Just a curious inquiry. We have been discussing this in general on another thread in the beginners forum, but I was curious if anybody could inform us on the general difficulty of creating aerodynamics in a racing simulator. It seems difficult to me to program, but perhaps it's not; I'm not a programmer
I guess aero will get really complex if you go into details, but actually aero is a lot simpler to simulate than for example tires. That's why we had flying simulators 20 years before racing sims started to get even remotely realistic.
Interesting point. How realistic were the flying simulators?
Granted, the only thing at play in a flying simulator is air, but in racing simulators there is a variable surface on a variable surface also exposed to the air. By that argument, a racing simulator is like a flight simulator with added elements.
I'm not terribly impressed by racing simulators yet. No simulator has what I would consider accurate physics. LFS does a great job, but even its developers want better. We have something close to what is needed, but it's accuracy to real life is lacking. Racing/driving simulators still feel like a universe separate from our own in regard to driving. That's not to say one VERY similar to our own, but it's not perfect.
Regardless, that's a great argument as to why aerodynamic physics would be easier to recreate than tire (tyre, for you non-American English learned linguists ) and surface physics
Current flight simulators aren't all that realistic. FSX uses lookup tables to decide what happens. X-Plane at least uses aero simulation, but even that is going to be simplified. Expensive commercial simulators may differ.
driving a car on track gives you quite accurate visual feedback, what the car did. Being off by 1-2m from your expectations will feel "wrong", if you are pro driver and you can lead your real car within centimetres.
Being few thousand feet above ground and missing the turn by 10-20m because of inaccurate physics... nobody can tell? So the critical part would be lift-off/landing simulation mostly, and for real pilots the corner cases like stalling the plane and saving it. And those would feel quite realistic even when based on tables, because once again you don't have good reference points to tell exactly what the plane did. And I forgot, the refueling during flight must be probably another weak spot where good pilot may notice something fishy.
Maybe having a Red Bull plane race simulator would be serious challenge, then again you don't feel the +-9G while playing behind computer, so it's already breaking the illusion for real pilots there a lot.
With weak cars my mind is sort of playing with me, giving me some fictional G forces feedback based on visuals (so I lean slightly with head and stiff the proper muscles), but it's still nowhere close to the body work I do in the kart and some tyre slips break the illusion immediately, just because the my body (and the inner ear "gyro") didn't move properly. But at least partially I can feel "there" with a good simulation, like for example LFS, it feels better, than I would expect, considering how it works and what are the inputs for the brain.
Well, the "Force Dynamics 401" for example is quite close. For weaker cars, with some future VR headset, with solid tracking, and stuff... I believe it can be really close to the real thing, to a point where the lower price may make you consider to drop the real thing completely. With proper race car and forces above 2G it's getting more difficult, then again most of the people will never ever experience this in real life, so getting close is not bad either.
@Cornys: and to the original question.
Actually if you want full aerodynamics simulation, with proper turbulences and slipstreaming, and with downforce modified by damage, you would have to do fluid dynamics simulation, i.e. like the CFD in F1 ( http://www.formula1-dictionary.net/cfd.html ), but in real time. Which the current computers are not even close to, in terms of horsepower.
So at the moment it's all "fake", although some games are very reasonable and sophisticated, but not simulating air molecules.
And to create a reasonable fake (or rather call it heuristic) is much more difficult than simulating basic atom/molecule interaction (in terms of formulas complexity and research).
Although if you don't care too much about the details, a simple "downforce = k1*speed^k2" will fake it quite nicely, for some arcade game...
Also has been a while since I watched that, but remember it was worth watching.
Another thing on the history of flight simulators VS driving, is what the focus of sim is.
If the focus is on instrument flight and navigation then simulators for that already existed in time of WW1.
Graphics, physics, forcefeedback was non-existant but just having moving instruments in front of pilot was appearently enough to realistic enough to give some training.
Flightsimulator can be a windowless cabin with nothing but some instruments, trained pilot can use that to take off, fly somewhere and land.
But driving a car around track just by speedo & RPM is impossible.
If one looks at early flight sims (not games) they all have most part of the screen taken up by the instruments:
If you tried to do loopings or such then an experienced pilot might notice that the physics are wrong/simplified. Graphics are simple too of course, but pilot of an 747 jet does not fly by sight anyway.
Extreme case is Space Shuttle simulator, there the view from window does definately not matter. The whole sim is basically staring at instrument panels..
Pilots are trained to avoid flying by eyes or feel. There is many stories of confused crews who could not even tell if they were upside down or climing or falling because the senses can not be trusted.
Most of the flying is by instrument, and there is no reason why even older computers could not give a somewhat real experience of that.
Driving race car is all about eyes and feel, so it could not be done until computers were more powerful to do graphics and physics.
Here's how I'm guessing LFS does it based on what you can see from the forces view.
Cars have a general downforce figure, and a downforce figure for any aerodynamic surfaces (front+rear spoiler, where fitted). These surfaces are represented as a point.
The forward component of that point's speed is measured (i.e. if the car's doing 90mph sideways the forward component is 0, but if it is going forwards then it is 90mph, and any angle between yields the relevent value between 0-90, through the magic of maths) and this is entered into a table. The table suggests that for 90mph, the downforce generated should be 200N, and drag 150N, and applies those values at that point. The table figures are multiplied based on the setup's wing angle for configurable surfaces, and the general vehicle downforce uses a set value. I never tested whether LFS inverts those values for reverse motion to create lift (a spoiler travelling backwards generates lift, which is sometimes the cause for cars taking off during high speed spins), but I suspect not, likewise I can't remember if slipstreaming is included in the equation, but this is usually 'when car B is within +/-10 degrees of the back of car A, measure distance between cars and calculate slipstream effect, deduct slipstream effect from drag). Would be interesting to find out from Scawen how close I am with that guess.
To simulate anything more complex would take big processing power, and in really complex CFD simulations it takes a computer minutes, hours, days even to simulate just a snapshot in time, let alone recalculating the whole lot 100s of times a second to account for the bodies being moved real-time.
From the forces-view it is not possible to make much assumptions like this.
It only shows the resulting forces and where they attack, but now how they came to be.
No matter how the downforce generated by wing is calculated, what it comes down to is that in the end *one resulting force* acts on the car a certain *point* and with certain strength.
That is what the forces-view shows.
If the force was calculated by CFD or a simpler formulas or by random number generator, we can not know.
On real wing many forces act, in some areas there is lots of downforce generated, in other areas is only little downforce, in some parts is maybe even lift generated, some elements pull to the sides.
Imagine the blue plane is some aereo piece, and the arrows at the downforce generated at this point.
If you want to simulate *the wing* then that is important to care of all those individual forces, because it could be used to calculate how the wing flexes. (for example)
From viewpoint of the car only the result matters, it is mathematically valid that you can sum up all those little forces to one resulting force.
What acts on the car is the resulting downforce through the point where the wing is mounted. (For example it is obvious that on a symetric wing the forces pulling to left/right sides will even out to zero.)
I think you misread my post, I was referring to the cars in LFS, not IRL. Of course aerodynamics on a real car aren't as simple as 'at x mph apply y force at point n'
Not completly sure what you mean, but the principle of having (for example) a rearwing simulated as two forces (downforce & drag) that act on the car at certain points is valid. It does not overly simplify things or so.
The question is more about how did these forces get calculated, but that is unrelated to what we see in force-view mode.
Thing is, that is not nessecary.
You can run a CDF analysis or even windtunnel test on real part or whatever and record the data. It does not matter how long these tests or pre-calculations might take, because you save the results and then reading the data ingame is very fast.
Just like the RPM/torque curve for engines, you do not have to simulate how fuel gets injected, how the cylinder fire, how the gasses flow through exhaust in realtime. That would be as impossible as complete realtime CDF for aero. Instead you take the data as you have measured it on dynamo/as you have it made it up based on guesses to similiar engines/as the producer has given you, and it gets turned into some function that is fast enough to be read during playtime.
I've no doubt that that's where the data that LFS uses came from, but I doubt for the sake of simplicity that it's any more complex than front, rear and general body downforce.
You're not making any sense. You're telling me I'm wrong, then saying exactly the same thing as I've just said.
For the sake of a live simulation, real world testing used to create a table of values for front wing, rear wing, and body aero is simple enough to work out the effect on the car in terms of wheel loading and net drag, and given the layout of the forces view I'd say that probably is the case.
Table data are accurate only in clear air.
Once you have something moving ahead of you (or in city the wind in tunnels and between buildings), it's just heuristic.
LFS does adjust a bit the table data by car in front/behind, although I can't recall for example effect of car going next to me, but I think Scawen didn't put months of research into those, and it's still reasonably complex, like losing downforce on rear, when somebody is really close on your back, etc.
Correct me if I'm wrong but current computer programs can only use lookup tables (or similar) to determine results in a program.
When they reach sentience and are capable of actual thought then this will change, but not before then !
All our current games are simulations, and the differences between them depend on programming with alternative lookup tables.
This is the difference between a simulation and reality.
I'll correct you then. A lookup table is just that: a table of inputs with results which the program looks through. A lookup table will have a finite size and may not contain every possible combination of input values, so the program will take the nearest.
The alternative is using formulas to calculate the results from the inputs, which will take more CPU time.
In LFS I think a car in front of another does result is something very similar to what it does in real life. There is a drop in down force and drag.
I'm curious as to rather or not ride height, body orientation (left side higher than the right for example) and such makes a difference on aero dynamic effects as well. I guess I haven't paid that much attention.
BUT !!!, a formula is actually a way to create a complex lookup table isn't it.
If x= and y= then z=. Vary x and y by random results, then z alters accordingly.
This is not AI, or in fact a system that allows for real world variables, i.e., a random event not included in the formula/lookup table, This is simply another, more complex way to create a variable lookup table without including every result in a pre-designed table.
Case in point, modelling aero effects on cars, the issue is with the ability to include EVERY variable to create a 'real world' effect within a sim game.
