Celestia discovers Newton!

[diegorodriguez]

Kepler is Ok, but when you need high precision, you can not confine yourself to the usual True Anomaly based calculations, especially with high eccentric orbits, namely many comets or small bodies.

I just played a bit with http://www.astrograv.co.uk/, a nice commercial product, but I do not see why can't we just plug a physical simulation engine into the already superior visualization capabilities of Celestia.

That probably would mean fetching rectangular heliocentric coordinates from a force calculator, instead of fetching them from polar Distance and True Anomaly coordinates. The force calculator could be easily decoupled and updated separately from the main Celestia body, as a classical server side process.

I am sure this wish has been heard on here many, many times.

But maybe it's time for change...

[rthorvald]

Why is this post a sticky?

- rthorvald

[selden]

I've removed the sticky bit. Discussions of math, etc probably aren't so urgent that they need to be at the top of the topic list forever.

[selden]

Celestia does not calculate gravitational effects because they are much too compute intensive.
It needs to run in realtime with the fastest possible framerate on older, slower systems.
Gravitational calculations, even if only for every visible body, simply cannot be completed in time.
As a result, it uses precalculated ephemerides (VSOP87 by default) when it needs accuracy or Keplerian ellipses when it doesn't. They both are much faster.

[diegorodriguez]

Thank you about dropping the sticky thing. I didn't mean to check it.

But don't drop the gravitational bit so fast!! We all are quite aware of the increased computational time, it's something we understand, but by no means make them unafordable for modern processors. Using a simple square matrix for 100 solar system objects running at one day steps doesn't cause any trouble.

I don't understand your 'real time' objection. You mean generating over 24 days per second?

[chris]

Thank you about dropping the sticky thing. I didn't mean to check it.

But don't drop the gravitational bit so fast!! We all are quite aware of the increased computational time, it's something we understand, but by no means make them unafordable for modern processors. Using a simple square matrix for 100 solar system objects running at one day steps doesn't cause any trouble.

I don't understand your 'real time' objection. You mean generating over 24 days per second?

The problem is that when incorporating gravitational effects, you lose 'random access' in time. What happens to an object in low Earth orbit (where one day steps are not an option at all) when I set the time rate to one million times normal? Or if I set the time to 2000 years in the past? Should Celestia just pause for a minute while it integrates the trajectory? The omission of gravitational modeling in Celestia is deliberate; incorporating them would affect the usage of the application in a very fundamental way. Also, there's so much to be done in the visualization domain that it hardly leaves time to implement gravitational effects.

--Chris

[t00fri]

...
But don't drop the gravitational bit so fast!! We all are quite aware of the increased computational time, it's something we understand, but by no means make them unafordable for modern processors. Using a simple square matrix for 100 solar system objects running at one day steps doesn't cause any trouble.

I don't understand your 'real time' objection. You mean generating over 24 days per second?

Calculations of time dependent gravitational effects that can REALLY be trusted are VERY complex.

Therefore, I doubt that many of the "voices" you represent really appreciate this fact! As a Senior Scientist in theoretical Particle Physics and Cosmology I know what I am talking about ...(unless MANY others, apparently).

Of course, to just display something on the canvas, one can always take some unrealistic limit of distant point sources of gravitational attraction. But that limit becomes quickly invalid. Notably, users have NO idea when it badly beaks down... And don't forget a basic design highlight of Celestia is to allow users to travel seamlessly from their backyard to the farthest galaxies...

Such a naive approach of time dependent gravitation is typically what is implemented at the level of space games...As a matter of principle, for Celestia we do not deal with such bad approximations, whose limitations are "open end" ...

Imagine for instance to navigate with your massive starship in the Iovian system, in-between it's moons. Already the substantially non-spherical shape of Jupiter makes the task VERY complicated. Radial symmetry is lost! There is no other solution but numerically integrating which leads to non-localities as indicated by Chris already...

If you have any solid PHYSICS arguments of which you think they can surmount the mentioned problems, I will be able to understand them. Just go ahead... I am listening.

Fridger

[diegorodriguez]

The gravitational model has had a place in many space simulators for years.

Highly successful Orbitersim is entirely based on a gravitational model. The funny thing is that they refuse the OpenGL part which is central to Celestia, while Celestia refuses the gravitational model absolutely critical to Orbiter's mission. Well, the last news is that a bunch of volonteers plan to add an OpenGL frontend and they have already released a beta...

But, well, You can take a look at Orbiter's DirectX Solar System simulation right now and here, or for a more simple approach, Astrograv, if you still think gravitational models do not work on personal computer simulations.

[t00fri]

The gravitational model has had a place in many space simulators for years.

Highly successful Orbitersim is entirely based on a gravitational model. The funny thing is that they refuse the OpenGL part which is central to Celestia, while Celestia refuses the gravitational model absolutely critical to Orbiter's mission. Well, the last news is that a bunch of volonteers plan to add an OpenGL frontend and they have already released a beta...

But, well, You can take a look at Orbiter's DirectX Solar System simulation right now and here, or for a more simple approach, Astrograv, if you still think gravitational models do not work on personal computer simulations.

As I wrote, is is fairly trivial to implement SOME time-dependent gravitational interactions, but they are bound to be badly WRONG in many configurations. This surely holds for Orbiter, too.

One undesirable consequence is that users will not be able to judge for themselves WHEN things are getting wrong. For us it is absolutely no reason to implement something that is largely incorrect, simply because space games and other software claim to have such features. We prefer NOT to implement some physical aspect altogether, before we do it INCORRECTLY.

What we emphasized is that a CORRECT implementation would strongly interfere with Celestia's main design strategy, also because of the resulting complexity.

I am ready to discuss these issues in PHYSICS terms either with you (if you are competent in this field) , or with anybody else who feels competent, e.g. from Orbiter The design goals of Orbiter are quite different from those of Celestia, although both simulations may superficially seem to address similar tasks...

Fridger

[diegorodriguez]

Random access time: the ability to specify a given date to start the simulation. You can skip there through usual Kepler to get the dynamical vectors and start from that point on using the heavier gravitational engine.

?Time frame? For virtually all other than artificial satellites, one day is fair enough. If you focus on a given craft, something Celestia is not well fitted to, you can always step down to minutes or seconds.

Let's be realistics, you can not go much farther in space or time with Celestia, either. VSOP87 says it ensures a decent precision for a few thousand years within the solar system..

The gravitational benefits for Celestia would come, funny enough, for one of its major appeals: ficticious solar systems, poorly known binary stars systems, etc.

[selden]

Here is my understanding of the situation:

The Celestia development team has no plans to implement gravity.
Celestia can display arbitrarily accurate orbits when using JPL DEnnn ephemerides, SPICE kernels or xyz trajectories downloaded from Horizons. It is not limited to using VSOP87.

If you feel strongly enough about imlementing gravity, please feel free to implement the code in Celestia.

If the results are mathematically correct, are a selectable runtime option, well documented, and interact cleanly with Celestia's existing code base, I suspect that there's a very good chance it might be seriously considered for inclusion in the official code. If it is crude and intended only for fictional systems, then it will not be considered.

Don't forget that Celestia also can use what it calls a "ScriptedOrbit" for any object which specifies it. It calls a Lua routine which can place the associated object anywhere. This might be a reasonable way for initial testing. Of course, Lua probably would be too slow for this to be used for many objects simultaneously.

[diegorodriguez]

The solution you point out looks good.

if you want seamless navigation, the gravitational model can be plugged into the trajectory calculator as an option for certain scenarios, so you really wouldn't notice which engine is on. The only major setback is this trip is one way only...

For instance, jump to 01.01.2009 deterministically and launch a 1-day step gravitational simulation for Jovian satellites. By the time you stop the simulation, let's say 31.01.2009, the final position and velocities wouldn't match those obtained if you jump directly to that date using Kepler. So, whether you can start with Kepler and go on with Newton, then you will no be able to continue once again with Kepler...

[jogad]

Hello

Please mercy for the simple guy like me who just want to visit the Universe to see how beautiful it is. After loading Celestia, I take a glance at the guide and I am ready to begin the journey. With Orbiter I must study at least for a week to be barely able to leave the Earth!
And with no assurance to go back safely! Gravity effects are perfect if you want precisely study them or test your skill in driving spaceships. It’s good that the two possibility exist but in two different programs. The beauty of Celestia is (among other things) its simplicity. Thus, no gravity!

Jogad

[diegorodriguez]

Hahaha... well pointed out... I liked the way you put it.

You are right, Orbitersim is definitely intended for another kind of users, who may share or not their interest with Celestia. Orbitersim is just an extreme example, and I don't think anyone would like them to merge.

But even when gravitational engines can be really complex inside, just like the semi anaylitical solutions Celestia uses, a frontend will be EXACTLY as simple as before, since the only difference takes place deep inside the orbital and position calculation routines.

Think of that like a hybrid car: electrical or gas powered depending on the situation, and you don't realy care about. It's just that by combining both, your car can do things it was not able before.