Brainstorm This Perfect turning system - no half-assed solution

[Neon_42]

I wonder if this could be simplified to a best approximation? Since the Inertia tensor is rather complex and is difficult to calculate for complex bodies like ships, even if they are made of blocks, it might be prudent to approximate all ships as boxes defined by their largest dimensions. Then the inertia tensor is the same for every ship. It would then only scale with 4 variables, Length, Width, Height, and Mass. While this would not be completely accurate,but It would be computationally simple and would give the appropriate scaling.

For reference:

Spoiler: Box inertia( about center)

I_i=(M/12)[(d_j)^2 +(d_k)^2]
Where i,j,k are separate dimensions, M is mass, and d is the length in a given dimension.

So, if you want the inertia around the x axis:
I_x = (M/12)[(d_y)^2 +(d_z)^2]

about y:
I_y = (M/12)[(d_z)^2 +(d_x)^2]

The total inertia is calculated on based on 3 equations.

You can further simplify the rotation by always having the point of rotation be about the center of the box.

Spoiler: Box inertia(not about center)

If you chose to not rotate about the center of the box the equations are still rather simple, though you have to calculate where the center is.

I'_i= I_i+M(r_i^2)

where I_x is the inertia about the center of the box and r is the offset along the same axis.

Turning is then simple to calculate.

Spoiler: Turning calculation

a=T/I

a is the acceleration. T is the thrust value assigned to that axis, and I is the inertia calculated above.

Edit: typos

 

[Malacodor]

...it might be prudent to approximate all ships as boxes defined by their largest dimensions.

This is exactly what I don't want, since it favours doomcubes.

 

[Mered4]

I wonder if this could be simplified to a best approximation? Since the Inertia tensor is rather complex and is difficult to calculate for complex bodies like ships, even if they are made of blocks, it might be prudent to approximate all ships as boxes defined by their largest dimensions. Then the inertia tensor is the same for every ship. It would then only scale with 4 variables, Length, Width, Height, and Mass. While this would not be completely accurate,but It would be computationally simple and would give the appropriate scaling.

For reference:

Spoiler: Box inertia( about center)

I_i=(M/12)[(d_j)^2 +(d_k)^2]
Where i,j,k are separate dimensions, M is mass, and d is the length in a given dimension.

So, if you want the inertia around the x axis:
I_x = (M/12)[(d_y)^2 +(d_z)^2]

about y:
I_y = (M/12)[(d_z)^2 +(d_x)^2]

The total inertia is calculated on based on 3 equations.

You can further simplify the rotation by always having the point of rotation be about the center of the box.

Spoiler: Box inertia(not about center)

If you chose to not rotate about the center of the box the equations are still rather simple, though you have to calculate where the center is.

I'_i= I_i+M(r_i^2)

where I_x is the inertia about the center of the box and r is the offset along the same axis.

Turning is then simple to calculate.

Spoiler: Turning calculation

a=T/I

a is the acceleration. T is the thrust value assigned to that axis, and I is the inertia calculated above.

Edit: typos

We already have a turning system based on dimensions. It's terrible for balance, because ships with lots of negative space (*cough* Star Trek *cough*) are unfairly handicapped by an illogical turning system.

 

[Snk]

This has 45 agrees. I feel like pointing this out, because hot damn.

 

[Ithirahad]

This has 45 agrees. I feel like pointing this out, because hot damn.

Considering the state of turning right now, it's understandable that ANY plausible fix/change for turning would gather up a whole ton of support... But yeah, damn.

 

[MacThule]

Simplification: instead of calculating blocks for the matrix, just use the existing mass figure?

I like the idea. I think there are a lot of other higher priorities, but would love to see this.

 

[Neon_42]

We already have a turning system based on dimensions. It's terrible for balance, because ships with lots of negative space (*cough* Star Trek *cough*) are unfairly handicapped by an illogical turning system.

This is blatantly not true. Unfortunately if you want realistic turning then sphere ships will always have the best, followed by boxes. This is basic classical mechanics that any first year engineer or physicist would know. This though assumes a solid object. If you look at the equations you will realize that it depends linearly on mass. so a solid ship of one set of dimensions will be less maneuverable than the ship that has hollow parts. If you look at the equations above a ship of the same size and thrust but about 1/2 the mass( I.E. Star Trek ships) then it will have twice the turning acceleration due to the fact that it is less dense. Once we have directional thrusts, which are planned if I am not mistaken, then this becomes more reasonable as you can push this advantage even further. Larger cubes will be further penalized as the turning rate goes up as twice the square of the length of the cube. Yes we have a system based on dimension currently, but it sucks and was done improperly with hard caps set way to low and change in turn rate that was is way to shallow. A semi-realistic system is the way to go as it is not computationally intensive and will scale properly with ship size.

Simple mass scaling will not work either. Imagine a ship with lots of negative space (I.E. the Enterprise) and one that is a box, but smaller made from exactly the same blocks. The negative space ship has a distinct disadvantage because it presents a much larger target. The advantage the Negatively spaced ship should be able to push is energy production since its dimensions are larger, but in the current state of the game that is difficult.

So to solve this problem power, thrust, and maneuverability need to be looked at in conjunction with each other. Turn rate should depend on mass and dimensions. This is exactly what the OP had suggested, and I suggested a simplified model. Building negative space sacrifices maneuverability for power. Directional thrust can make your ship more maneuverable at the cost of shields and weapons. In this manner there is always a system of trade offs, and no one ship will be best.

Ithirahad , Lecic I have included you as some very basic physics knowledge is a good thing. The inertia tensor article is way more complex than you will ever need, however the concepts are good. If you want any help understanding how this works. As an Astrophysicist I will be very happy to help you.
http://en.wikipedia.org/wiki/Moment_of_inertia#The_inertia_tensor
http://en.wikipedia.org/wiki/List_of_moments_of_inertia

I am quite surprised Malacodor since I have only suggested a simpler model of your earlier approach.

 

[psteiner]

This is blatantly not true. Unfortunately if you want realistic turning then sphere ships will always have the best, followed by boxes. This is basic classical mechanics that any first year engineer or physicist would know. This though assumes a solid object. If you look at the equations you will realize that it depends linearly on mass. so a solid ship of one set of dimensions will be less maneuverable than the ship that has hollow parts. If you look at the equations above a ship of the same size and thrust but about 1/2 the mass( I.E. Star Trek ships) then it will have twice the turning acceleration due to the fact that it is less dense. Once we have directional thrusts, which are planned if I am not mistaken, then this becomes more reasonable as you can push this advantage even further. Larger cubes will be further penalized as the turning rate goes up as twice the square of the length of the cube. Yes we have a system based on dimension currently, but it sucks and was done improperly with hard caps set way to low and change in turn rate that was is way to shallow. A semi-realistic system is the way to go as it is not computationally intensive and will scale properly with ship size.

Ithirahad , Malacodor, Lecic I have included you as some very basic physics knowledge is a good thing. The inertia tensor article is way more complex than you will ever need, however the concepts are good. If you want any help understanding how this works. I will be very happy to help you.
http://en.wikipedia.org/wiki/Moment_of_inertia#The_inertia_tensor
http://en.wikipedia.org/wiki/List_of_moments_of_inertia

I'm still concerned about the accuracy of some... unusual shapes that are made when you ignore physics and forget to connect the ship into one solid mass. The inertia tensor appear to only calculate for shapes in a given level of accuracy defined by a matrix of reference points. How accurate would this be, and if made accurate enough, would that affect performance.

If we go for defining the inertia of the ship in a box, then we are right back at square one in terms of mass distribution to turning performance (I see what you're getting at but who builds a giant hollow ship that needs to turn [ie. deathstar, hiveship, hollow planet, halo ring]). So, to reiterate, this system calculates the moment of inertia for a rectangular prism with a constant density.

And I must say, if thrusters are planned to improve turning, schema might need to hire a physicist to design the turning algorithms.

Note: but if this is implemented, then it will be an improvement, but only a placeholder for an inevitably more advanced system later on.

 

[Neon_42]

To clarify when I say box, I mean rectangular prisms. All rectangular prisms are boxes, but not all boxes are cubes. If you wanted to assume all ships are cubes then the rotation calculation becomes one calculation based on mass and length of one side.

To perfectly describe any object would be computationally difficult and unnecessary. For ships which are millions of essentially points this would be fairly laborious. I work with 4 dimensional data sets of about a million points. when compressed each object is about 128MB. This is a simple array consisting only of numbers. Starmade ships are much more complex containing different blocks in addition to the 3 coordinates that describe its position. Calculating the real world accurate inertia would be no simple matter. The simple way to calculate it is to assume a simple shape. In that way we don't have to worry about off axis terms which add exponential difficulty to the calculation. That is why I suggest you restrict the calculation to that of a box with constant density. Then it is 3 calculations based on the dimensions and the mass of the box instead of 9. If you have to do parallel axis theorem for every block about some point you'd have a 9 item matrix for every block in a ship. So,having an accurate system is not necessary. The scaling will be pretty much the same with a box, and you can tweak the constants in the equations to your liking maybe weighting density more than dimension for example. In terms of ships that don't connect to themselves, they still have to move as one object less they be different ships. Then to a good approximation they are boxes. Should you make ships that don't have some sort of dimensional constraints we can expect ships with very large dimensions and very few blocks, making them extremely hard to hit. A worse version of the death cube, the Swiss cheese cube if you will, with larger holes.

 

[Malacodor]

And I must say, if thrusters are planned to improve turning, schema might need to hire a physicist to design the turning algorithms.

The physics is simple, the problem is that it requires advanced math knowledge. However, creating a 3D engine requires
advanced math as well, so schema should be sufficiently qualified.

 

[Ithirahad]

So to solve this problem power, thrust, and maneuverability need to be looked at in conjunction with each other. Turn rate should depend on mass and dimensions. This is exactly what the OP had suggested, and I suggested a simplified model. Building negative space sacrifices maneuverability for power. Directional thrust can make your ship more maneuverable at the cost of shields and weapons. In this manner there is always a system of trade offs, and no one ship will be best.

No thanks... If we're going to do things this way we might as well keep the current system. This will just force people who want lots of power to have lots of empty space, which is annoying and kind of pointless, and it won't fix the issue the original post is trying to address. That guy over there building the Enterprise doesn't care if he gets to generate more power, his current power grid powers his beams totally fine. He wants better turn rates than the guy with the solid doomcube that's got the same boxdims. Either we need more realistic turn rate calculation that factors in actual blocks or chunks/segments rather than the box dimensions, or we leave things as they are. (Or we just implement purely mass-based turning, which is probably the easiest thing to do right now, but meh.)

Anyway, while I still don't quite get the maths involved, I suspect that perhaps using total masses of chunks rather than trying to count every block as its own mass-point might help things... (However, since this would imply that mass numbers in the calculations have to change rather than just totally removing or adding identical masses, it may make things worse. I have no idea; need to finish this year's math class first and read up on some calculus stuff.)

 

[Mered4]

This is blatantly not true. Unfortunately if you want realistic turning then sphere ships will always have the best, followed by boxes. This is basic classical mechanics that any first year engineer or physicist would know. This though assumes a solid object. If you look at the equations you will realize that it depends linearly on mass. so a solid ship of one set of dimensions will be less maneuverable than the ship that has hollow parts. If you look at the equations above a ship of the same size and thrust but about 1/2 the mass( I.E. Star Trek ships) then it will have twice the turning acceleration due to the fact that it is less dense. Once we have directional thrusts, which are planned if I am not mistaken, then this becomes more reasonable as you can push this advantage even further. Larger cubes will be further penalized as the turning rate goes up as twice the square of the length of the cube. Yes we have a system based on dimension currently, but it sucks and was done improperly with hard caps set way to low and change in turn rate that was is way to shallow. A semi-realistic system is the way to go as it is not computationally intensive and will scale properly with ship size.

Simple mass scaling will not work either. Imagine a ship with lots of negative space (I.E. the Enterprise) and one that is a box, but smaller made from exactly the same blocks. The negative space ship has a distinct disadvantage because it presents a much larger target. The advantage the Negatively spaced ship should be able to push is energy production since its dimensions are larger, but in the current state of the game that is difficult.

So to solve this problem power, thrust, and maneuverability need to be looked at in conjunction with each other. Turn rate should depend on mass and dimensions. This is exactly what the OP had suggested, and I suggested a simplified model. Building negative space sacrifices maneuverability for power. Directional thrust can make your ship more maneuverable at the cost of shields and weapons. In this manner there is always a system of trade offs, and no one ship will be best.

Ithirahad , Lecic I have included you as some very basic physics knowledge is a good thing. The inertia tensor article is way more complex than you will ever need, however the concepts are good. If you want any help understanding how this works. As an Astrophysicist I will be very happy to help you.
http://en.wikipedia.org/wiki/Moment_of_inertia#The_inertia_tensor
http://en.wikipedia.org/wiki/List_of_moments_of_inertia

I am quite surprised Malacodor since I have only suggested a simpler model of your earlier approach.

Is it more realistic?

I'm not sure. The grass is always greener, though.

That said, it would make sense from a purely GAMEPLAY AND BALANCE standpoint to implement a mass-based turning system. It would help to standardize ship classifications (because some people are fans of Dalmont and build skinny, long builds and thus skew their ship classes) across the board, which would also help with balance. TBH, everyone wins.

The only people who don't win are the noob-cubes. Don't be a noob cube.

 

[EtherEel]

Forgive my clearly lower IQ, but isn't all this quandry over turning needless? Are yall are forgetting that these ships are objects in space? The current turning mechanic appears to me to emulate vector change as if your ship were turning in fluid (air, water) and subject to gravity (the earth) -- in essence, as if your space-ship were an airplane or submarine.

Haven't we observed via our primitive space-crafts that Inertia in space is essentially irrelevant? That a mass (no matter how large) needs only a minimal, short burst of thrust to set it moving in a given direction indefinitely? That to create spin of said mass one only need a synchronized equal thrust in the opposing direction? Thus, even the notion that we must continually hold-down a button to invoke our thrusters to move forward violates actual physics.

Actually, the Star-Trek consultants back in the 1960's got this right by making the distinction between Impulse DRIVE and "thrusters" -- thrusters being small, short bursts of propellant used for positional refinement and slow movements, which then could be easily overcome by counter-thrust. (This is one reason why the Star-Trek universe generally had no use for "fighter"-sized craft -- the big ships are just as nimble and can support larger, more effective systems.)

Therefore, turning even an extremely massive object should involve minimal effort. Say you're battle-cruiser is being chased by another big craft, you should be able to spin about and face your pursuer (and fire at it) while still maintaining your previous direction and speed (only now you're moving "backward"...)

I'm probably overlooking some finer points of various aspects of advanced physics, but in my experience "simpler is better" are words to live by. And, or course, using real physics will feel strange when our experience tells us that objects (because we and they are bound to the earth) "should" act a certain way. (Thus, you need to hold your "Forward Thrust" key to maintain speed...)

 

[Neon_42]

Forgive my clearly lower IQ, but isn't all this quandry over turning needless? Are yall are forgetting that these ships are objects in space? The current turning mechanic appears to me to emulate vector change as if your ship were turning in fluid (air, water) and subject to gravity (the earth) -- in essence, as if your space-ship were an airplane or submarine.

Haven't we observed via our primitive space-crafts that Inertia in space is essentially irrelevant? That a mass (no matter how large) needs only a minimal, short burst of thrust to set it moving in a given direction indefinitely? That to create spin of said mass one only need a synchronized equal thrust in the opposing direction? Thus, even the notion that we must continually hold-down a button to invoke our thrusters to move forward violates actual physics.

Actually, the Star-Trek consultants back in the 1960's got this right by making the distinction between Impulse DRIVE and "thrusters" -- thrusters being small, short bursts of propellant used for positional refinement and slow movements, which then could be easily overcome by counter-thrust. (This is one reason why the Star-Trek universe generally had no use for "fighter"-sized craft -- the big ships are just as nimble and can support larger, more effective systems.)

Therefore, turning even an extremely massive object should involve minimal effort. Say you're battle-cruiser is being chased by another big craft, you should be able to spin about and face your pursuer (and fire at it) while still maintaining your previous direction and speed (only now you're moving "backward"...)

I'm probably overlooking some finer points of various aspects of advanced physics, but in my experience "simpler is better" are words to live by. And, or course, using real physics will feel strange when our experience tells us that objects (because we and they are bound to the earth) "should" act a certain way. (Thus, you need to hold your "Forward Thrust" key to maintain speed...)

To put it simply, No. Inertia matters in space. Being in space does not magically alleviate all the problems of having mass. In our primitive space craft inertia is one of the most complicated and important aspects of spaceflight. There are entire companies who own the rights to specific inertia tensors that predict how a spacecraft will move.

 

[EtherEel]

Well, I DID say my IQ was lower.... ~.o
[DOUBLEPOST=1426111517,1426111034][/DOUBLEPOST]Perhaps I should add, however, that in similar vein to "holding your forward thrust key", TOO MUCH slavishness to advanced physics is a detriment to fun. Simpler really is better.

 

[Neon_42]

No thanks... If we're going to do things this way we might as well keep the current system. This will just force people who want lots of power to have lots of empty space, which is annoying and kind of pointless, and it won't fix the issue the original post is trying to address. That guy over there building the Enterprise doesn't care if he gets to generate more power, his current power grid powers his beams totally fine. He wants better turn rates than the guy with the solid doomcube that's got the same boxdims. Either we need more realistic turn rate calculation that factors in actual blocks or chunks/segments rather than the box dimensions, or we leave things as they are. (Or we just implement purely mass-based turning, which is probably the easiest thing to do right now, but meh.)

Anyway, while I still don't quite get the maths involved, I suspect that perhaps using total masses of chunks rather than trying to count every block as its own mass-point might help things... (However, since this would imply that mass numbers in the calculations have to change rather than just totally removing or adding identical masses, it may make things worse. I have no idea; need to finish this year's math class first and read up on some calculus stuff.)

The current system is an obvious undersiplification of the problem. You all seem to think that including dimensionality is a bad thing. This cant be further from the truth. If you want realistic turning then you need to include the dimensions of an object. This is basic physics. A solely mass based system doesn't work either. as for the swiss cheese cubes as I called them. To quote the OP:

The Advantages:
- It's real world physics and will thus result in a realistic feeling when maneuvering ships.
- No artificial limitations, which would add to the game's complicatedness (even the simple system we have now isn't very intuitive).
- One system for both ships and turrets.
- It only punishes minimally for decorative ship parts, which increase a ship's box dimensions, but have low mass (e.g. antennas) - unlike other systems (like the current one) which work on box dims.
- Ships with hollow spaces inside (hangars, interior) turn faster than completely stuffed ships of the same size.
- Due to limiting angular acceleration rather than angular velocity (= turn rate) big ships and turrets still can't react fast enough to agile fighters' evasive maneuvers, while on the other hand it doesn't take ages to aim at a target in a completely different direction.
- Isn't a half-assed solution, "uses inertia tensors" looks cool in the game discription, and this post, which took me hours, wasn't in vain.

It should be clear that calculating the real inertia tensor for a ship will be overly complicated(look at the wiki page), THUS A SIMPLIFICATION IS NECESSARY. If you simplify all ships to a box this works. More realistic turning than what we have with the appropriate scaling. Yes it is more realisitic Mered4 . It's amazing how many people claim to know more about my field than I do.

In terms of BALANCE and GAMEPLAY it should be clear that solely a mass based system is a more of a boost to the doom cubes, and they need to be balanced some other way.

 

[Ithirahad]

I'm probably overlooking some finer points of various aspects of advanced physics, but in my experience "simpler is better" are words to live by. And, or course, using real physics will feel strange when our experience tells us that objects (because we and they are bound to the earth) "should" act a certain way. (Thus, you need to hold your "Forward Thrust" key to maintain speed...)

Uh, for me, having to continually fire forward thrusters to maintain speed feels awkward and ridiculous when I'm not braking. I kind of expect my object in motion to remain in motion until I decide to apply another force to it, whether that be my autobraking or manual direction-of-motion burning. (The only thing we need to make the already-present option to turn off the weird space drag more viable is autobraking when there's nobody aboard, or BOBBY AI is activated with no enemy targets nearby, but that's a subject for another topic.)

 

[Keptick]

Using chunks as discrete masses in the inertia calculation would probably be the best way to go. It wouldn't be nearly as computationally intensive as calculating the inertia of every single block (instead using chunks as massive 'blocks') and would also take into account negative space up to a certain degree of accuracy.

The total mass of a chunk would be calculated and the center of said chunk would be the location of the mass point. So basically, an empty chunk would rate as 0 mass, a filled one for 409.6 and then there's every value in between. Smaller ships wouldn't really be affected by the rather low accuracy (16x16x16 chunks) as they already have good turning speeds. A system like that would require something like 4000 times less points to be calculated and would still retain decent accuracy.

I don't code btw, so this is all just speculation. I am fairly familiar with the concepts though.

 

[spacie]

Using chunks as discrete masses in the inertia calculation would probably be the best way to go. It wouldn't be nearly as computationally intensive as calculating the inertia of every single block (instead using chunks as massive 'blocks') and would also take into account negative space up to a certain degree of accuracy.

The total mass of a chunk would be calculated and the center of said chunk would be the location of the mass point. So basically, an empty chunk would rate as 0 mass, a filled one for 409.6 and then there's every value in between. Smaller ships wouldn't really be affected by the rather low accuracy (16x16x16 chunks) as they already have good turning speeds. A system like that would require something like 4000 times less points to be calculated and would still retain decent accuracy.

I don't code btw, so this is all just speculation. I am fairly familiar with the concepts though.

This. I have been thinking about this for a while now and this seems like the only viable and reasonably realistic solution. Using every individual block isn't doable since you'd have to recalculate inertia abd centre of mass for every block added or subtracted. Which is especially nasty for the subtraction part. Using a shipsized block is still quite unrealistic.
A few additions to the above post. I would suggest calculating the moments of inertia of each chunk as a cube of variable size but constant density, since moments of inertia scale linearly with mass but quadratic with size. Only changing the mass would have a 'dampened' effect.

Also using the full inertia tensor would cause some *interesting* things to happen if you think of user inputs as torques on the ship. Asymmetric ships would start start rotating about a different axus than the one you apply the torque to for example.

I'd say only use the x y and z moments of inertia and their 2d rotation equations. If you use the 3d set of equations you also have to take the rotational velocity about all other axes into consideration. This may give a more realistic feeling but is also most likely harder to implement.

Tl;dr: use chunkbased moments of inertia and center of mass calculation, calculate individual chunk values with mix of inferred size and mass of a cube, use only the principle momenta of inertia.

 

[Ithirahad]

Using chunks as discrete masses in the inertia calculation would probably be the best way to go. It wouldn't be nearly as computationally intensive as calculating the inertia of every single block (instead using chunks as massive 'blocks') and would also take into account negative space up to a certain degree of accuracy.

The total mass of a chunk would be calculated and the center of said chunk would be the location of the mass point. So basically, an empty chunk would rate as 0 mass, a filled one for 409.6 and then there's every value in between. Smaller ships wouldn't really be affected by the rather low accuracy (16x16x16 chunks) as they already have good turning speeds. A system like that would require something like 4000 times less points to be calculated and would still retain decent accuracy.

I don't code btw, so this is all just speculation. I am fairly familiar with the concepts though.

Yaay, I was right! ...I think? This makes sense to me, but I'm no mathematician (currently) so I didn't want to sound too sure of myself.