Are KSP's physics, at this stage, a good approximation of this universe's physics?

[comham]

Ok a legitimate question for the hardcore physics and astrophysics guys--do you think that KSP cannot demonstrate rigid body dynamics and/or n-body problems at its current state right now, even at the most rudimentary level?

I thought rigid body dynamics was basically how crafts work in KSP. Undeformable parts with collision meshes, and they either bounce of each other or explode. Attached at a point with a certain amount of flex and a breaking point.

[rkman]

It is not that complex. For example it is implemented in Orbiter. There is no universal analytical solution to the problem, but integration gives you VERY accurate results.

Probably not at high timewarp, with multiple missions in progress.

[asmi]

Probably not at high timewarp, with multiple missions in progress.

Yes, even in high timewarp with multiple missions in progress.

[solomon]

It's about the same, except that you cannot be influenced by more than one source of gravity at a time, making Lagrangian points impossible.

[asmi]

It's about the same, except that you cannot be influenced by more than one source of gravity at a time, making Lagrangian points impossible.

Lagrange point is not the only thing we're missing. Current implementation also precludes implementing cool things like multi-star systems (if we ever get another star systems in KSP).

[paleorob]

Another point - either the scientific instruments are broken OR readings derived from them indicate physical processes that are very much unlike what we have in our universe operating in KSP.

[rodion_herrera]

Another point - either the scientific instruments are broken OR readings derived from them indicate physical processes that are very much unlike what we have in our universe operating in KSP.

Please cite a specific example? Re temp/baro readings?

[Jokurr]

As far as temperature goes it is only a function or altitude for each of the planets, though I believe this is a placeholder for the moment and a more complex model will be introduced eventually. The temperature on Kerbin at the equator and the pole is the same for a given altitude.

[rodion_herrera]

As far as temperature goes it is only a function or altitude for each of the planets, though I believe this is a placeholder for the moment and a more complex model will be introduced eventually. The temperature on Kerbin at the equator and the pole is the same for a given altitude.

I didn't know that! Thanks to the intrepid explorers for actually checking that out. And thanks to you that I know about it now

[Moon Goddess]

Another point - either the scientific instruments are broken OR readings derived from them indicate physical processes that are very much unlike what we have in our universe operating in KSP.

Haven't really payed much attention to the values from any of them, but I do know the devs have stated that temperature is very broken, when on a planet with atmo it just changes by altitude not paying any attention to latitude or other factors.

[Aphox]

Popping this over to the Science Labs.

For science.

Edited May 30, 2013 by Aphox

[sporkafife]

The patched conic approximation works great in a two body system where long as the mass of one of the bodies is completely negligible compared to the planet (i.e. just a ship orbiting a planet). So if you just have a ship orbiting a planet, the way KSP models the forces is almost perfect. It doesn't model relativistic effects but as of right now relativistic speeds are impossible to reach in KSP without cheats anyway.

But the patched conics system only ever models the gravitational influence of a single body on your ship, which while it is an okay approximation it cuts out the ability to use lagrange points to your advantage for low delta-v interplanetary missions. The fact that the orbits of the planets are fixed, and the moons simply orbit the planets as if they were massless instead of both bodies orbiting the common barycenter is a bit more of a fudge that isn't very accurate, but it's an approach that is close enough for the scope of the game as it is right now.

[quantumpion]

In addition to the lack of Lagrange points, another unrealistic aspect to the patched conics approximation is the stability of orbits. In reality, if you were not at a Lagrange point then your orbit would not be stable. For example you couldn't be in a stable kerbolcentric orbit just outside the SOI of kerbin, kerbin would pull you back in eventually. For example, the Apollo 12 third stage recently returned to Earth orbit (and was mistaken for an asteroid initially): http://neo.jpl.nasa.gov/news/news134.html

[The Jedi Master]

I'm sure this has already been mentioned, but if aerodynamics count as physics, then they are VERY different.

But they're trying to fix that.

[CobraA1]

Yes, even in high timewarp with multiple missions in progress.

N-body systems tend to "drift" out of what would actually be their orbits due to numerical rounding, and there's no really good way to correct for that. At extreme levels of time warp, this can cause systems to fly apart.

Two body systems (ie, conics) is the only one that has been solved completely by mathematicians, and thus is the only one where you can completely predict the position at any arbitrary point in time.

With three or more bodies, you generally have to perform actual simulation of the orbits in real time, and can't completely predict the position at any arbitrary point in time. This means that errors tend to accumulate over time, leading to instability, especially at high levels of time warp.

I should note that this has been discussed before many times, and as a result KSP is unlikely to change. It will likely always be based on conics, unless there's a breakthrough in current mathematics involving three-body systems.

EDIT: This is why:

B10. Orbit deteriorates at high time acceleration.

Orbiter uses numerical integration methods for propagating spacecraft states from one simulation frame to the next. The accuracy of this method depends on the simulation time interval between frames. While Orbiter's numerical implementation is quite sophisticated (including high-order Runge-Kutta and symplectic integrators, and subsampling of time steps), it may still lead to problems when extreme time accelerations are used during high-acceleration phases of a spacecraft (e.g. in low orbit).

It is generally a good idea to avoid high time accelerations when in low orbit, and in particular during powered flight. Orbiter also provides an "orbit stabilisation" mechanism that disables the numerical state integration in critical phases. Orbit stabilisation can be configured in the "Extra" tab of the Orbiter Launchpad.

The authors of KSP decided that for their game, this type of orbital deterioration is unacceptable. The goal of KSP is to be a game first and foremost, and gamers will expect not to have these types of bugs. The goal of Orbiter is to be a realistic simulation, and to some degree these types of bugs are acceptable if it's still an overall more realistic simulation.

Edited May 30, 2013 by CobraA1
orbiter Q&A

[EatThePath]

There is a recent thread on n-body physics:

http://forum.kerbalspaceprogram.com/showthread.php/29938-A-question-on-n-body-physics?highlight=lagrange

Developers said (in a lost post I guess) that n-body physics would not be implemented because it raises problems with time warp.

This can be easily understood. Patched Conics Approximation (current KSP model) allows exact predictions. It means that you can easily know where your ship will be any time in the future. Real n-body modelling does not allow exact prediction, you can only make approximations.

Consequently, the behaviour of your ship will depend on the time warp mode. To know where you ship will be at the next frame, in normal speed, you compute the position of your ship a fraction of a second later. In maximum time warp, you compute the position of your ship (let's say) one minute later. The cumulative error will be much higher in the second case after the same amount of (game) time, whatever prediction algorithme you use.

In terms of gameplay, knowing that the time acceleration changes the trajectory is difficult to accept. I don't know how Orbiter is handling that problem.

This is actually the problem that was in my mind when designing the core structure of Atomic Space Navy/Tools/Race. As a consequence of that effort I think a version of KSP with a proper n-body simulation and predictable physics out to a useful duration is possible, but it's something that you'd have to build the whole game around, not something that could reasonably be patched in at this stage.

To the original question, it all depends on what you're using it for. Broadly I'd say it's pretty good, but it breaks down in certain areas, and while the mechanics are pretty good the properties of things are rather wildly unrealistic.

[SuperFastJellyfish]

The most obvious I can think of is the lack of proper drag model. It can be seen as a technical issue, and it has been clearly stated that it will change in the future.

However, I guess it is also a gameplay issue. With a vaguely realistic drag model, the silly, fat, cathedral-like rockets that you can send now in orbit would become useless. Basically, rockets would have to look like rockets, and I am not sure most people would like that. I really wonder how squad will handle that...

I don't see how the 'silly, fat, cathedral-like rockets' would be hindered with a better drag model. It's essentially like multiple rockets flying in formation. I'm no aerodynamics expert or anything, but it seems feasible to me.

[Jokurr]

Well firstly you won’t be seeing any rockets flying around without a nose cone like you currently see so often in the game, that creates a huge amount of drag. Depending on how complex the drag model Squad implements is, flying several rockets close together is also a very inefficient idea. You now have multiple objects flying in each other’s boundary lairs, where all sorts of drag can occur. When compressible flow is introduced the situation can become very nasty. This is why rockets such as the Soyuz that have stages in parallel with the main engine, sort of “stem out†from the main rocket body, if you know what I mean. The space shuttle was not like this, but that wasn’t really the best design for that matter.