Русский New site

Advanced search

[ New messages · Forum rules · Members ]
Page 1 of 11
Forum » SpaceEngine » Feedback and Suggestions » Procedural generation elements (A few issues)
Procedural generation elements
AycemanDate: Sunday, 26.08.2012, 00:42 | Message # 1
Space Tourist
Group: Users
Romania
Messages: 22
Status: Offline
Hello forum! I've been using Space Engine for a while now and I love it, but I registered here as whenever I'm flying around looking for perfect habitable systems it's hard finding them because of a few elements in the procedural generation system that causes unrealistic situations around the most common low mass stars. Starting with the most glaring issue:

1. For some reason, half the time, SE generates M-dwarfs that are too massive. Some M0V/M1V stars have up to 0.99 Solar masses which is totally off. M4 stars with around 0.65 MSol are quite common too. Alone, this isn't a huge problem, apart from the unrealistic orbital periods produced, but this leads to a frustratingly large body of K/M binaries, where the M dwarf is more massive than the K dwarf. Just recently I had to go through over 10 of these binaries before finding a realistic system.

2. Tidally locked Terras/Oceanias/Deserts around stars that contribute less heat than a nearby hotter star still have anti-solar ice caps.

3. The presence of large satellites around tidally locked planets is unrealistic. One can expect temporary asteroid moons to exist, however. I'd suggest having SE check for tidal lock and avoid placing any round satellites around tidally locked planets, maybe with the exception of a single one per system around a gas/ice giant, as its Hill sphere is large enough to allow enough time for initially farther away satellites to exist before being destroyed.

4. Brown dwarfs don't have a mass-age-spectrum variance yet. I see L2-T9 dwarfs at just over 84 Mj (which is a bit high for a BD unless it's a subdwarf) at any age.

5. Normally axial tilts tend to cluster around 0 in real life, because a planet initially spins in the same plane as the debris disk that formed it. In game there are too many planets with axial tilts close to 90 deg. There should probably be 3 tiers of probability (0-30 - biggest; 30-50 and 160-180 - medium; 50-160 - lowest). Tilt altering impacts do occur, but it's strange to see many planets around 60-90 degree tilts in a single system, especially since most of the energy of the impact is likely going into changing the rotation rate (which is why I also put 160-180 in the medium category).

EDIT: BTW, has the frequency of life planets changed from 0.96.1 to 0.96.2? It seems to take longer to find a habitable system.

EDIT2: In 0.96.1 I did find 2 Cool terras with life during the time that I had it, but now in one day I found 2 Cool terras with life (one of them at 126K) and one Frozen oceania with life at 90K (all of them had water/liquid at the surface as expected from this class of planet). I think this needs a check.


Edited by Ayceman - Sunday, 26.08.2012, 01:09
 
SpaceEngineerDate: Sunday, 26.08.2012, 13:38 | Message # 2
Author of Space Engine
Group: Administrators
Russian Federation
Messages: 4795
Status: Offline
Hi Ayceman, thanks for bug reports!

Quote (Ayceman)
1. For some reason, half the time, SE generates M-dwarfs that are too massive.

Yes, this is a bug in binary systems. Single stars should be generated normally.

Quote (Ayceman)
2. Tidally locked Terras/Oceanias/Deserts around stars that contribute less heat than a nearby hotter star still have anti-solar ice caps.

Tidal locked planets don't take into account secondary star. There should be some climate simulation and animated snow texture, that is impossible with current landscape engine.

Quote (Ayceman)
3. The presence of large satellites around tidally locked planets is unrealistic.

When SE generates a planet and a moon, it performs few calculation of their orbits stability using well known formulas. If moon don't pass the text, is being skipped. If you see a moon in the SE, this means that it is passed orbit stability test.

Quote (Ayceman)
4. Brown dwarfs don't have a mass-age-spectrum variance yet. I see L2-T9 dwarfs at just over 84 Mj (which is a bit high for a BD unless it's a subdwarf) at any age.

Yes, this was a bug too, already fixed.

Quote (Ayceman)
5. Normally axial tilts tend to cluster around 0 in real life, because a planet initially spins in the same plane as the debris disk that formed it.

This is probably a bug too, I'll take a look on it.

Quote (Ayceman)
EDIT: BTW, has the frequency of life planets changed from 0.96.1 to 0.96.2? It seems to take longer to find a habitable system.

No, in 0.9.6.2 I has fix a bug there habitable planets was generated only in binary star systems.

Quote (Ayceman)
EDIT2: In 0.96.1 I did find 2 Cool terras with life during the time that I had it, but now in one day I found 2 Cool terras with life (one of them at 126K) and one Frozen oceania with life at 90K (all of them had water/liquid at the surface as expected from this class of planet). I think this needs a check.

This is an old issue that cannot be fixed in currrent model. The planetary system generation function should perform some sort of climate evolution calculation to obtain realistic climate and surface textures on a planets with water. And surface texture itself should support animation, at least for snow cover.





 
AycemanDate: Sunday, 26.08.2012, 14:27 | Message # 3
Space Tourist
Group: Users
Romania
Messages: 22
Status: Offline
Ok, thanks for the replies, but:

Quote (SpaceEngineer)
Yes, this is a bug in binary systems. Single stars should be generated normally.


Not quite, I've seen a few single stars too, mostly in the 0.6-0.7 MSol range for M4V and 0.95-1 MSol range for M0V, the binary anomalies seem more spread across stellar classifications, but it could be my bias.

Quote (SpaceEngineer)
When SE generates a planet and a moon, it performs few calculation of their orbits stability using well known formulas. If moon don't pass the text, is being skipped. If you see a moon in the SE, this means that it is passed orbit stability test.


Orbit stability taking into account perturbations by other bodies is one thing, but the slow rotation of the planet itself can force a nominally stable orbit to decay (which is happening at Mars with Phobos) and I am sure this is not taken into account. On tidally locked worlds the difference is large (50+ day rotation versus <5day orbit seems quite common), and this will make the orbit decay very fast. There is the additional problem that a satellite will also have its orbit decay during the period when the planet slows down its rotation rate to match the host star, so most satellites are likely to be broken apart before the planet is tidally locked (The decay of a subsatellite of Iapetus is the reason it has a ridge now, and the initial collision occurred before tidal locking was achieved). Since giant planets are the only ones massive enough to have moons orbiting farther away when very close to their star, they could potentially retain a single large satellite for a longer period. Neptune - Triton is a good example, where instead the satellite orbits retrograde, resulting in effectively the same situation. Triton was captured later, and its orbit is decaying, but it will still take about 3 billion years to enter the Roche limit for solids because it started off far enough away.

Also, is the terrain generated after the planet itself and its orbital parameters are generated? If that is so, a check could be inserted in between to see what the heat flux is at its current distance from the various components of a multiple star system and every star's contribution. If the host dominates (contributes more than 50%), the generation algorithm can proceed as normal, if it doesn't, it skips generating an antisolar ice sheet.


Edited by Ayceman - Sunday, 26.08.2012, 14:45
 
Joey_PenguinDate: Sunday, 26.08.2012, 18:34 | Message # 4
Pioneer
Group: Users
United States
Messages: 311
Status: Offline
I have a few issues, too:

1) Epistellar planets are too common. Just about every system I've seen has a planet in a 2-4 day orbit, while they should be quite rare in reality. This leads me to

2) Not too much variety in solar system structure: You start with one epistellar planet, than find 6-9 more out according to the (outdated) Titus-Bode law, with a few debris belts and Nibiru-type elliptical orbit glitches. Can't we have less predictable star systems thrown in, like one with just one or two planets in wide orbits, or one with all 27 planets in chaotic, unstable orbits?

I love the program, though. So much potential!! smile





Careful. The PLATT Collective has spurs.

Edited by Joey_Penguin - Sunday, 26.08.2012, 18:37
 
SpaceEngineerDate: Sunday, 26.08.2012, 22:53 | Message # 5
Author of Space Engine
Group: Administrators
Russian Federation
Messages: 4795
Status: Offline
Quote (Ayceman)
and I am sure this is not taken into account.

You're right particulary. The tidal breaking caused by the sun is calculated first, then tidal breaking (or accelerating) caused by moon is calculated. In your system the moon can't accelerate planet's rotation much, so it rotates slowly. If moon is massive enough, you may obtain a synchronized binary planet system. To obtain the right rotation periods, the integral equation should be solved, this is an area for future work. Also, I have no analythic formula to calculate evolution of moon's orbit semimajor axis, so there are many moons with unrealistic low SMA generated.

Quote (Ayceman)
Also, is the terrain generated after the planet itself and its orbital parameters are generated? If that is so, a check could be inserted in between to see what the heat flux is at its current distance from the various components of a multiple star system and every star's contribution.

Yes, terrain parameters are generated based on physical parameters. But for suns too - and their luminosity may change a bit. Also, in the generation procedure, the temperature of the planet is computed for piriastron (i.e. planet and all other suns are placed at minimum distance to each other). This is used to compute right atmosphere dissipation. When simulations are started, the planet and suns come to their real positions, and the temperature of the planet drops. So sometimes terras and oceanias become very cold. But once a planet is generated, its class cannot be changed, so SE cannot change cold terras and oceanias to ice worlds.

Quote (Joey_Penguin)
Nibiru-type elliptical orbit glitches.

This is not a glitch, the high-elliptical planets do really exists. SE generates them with certain probablility (but won't perform orbit stability and intersection calculation, though).

Quote (Joey_Penguin)
Can't we have less predictable star systems thrown in, like one with just one or two planets in wide orbits, or one with all 27 planets in chaotic, unstable orbits?


SE cannot model unstable orbits. Only stable Kepler ellipses. So it can be only implemented "instantly unstable" systems, i.e. planets on elliptical orbits, but with a limited time range of simulation. I.e. as we see that system at this time +/- 1000 years, for example.

All these problems will be solved when new planetary system formation simulation is implemented (like in StarGen).





 
AycemanDate: Monday, 27.08.2012, 18:18 | Message # 6
Space Tourist
Group: Users
Romania
Messages: 22
Status: Offline
Quote (SpaceEngineer)
You're right particulary. The tidal breaking caused by the sun is calculated first, then tidal breaking (or accelerating) caused by moon is calculated. In your system the moon can't accelerate planet's rotation much, so it rotates slowly. If moon is massive enough, you may obtain a synchronized binary planet system. To obtain the right rotation periods, the integral equation should be solved, this is an area for future work. Also, I have no analythic formula to calculate evolution of moon's orbit semimajor axis, so there are many moons with unrealistic low SMA generated.


Until better generation algorithm can be implemented to handle these situations, you could just enforce an all out rule to not generate satellites if the planet is tidally locked (at least on planets below 10MEarth).

Quote (SpaceEngineer)
Yes, terrain parameters are generated based on physical parameters. But for suns too - and their luminosity may change a bit. Also, in the generation procedure, the temperature of the planet is computed for piriastron (i.e. planet and all other suns are placed at minimum distance to each other). This is used to compute right atmosphere dissipation. When simulations are started, the planet and suns come to their real positions, and the temperature of the planet drops. So sometimes terras and oceanias become very cold. But once a planet is generated, its class cannot be changed, so SE cannot change cold terras and oceanias to ice worlds.


I wasn't referring to the eliptical terra/oceania case which I understand. I was talking about close, tidally locked planets in multiple star systems. If the generation is based on periastron, it's good as the highest likely temperature will be used. If this thing can do the same calculation taking into account how other stars influence the planet at their periastron, the ice shell on half of a terra/oceania/desert problem can be avoided where necessary.

Thanks for taking the time to answering to these aspects, btw.

PS: When creating a cylindrical projection of a planet, is it possible to add an option to export the color/bump/alpha maps directly to png format?
 
SpaceEngineerDate: Monday, 27.08.2012, 20:31 | Message # 7
Author of Space Engine
Group: Administrators
Russian Federation
Messages: 4795
Status: Offline
Quote (Ayceman)
Until better generation algorithm can be implemented to handle these situations, you could just enforce an all out rule to not generate satellites if the planet is tidally locked (at least on planets below 10MEarth).

No, this is a question of the satellite maximum mass.

Quote (Ayceman)
If this thing can do the same calculation taking into account how other stars influence the planet at their periastron, the ice shell on half of a terra/oceania/desert problem can be avoided where necessary.

Maybe, I will work on this problem later.

Quote (Ayceman)
PS: When creating a cylindrical projection of a planet, is it possible to add an option to export the color/bump/alpha maps directly to png format?

Alpha channel has only a color map, and it's an ice mask. Water color/specular is calculated directly in the render shader, and not needed in the texture. But I will add an option for filling the oceans with blue in the future.

*





 
AycemanDate: Thursday, 30.08.2012, 22:55 | Message # 8
Space Tourist
Group: Users
Romania
Messages: 22
Status: Offline
Quote (SpaceEngineer)
No, this is a question of the satellite maximum mass.


Oh, in that case an upper limit of 200km (therefore asteroid) can be applied in this case for the moment. (if I'm understanding you right)
 
smjjamesDate: Tuesday, 04.09.2012, 04:00 | Message # 9
World Builder
Group: Users
United States
Messages: 913
Status: Offline
It may be a limitation of the generation engine, but I've noticed that, outside of globular clusters, there are no giant and larger procedural stars. I don't think I've seen any procedural subgiants either.




 
HarbingerDawnDate: Tuesday, 04.09.2012, 04:20 | Message # 10
Cosmic Curator
Group: Administrators
United States
Messages: 8711
Status: Offline
Quote (smjjames)
It may be a limitation of the generation engine, but I've noticed that, outside of globular clusters, there are no giant and larger procedural stars. I don't think I've seen any procedural subgiants either.

Do you mean anywhere in any galaxy, or just anywhere in the Milky Way, or just in the Milky Way's halo? Because there are plenty of giant procedural stars in SpaceEngine, though in the Milky Way they are generated anomalously and only appear in the region of the central bulge, and are also not as dense as in other galaxies. But as far as I can tell this is a problem specific to the Milky Way.





All forum users, please read this!
My SE mods and addons
Phenom II X6 1090T 3.2 GHz, 16 GB DDR3 RAM, GTX 970 3584 MB VRAM
 
smjjamesDate: Tuesday, 04.09.2012, 04:46 | Message # 11
World Builder
Group: Users
United States
Messages: 913
Status: Offline
Quote (HarbingerDawn)
Quote (smjjames)
It may be a limitation of the generation engine, but I've noticed that, outside of globular clusters, there are no giant and larger procedural stars. I don't think I've seen any procedural subgiants either.

Do you mean anywhere in any galaxy, or just anywhere in the Milky Way, or just in the Milky Way's halo? Because there are plenty of giant procedural stars in SpaceEngine, though in the Milky Way they are generated anomalously and only appear in the region of the central bulge, and are also not as dense as in other galaxies. But as far as I can tell this is a problem specific to the Milky Way.


Oh, I don't think I've been to the Milky Way's halo, and I did say that they show up in globular clusters.

Well, I mean in the Milky Way's disk (or that of any other galaxy, though I haven't really explored other galaxies). There are quite a few giants, subgiants, etc in the catalog stars, so I thought there should be some being generated in the disk itself. I don't know how representative the catalog stars really are for typical numbers of subgiants, giants, and supergiant stars in the disk itself, but I'm finding absolutely zero procedural subgiant+ stars (aside from an anomalous binary earlier) in uncharted areas.





 
HarbingerDawnDate: Tuesday, 04.09.2012, 05:02 | Message # 12
Cosmic Curator
Group: Administrators
United States
Messages: 8711
Status: Offline
Quote (smjjames)
but I'm finding absolutely zero procedural subgiant+ stars (aside from an anomalous binary earlier) in uncharted areas.

As I said in my post, there is a bug associated with the Milky Way that generally prevents giants from being generated in the disc/halo, and can only be occasionally found in the bulge. In other galaxies they appear as they should.





All forum users, please read this!
My SE mods and addons
Phenom II X6 1090T 3.2 GHz, 16 GB DDR3 RAM, GTX 970 3584 MB VRAM
 
smjjamesDate: Tuesday, 04.09.2012, 05:06 | Message # 13
World Builder
Group: Users
United States
Messages: 913
Status: Offline
Okay, off to Andromeda then, or maybe one of the Magellanic clouds.

Edit: I think the magellanic clouds may be affected by the same bug as with the Milky Way.







Edited by smjjames - Tuesday, 04.09.2012, 06:02
 
Forum » SpaceEngine » Feedback and Suggestions » Procedural generation elements (A few issues)
Page 1 of 11
Search: