Mass-based turning system

[Keptick]

The best solution is actually to just apply real life angular dynamics, which take dimensions AND mass distribution into account...

 

[aceface]

which is what it already does, mass does affect the turning rate to, but not at all as much as before the dimension option was implemented

 

[NeonSturm]

With a ship >2km minimum radius, you can make a distance-holder like:
H . . . H . . . H . . . H
. . T . . . . . . . T . .
. . . . . T . . . . . . .
. . . . . . . . . . . . .
H . . . H . . . H . . . H
. . . . . . T . . . . . .
. . . T . . . . . . . . .
. . . . . . . . . T . . .
H . . . H . . . H . . . H
. . . . . . . . . . . . .
. T . . . . T . . . . T .
. . . . . . . . . . . . .
H . . . H . . . H . . . H
And use free floating drone-sized sniper turrets, as evasion with passive push is a viable strategy again ((as your core is just out of range))

in just what way does this even remotely relate to the thread...

If box size does not matter anymore... :p

---

Imagine a cube:
-- -- -- -- y: -- -- --| y blocks
-- -- -- -- +3 -- -- --| 3
-- -- -- -- +2 -- -- --| 7
-- -- -- -- +1 -- -- --| 9
x: -3 -2 -1 00 +1 +2 +3| 9
-- -- -- -- -1 -- -- --| 9
-- -- -- -- -2 -- -- --| 7
-- -- -- -- -3 -- -- --| 3
–––––––––––––––––––––––––-––––––––––––––––––––
x: .3 .7 .9 .9 .9 .7 .3 mass in yz plane

You could reduce the size of all blocks to just 1 number per plane.
You have with + height + length planes (yes, + not *)

But to find the Center of Mass (CoM) (Example plane-weights: 3799973) you have to calculate:

aOriginal[] = 3,7,9,9,9,7,3 given (modified on block addition/loss)

loop all "i"s
> aFind1CoM = aOriginal + (aFindCoM[i-1] defaults to 0)
> aFind1CoMReverse = aOriginal + (aFindCoM[i+1] defaults to 0)
> aFind2CoM = aFind1CoM + (aFind2CoM[i-1] defaults to 0)
> aFind2CoMReverse = aFind1Com + (aFind2CoM[i+1] defaults to 0)


What the loop result is:
> the closer to go away from the very rear or front, the heavier this plane is for rotating.
> aFind1... fields contain the sum of plane-blocks from itself and all previous.
> aFind2... fields goes one step further and contains the sum of it and all previous plane blocks + the sum of previous which also contains the sum of previous.
> Thus aFind2 #2 adds #1 one time. aFind2 #3 adds #1 two times and #2 one time. Basically weight * distance without multiplications AND intermediate results in aFind1CoM

The CoM is where aFind2CoM and aFind2CoMReverse is equal.


​

Performance

You have to do 2 additions * axis-planes count * axis count and access the array values.

Each time it's modified, everything has to get calculated again.

Maybe you can save some calculations by only updating it each second or at 10% blocks lost in a significant (outward) plane.
Maybe you can save some calculations by calculating more details on outermost parts.​

It is still a huge task each time significant amount gets modified.

Especially since you have to update 3 plane weights per update-block-group-plane: each occupied {xz plane, xy plane, yz plane).​

​

Why you should read the math above

Maybe this is not the optimal solution for updating CoM (even though it is pretty fast calculating it from scrap) but it definitively can give you some insight into the complexity of CoM calculations.

I still try to learn more about it - where are the good manuals when you need them?​

 

[Commodore]

I already explained why I'd prefer mass only based turning, but if calculating the center of mass and moment of inertia is the goal, the way I would do it is count the number of blocks (mass) in each chunk (16*16*16 area I believe) and treat them as a point mass in the center of the chunk. Then the center of mass is just the weighted average of the all the chunks, and the moment of inertia on each axis would be found using I = mass * distance^2 (perpendicular to the axis).

Calculating the exact values would involve the same calculations on every single block, which is not practical.

EDIT: because I stuffed up formula, also, as @ltmauve said, both these calculations need to be done on X, Y and Z axis.

 

[ltmauve]

Calculating the exact values would involve the same calculations on every single block, which is not practical.

Actually, you need six new operations per block. For turning, you need the mass times the distance squared in each plane. For center or mass calculations, you need to multiply the mass by each coordinate and add it to the sum. Then you get the average by dividing it by the sum.

 

[RedDragonflyxx]

Makacodor explained my physics based system well, he posted a link on the first page. I like this system better but it just doesn't work as good as doing the actual math which wouldn't be too intensive unless the code was completely devoid of optimization.
Edit: Malacodor

 

[Proxima_Centauri]

...but turning rate IRL is determined by a ship's moment of inertia, so the current system is more accurate

and borg cubes aren't used because they turn faster, they're just easier/quicker to build for griefers

 

[Comr4de]

I'm going to provide a veteran's view on the whole situation.


The dimensional turning system was introduced a while back to help ships turn faster; when mass was the primary dictating factor in the turning formula, it became VERY easy to get a ship that couldn't turn for shit.


I think schema should lower the minimum turn rate by default to something around 0.5. But that's just me.

 

[Commodore]

I'm going to provide a veteran's view on the whole situation.

The dimensional turning system was introduced a while back to help ships turn faster; when mass was the primary dictating factor in the turning formula, it became VERY easy to get a ship that couldn't turn for shit.

I think schema should lower the minimum turn rate by default to something around 0.5. But that's just me.

That's kind of the effect I was going for with my logarithmic formula, but rather than hitting a minimum and having a 1 million block ship turn at the same rate as a 10 million block ship, the 10 million block ship would turn ~20% slower, giving the smaller ship a much needed advantage. The difference between a fighter and a titan would be smaller though.

 

[BeefBacon]

I'd like to see the rotational axis be moved from the ship's core and to the ship's center of mass. Having the core right at the back or right at the front on a large ship looks pretty weird when it turns.

 

[Planr]

i'm not math-savvy so i'll let you guys handle the equations and stuff. honestly, i can't even begin to understand all that math stuff you guys are doing in the thread...

 

[Keptick]

i'm not math-savvy so i'll let you guys handle the equations and stuff. honestly, i can't even begin to understand all that math stuff you guys are doing in the thread...

Mechanics course #1 :P

 

[RedDragonflyxx]

i'm not math-savvy so i'll let you guys handle the equations and stuff. honestly, i can't even begin to understand all that math stuff you guys are doing in the thread...

Ok, I meant that it takes a certain amount of force to move a certain weight (mass) and was talking about calculating the mass that was being moved when you turn the ship one way; then you take that and compare it to the amount of energy going into rotating the ship. I'm not very literate in physics so I'm not sure as to specific values on what is being moved and how much force is needed to move it.