Atmospheric flight model 2.0

[selbie]

Also wondering if a solution to balancing could be adding heat that can destroy a ship during reentry. Personally I would love this realistic, challenging aspect of reentry, while overlooking some of the more mathematically challenging atmospheric flight mechanics.

I’m wondering the same. I’m all for a mix of BVR combat and guns dogfighting so I think in terms of gameplay making the planetside combat speeds determined by the atmospheric pressure (ie. related to altitude) would help keep players confined to a particular hotspot lower down. If they want to escape or draw someone away, they have to increase altitude…thereby exposing them to intercept from above. Flying faster than the max dynamic pressure at the wrong altitude will get you cooked or cause Rapid Unplanned Disassembly

[Jamin62]

I’d like to throw in my opinion in favour of boosting the lift characteristics of the 3 smaller ships, with the goal of encouraging the bank/roll and pitch technique for atmospheric flight which added so much depth to the early prototype videos (canyon race track?!).
I’m 100% pro realism, but I don’t think it’s necessarily unrealistic just… optimistic, in terms of future starship aerodynamic design (the ‘Shield envelope aerofoil’ idea mentioned earlier would make a good in-lore excuse for this). I would love to see this fantastic physics system Flavien has created as a force to encourage players to use and engage with physics to gain an edge in dogfights- it should feel like atmospheric lift gives you a BOOST to maneuverability in addition to drag being a hinderance to speed- therefore adding depth and variation to the strategies required to win in different environments

[Zen]

@Flavien - no worries, we understand you have been busy. You and all the other DEV’s hard work is greatly appreciated. But I would like to weigh in on the joystick/mouse comment you made. The stick will never be able to out aim the mouse but where the stick can compensate is maneuverability. So instead of worrying about aiming mechanics maybe you should consider balancing speed. In short the slower the ship the greater the advantage to mouse users, as where faster flight allows stick users to turn and engage quicker than mouse users. Anyway, my 5 yen’s worth.

[nanit]

That was a very interesting blog post Flavien!
So interesting, that I have spend some hours to research lift and drag for myself. While I am not allowed to test the flight models myself, I thought I share here some of my findings and ideas to the science:

One thing I still haven’t solved are torque ( rotational accelerations ). Without introducing an arbitrary factor in the equation, I found that the amount of torque is very high. In fact, it makes a sky-diving human quickly lose its stability and start to rotate in all directions. I do not know if I’m missing something in the physics or if I can simply blame it on the method ( a low amount of ray casts means every frame, the flow vectors hit the object at random positions. It’s no problem for the linear acceleration, but this randomization in the surface positions causes a random amount of torque )

If you only want to reduce the randomness of the positions, you could experiment with symmetry. For example you could mirror the origin of your rays over the middle point of the flow (yellow arrow).

That being sad you can definitely blame the method. These random air molecules are a nice way to simulate chaotic turbulences at high velocities but they shouldn’t be as good to simulate normal torque because of drag and lift. To change that, you could try to define an “Center of Pressure” (cp) where all your drag and lift forces act upon. If you put the cp behind the Center of Mass the ship will allways try to put the nose in the wind, if you put the cp in front of the Center of Mass, the ships will become unstable at high drag. Do you also somehow decelerate ship rotation in atmosphere because of drag and friction?

So, what the algorithm does for each ray is to cast a ray, determine if it hits that box, and if it does, generate a feedback force along the normal of the box surface. This will naturally generate drag and lift.

Even If your new atmospheric flight model 2.0 is not very realistic, it is elegant. It easily measure the area facing the flow and even considers how “sharp” the shape facing the flow is, by using the normals of the box surface.

The most interesting part of this new model is that I was able to get ride of the arbitrary coefficients of the old model.

I am afraid if you can’t get rid of these arbitrary coefficients if you deal with aerodynamics, there are the whole point. If you don’t want to rent some time on a supercomputer or want to make some complicated experiments with 3D-printed ship models in your bathtub you have to life with the arbitrary nature of these values.

I calculated the drag coefficient in the “atmospheric flight model 2.0”- testbed app with the equation for drag and the values out of the examples of the belly first falling and the maximum trust fighter brick.
The drag coefficient, of the flat side of a box, the app uses seams to be about 0.25, witch is about the Cd of the front of a normal car. If I have to compare it to other values and don’t bother with the effect of Reynolds numbers (Re should be about 10^7) I would say that is a good value for the front side of the fighter, maybe a little bit low for the belly side of a giant brick.

If I understand your method correctly the Cd of the edge of a cube (45° angle of attack) should be the exactly 1/2 of the Cd of a face of a box in the model, witch seams also a little bit low but still reasonable.

Because the ships are not streamlined but bluff bodies and the complex nature of the problem any value for Cd between 0.1 and 1.5 seems okey.

---

You will not get a 100% exact drag and lift simulation in an computer game working.

Personal I think a refinement of the old atmospheric flight model (1.0) to act a little bit more like bricks wouldn’t be so bad.

Old atmospheric flight model ( 1.0 )

In the old flight model, the two equations were implemented directly. To determine the lift and drag coefficients Cd and Cl, based on the angle of attack, I used the following curves:

http://www.aerospaceweb.org/question/airfoils/q0150b.shtml

As you can see, those simplify the object as an airfoil. Our ships aren’t wings, they’re actually more like bricks. So, in many ways, the old atmospheric flight model was incorrect. More importantly, the overall shape of the ship in the travel directory ( the “flow vector” ) wasn’t a factor.

I think if you modify the curves a bit you can use them to calculate the coefficients Cl and Cd of the faces of a box.
The curves should be okey between the stalling angle (25°-155°).
If you calculate the angle of attack between the flow and 3 faces of the box (top, front, rights side) and look there Cl and Cd values up in your curve and add the 6 forces in the centre of pressure up, you should have a good little simulation.

[Ripper]

I’d consider a radomized number of rays. Something like 20-30. Then have that variably change on the fly, so the pilot would need to constantly adjust for turbulence.

And then different values for different atmospheric densities.

[TheFox]

^Starts Clapping^
Bravo, well done.

Just my two cents, that might be only worth one:
I think there is difference between modeling realism and creating a realistic experience. In that putting too much realism may take away from a fun realistic experience. I think what you did here is perfect, it will add an interesting layer to the game: you could stop faster by using the broad side of your ship to crate drag.

Just watch those deep rabbit holes, I say keep the focus on gameplay as that is what we are here to do. Also you are creating Science Fiction, physics can be bent with technology and people will accept it.

[Zen]

Absolutely - couldn’t agree more. And using ship design like turning to create drag is a good idea. Or perhaps extendable flaps to create drag. Of course this will add time to ship building but it maybe a simple solution to an otherwise never ending quagmire of mathematical equations.

[bavellone]

Thank you for taking the time to write this up and showing us what you’ve been working on! Ever since I first found I:QfE I’ve enjoyed reading your technical blog posts on the inner workings of the engine

[selbie]

A few sound effects for your sound designers

[Freeman]

Did you like the roll-then-pitch approach of the old versions ?

Yes, I like it.

[wensir]

If the drag only depend on the ship model, flight direction, and speed, is that possible that you pre-calculate some of angles (say ~ every 5 degree in theta and phi directions), and do a interpolation during actual flying? I guess that will make the calculation much faster.

[drzerg]

why not just precalculate in some physic model all angles of attack for each speed and just make interpolation between them? what you need is a table consist of [alpha][beta]=[drag coef][lift vector coef] i think something like 16 alpha and 16 beta positions will be enough. coefficiens will be inserted in formula that calculate drag and lift according to angle of attack, speed, density.

if ship has symmetry you just need half of that data

[drzerg]

also what about hypersonic drag and lift? for high speed atmospheric flight there should be at least 4 models for M0-M0.8/M0.8-M1/M1-M5/M5+
i do not argue for precise calculations. if it will just feels like true will be great. for example at hypersonic speeds lift is very small regardless of the ship geometry

[Lomsor]

also what about hypersonic drag and lift?

Quick note: using raycasts is, at hypersonic speeds, surprisingly accurate. At subsonic speeds, it looses accuracy.

---

The current model is quite close to reality. We discussed it a lot in the dev forums and Flaviens test (see opening post) also supports it:

https://forums.inovaestudios.com/t/0-1-3-0-drag-tests-and-approximations/3464
https://forums.inovaestudios.com/t/patch-0-1-4-0-2016-05-01/3478

Sure there are ways to optimize the performance of the calculationt. Handling tables that need to be updated after every balance change sound tedious to me though.
We don’t even know if there’s a performance problem yet. The formula used in the last version wasn’t extremely complex and casting a few rays every physic frame doesn’t sound a lot more complex either.

[DirkLarien]

well what do we have here ? another observer ?

[Skyentist]

So this is something I’ve had on my mind for awhile that I’m just going to put out there. Its pretty unreasonable, but maybe someone else will think of something.

I thought it would be really cool to 3d print models of the different ships and put them in a real wind tunnel to find their coefficient of drag. I think it would be really cool to advertise that not only is your universe true to scale, but your spacecraft would also fly(mostly) true to physics as well.

I looked up pricing and it seems to be, unsurprisingly, pretty expensive. The first result was ironically UW and it’s 500$ an hour. I’m not even sure if that comes with any sensors and number crunching either. I say ironically because I applied to UW for grad school and unfortunately didn’t get in. If I did, I was going to try and do it myself but that’s not quite how the cookie crumbled.

[Aleon]

Might sound like a cool idea at start, but in reality aerodynamic properties are difficult to scale. If I recall correctly you’d have to use fairly big models, and really high speeds to get measurements with the correct Reynolds number, as the ships of IB zip around really fast. If you wanted to pre compute drag coefficients, you’d be a lot better off running numerical simulations.

[Skyentist]

The issue with Coeffecients of drag, is that they are empirically found fudge factors. You can sort of try to run numerical simulations based off of similar shapes but the best way to find the Cd is to find it empirically. This is especially the case when the object isn’t a well defined shape like a sphere, cylinder, etc. It doesn’t have to be perfect, this is only a videogame afterall.

Having a wind tunnel is what allows you to get high speed measurements in the first place. Otherwise I could build my own with a fan but that doesn’t really help if I only have the measurement of the Cd up to 30 mph or whatever.

[Red_Syns]

Subsonic, scaling for aerodynamics is not as simple as simply scaling the size. Supersonic, everything is found experimentally and “calculated” by looking up the values in a chart.

Realism is incredibly hard, when you are talking flight, and especially with non traditional bodies, where you are unlikely to find any sort of useful information.

[innociv]

The cD may be lower, but I believe it’d actually be more drag due to the greater area, no?
Well, it depends on how closely the shield hugs as well.

Lift at hypersonic speeds is also much more to do with angle of attack and surface area than aerofoils, as well, I believe?