has this kind of aproach for Earth, Mars and many other objects


[youtube]VPIvZMqEbBk[/youtube]

[youtube]NWZAg6qpMlE[/youtube]

For Earth and the Moon they have also solved a problem associated with the representation on programs like Google Earth: where satellite and aerial imagery looks patchy at best with different light conditions, colors associated with different seasonal conditions and different quality of the image, all added side by side. In the Mars video you can see this problem. It looks weird from far distances. But they seem to have solved this quite well for Earth as I said, their textures are very nice and continuous.

Since the project is open-source (here is the code) I was wondering if this could be implemented easily on SE. Having LRO and MRO resolution textures for the Moon and Mars would be a great feature in the future.
The new PBS series Planets showed the planets, their rings and moons in amazing detail from the Galileo, Cassini and New Horizons missions, can we incorporate that data into SE?

SpaceEngine TODO

Posted: 11 Sep 2019 07:04
by epoch5
Do you plan to support Windows Mixed Reality VR in the future?

SpaceEngine TODO

Posted: 06 Oct 2019 08:25
by longname
How come this hasn't been updated yet?

SpaceEngine TODO

Posted: 11 Oct 2019 13:32
by FastFourierTransform
It is called OpenSpace, and as you can see here it has this kind of aproach for Earth, Mars and many other objects
Apparently the new version of Microsoft Flight Simulator has similar strategies for Earth's terrain rendering that OpenSpace has.
[youtube]ZGDb-hlqQR0[/youtube]

More on how this works:

[youtube]-c13VFfHxdc[/youtube]

SpaceEngine TODO

Posted: 02 Nov 2019 13:47
by FastFourierTransform
Regarding these two incredibly difficult goals:

SpaceEngineer:
Modeling of asteroid collisions with lighting, explosion and the formation of the crater
SpaceEngineer:
Simulation of the collision, tidal or artificial destruction of the planets, formation of the asteroid belt and the dust disk around the sun, its further evolution
There is a smoothed-particle hydrodynamics (SPH) N-body integrator, open-sourced and available for free, for researchers, called OpenSPH. The code is written in C++ (like SpaceEngine). It has been developed by the Astronomical Institute of Charles University in Prague.


This week ESO released a simulation, made with OpenSPH (and visualized with OpenVDB), of the possible formation of Hygeia (the results are published in Nature). I find outstanding the fact that this is not only an animation but a physical simulation of the event:
[youtube]qhTEQsH2xMA[/youtube]
This is also a beautifull simulation made a year ago (when this was still in its infancy):


[youtube]GYCHTdJNlvo[/youtube]

Image

I hope this continues to evolve to the point SpaceEngine can implement it for real-time rendering of Planetary and Asteroidal collisions.

SpaceEngine TODO

Posted: 05 Nov 2019 01:21
by Watsisname
Wow, those are incredible simulations.  Surreal.  Now I dream of seeing the Earth-Theia impact rendered this way as well -- presumably much harder than for asteroids, but maybe it will be possible in the near future. :)

SpaceEngine TODO

Posted: 14 Nov 2019 03:18
by PlutonianEmpire
My favorite bits of the MSFS2020 previews are the volumetric clouds and crepuscular rays. Now imagine that in space engine for Binary and Multiples systems :D

SpaceEngine TODO

Posted: 15 Nov 2019 09:12
by Lujo
I'm waiting for volumetric clouds and 3d water. After that update i wouldn't mind if it costs a 100€, i would pay it in a second. P.S when will this TODO list be updated?

SpaceEngine TODO

Posted: 15 Nov 2019 15:49
by N0B0DY
Quoted from the dev posts:
SpaceEngine lacks shadows, this is very noticeable here  :) This is one of the highest priority items to implement in future updates. Together with finishing the PBR implementation (Physically based rendering). It wasn’t completed for the 0.990 release, so all SE ships still use the old rendering method (though they have reflections now).
I am really looking forward to this..
Then we can start talking about volumetric effects, weather and other MSFS2020 eye candies :).

SpaceEngine TODO

Posted: 17 Nov 2019 02:17
by vlad01
volumetric clouds with shadows would be an awesome thing to add.  As the couple posts up mentions, water improvements are something thats needed.