The Flight Model

These simple words seem to summon the desire of every simulation advocate and nay-sayer to muster their personal wealth (or dearth) of accrued information and passion regarding the accuracy and validity of any given flight simulation's true worth. In the process, they heap accolades on developers or crucify them as a group.

This raises the question, "What is up with the flight models today and what makes them the center of such debate?"

"Into the breech…"

My decision as to whether to base this article on jet simulations or prop simulations was an easy one. Aside from the odd zealots determined ranting, jet simulations benefit from the access to complete and recent specification documentation. Prop simulations on the other-hand suffer from their age of inception, resulting in a scarcity of complete and accurate documentation.

In preparation for this article I structured a questionnaire and sent it to many of the most recent and upcoming simulation developers with the hope of "picking their brains." Surprisingly, I received few responses (one of with was rendered mute due to Sierra's cancellation of Desert Fighters). Those who did respond provided me a wealth of information that I believe will shed a great deal of light on this subject.

"Enter the players…"

• Chris Sherland representing Playnet and their WW2Online simulation using their "as of yet" unnamed flight loop technology.
• Constantine Hantzopoulos and Mike Valdez representing Looking Glass and Flight Combat based on the Flight Unlimited series Flight of flight models.
• Andrew Walrond representing Wayward Design and B17-II the Mighty Eighth using an all-new engine developed by Wayward Design.
• Michael Harrison representing Parsoft / Inertia and Fighter Squadron, using the Openplane engine.

"And the Question Is…"

Firstly to better understand the state of flight models, I thought it best to allow the developers to clue us in as to what kind of flight model engines they are currently using.

Q 1.) Describe your flight model engine and how it works within its virtual environment.

Q 2.) What advantage do you feel your engine has over existing simulations?

Playnet

Playnet Q1) Our flight model looks at each vehicle in components. First the plane is measured and divided into logical sections. Each section is given accurate weight, roll inertial values in all axis, and sub components such as control surfaces, fuel and oil tanks. All the sub components share the same level of detail as well. The airfoil's sections are all calculated for lift/drag/stall through 360 degrees of AOA.

That is simply obsessive, but it gets us much more dynamic interaction with airflow at severe performance regimes, and it's fun to claim. After all that is set we send the model through a wind tunnel program that was written in house. Here is where we can see how close the performance values are, right off the bat. Then it's a matter of detailing the control surface dynamics and entering dedicated detail and oddities to the individual model.

Playnet Q2) We feel our advantage is that our flight engine works with real-life numbers, weights, pressures, HP, drag, lift, and velocities etc. All that the flight "modeler" needs to do to create a model in our engine is to accurately transfer the dimensions, weights, and sub components of an aircraft into flight model data.

I'm not sure that we are making any huge leaps forward in flight modeling in general, but we are certainly increasing the overall fidelity and upping the level of detail on flight regimes for the industry standard. I'd say the big advantage in our model is that it's all based on real numbers, and it's very easy to check your specs that way. This translates into more accurate depiction of the actual performance of the aircraft with less guesswork.

Looking Glass Q1&2

Q1) Our flight model is what we call a piecewise force-based model. Let me explain these separately. "Force-based" means that we calculate the forces on the plane (lift forces, for example), then use Newton's First Law (F = ma) to calculate the resulting motion of the plane, that is, its new position and velocity.

I believe that most flight Sims today use the force based approach, and with good reason. Since nature is "force-based", it is no surprise that modeling a plane this way would result in the most natural and realistic flight behavior. "Piecewise" means that we do not model the forces on the plane as a single, indivisible piece. Instead, the plane is made up of separate pieces (up to 70) which individually calculate their force contributions and then all these contributions are added together to give the net force on the plane.

I'm pretty sure that few flight sims use the piecewise approach, which has advantages as described below. Within the simulation our flight model is merely one of several physics models that can "take control" of an object. The player plane is always under control of the flight model, and at certain times, other planes are switched from a simpler model to the full flight model.

LGlass Q2) I believe that relatively few flight sims use the piece-wise approach described above, which I think is the biggest strength of our model. This is because it captures the behavior of a plane with missing or damaged pieces almost automatically - we just need to damage or remove the corresponding piece in the model, and the model just starts behaving accordingly.

For example, if the left wing tip breaks off, then we just remove that piece from the flight model. The remaining pieces of the plane keep providing their force contributions and the plane starts tailing to the left because there is no longer that lift contribution from the left wing tip.

Wayward Design Q1&2

Q1) Firstly, we have modeled the world. This includes simple things such as the height of the ground and gravity, but also complex things such as the physical properties of water, bumpy grass, icy tarmac etc. Also in the world we model the atmosphere, with pressure, density and temperature all varying with altitude. Then weather, clouds, wind, gusts, turbulence etc.

Next we break the vehicles in our world, (that includes the ground vehicles as well as the aircraft) down into their component parts, which we model separately. Weight distribution and payloads, wheels, brakes, suspension, airfoils and their modifiers (e.g. flaps), engines, propellers etc are all modeled individually.

To implement an aircraft, we create a wrapper and place all the relevant parts into it, at the correct physical locations. Some of the components, most notably the propellers and airfoils cause secondary effects in the world: turbulence, vortices etc which not only effect the other components of that aircraft, but other aircraft nearby as well.

Q2) For us, the advantages of the component approach are simple. Each component is self-contained, easy to maintain and can be tested in isolation. Any aircraft can be modeled quickly and easily by assembling the correct parts with the correct specifications. We can have a new plane in the world within an hour and expect it to handle very realistically from the start. Only mild tweaking of the numbers is usually required to finalize the model. But we can just as easily create physical models for jeeps, trucks etc using the same components.

Parsoft Q1&2

Q1) The OpenPlane engine models forces, which simulate the forces that a real aircraft is subjected to. Each aircraft is composed of many different parts which all have weight, create drag and some cause lift. We build our aircraft in much the same way that a real aircraft is built, by the aggregation of individual parts which all work together. We start by creating the fuselage and adding objects with mass and aerodynamic properties such as wings, fuel cells, gear, guns, etc. By the time we're done, we have an aircraft which very closely matches its real-world counterpart.

Q2) Our greatest advantage is that our model is based at its core, on real-world physics. This is also our greatest bane; if an aircraft "looks" right, but isn't balanced correctly, it can be very difficult to fly. The result of the work required is a flight model that re-creates "flight" as no other simulation can.

Response

I must be missing something ... Based on these responses; great length is taken to micro-manage every aspect of the flight model. Seemingly every possible piece and nuance was being modeled.

It's obvious that the fidelity of today's flight models are head and shoulders above older simulations like WarBirds (which incidentally seems to be the "measuring-stick" all simulations are held up to), yet they still seem to fall under microscopic scrutiny by players. If there is a problem out there, maybe it's not the "engines" themselves but the data applied to them. Having said that, I asked:

Q 3.) In user groups and forums, one of the greatest debates focuses on the "accuracy and authenticity" of today's flight simulations. Explain how you've collected information regarding flight performance data and how do you qualify it for authenticity?

Q 4.) In hindsight, what is the most difficult thing in respect to modeling WWII aircraft?

Looking Glass Q3&4

Q3) First, we gathered all the ORIGINAL pilot manuals and performance charts for each and every aircraft modeled in our game, a difficult and exhausting task (so much so that some of our German plane manuals had to be translated from German to English). Bob Shaw (yep, THE Bob Shaw), helped us out immensely with this.

Second, we built what we call the "Flight Model Editor" - a Wind tunnel tool that allows us to plug in values directly from performance tables and tweak those values in real time as we are running the simulation. This really helps us fine-tune and hone the physics and flight modeling, as we can bring in pilots, have them fly a particular model, comment on it and instantly tweak characteristics.

Third, to apply the gathered plane specifications to our flight model, we basically have to "work backwards" from the specs to come up with appropriate values for the parameters in our model. For all its advantages, the force-based approach does present one significant challenge. It is often very difficult to translate high-level performance characteristics into force values.

For example, if the spec says a plane has an optimal climb rate of 2000fpm, there's no simple formula for figuring out what the corresponding parameter values in our model should be. In fact many parameters affect the optimal climb rate, as an example: air pressure. Also if you manage to set the parameters such that the model does indeed have that optimal climb rate, its very likely that the accuracy of the model with respect to some other performance parameter, like max roll rate, has been compromised.

Q4) For the most part, we don't have access to the pilots or the planes of WWII so we can only go by the high-level performance data that is generally available, such as top speed, rate of climb, takeoff distance, etc. We said we have a force-based model, and there is basically no force data available for WWII planes. Even to model a plane of today, the manufacturer might be hard-pressed to find such data and even if they had it, might not be authorized to give it to us.

Playnet Q3&4

Q3) This is a good question. We are using lots of NACA reports, Wartime flight test evaluations, pilots manuals, and snippets of stuff we dig up out of the strangest places. We use ex-military sources, web collections, and libraries of the truly insane (Imagine having detailed info on the MS 405!). If you dig deep enough, you can find anything! The first part of modeling our planes is all math...areas, weights, roll inertias etc. We get that stuff from detailed construction stats and then turn it into data that our engine reads.

Then it's wind tunnel time, check to see if the math worked or if there were any errors. Then we detail the performance parameters, this is where you find out how close you are by cross referencing the way it flies in the sim with what the real aircraft's flight tests say, and what the pilots reported. Stall, spin, pre and post departure character, and any other stuff that's specific to THAT plane needs to be looked at. Sometimes it's not easy to get a plane to "sag" or "wobble", or "burble", those are very subjective terms. But that's where the factual nature of our engine is so rewarding.

Usually, if you plug in all the numbers right, and you took a LOT OF CARE in measuring and calculating the specs, the plane will act like the reports. The first few times we were kind of dumbfounded actually, and wrote it off to "luck", but it's turning out to be fairly consistent, so we're not going to complain. But again, every once in a while you run across a plane that has very little data on it, or widely flawed or even conflicting info.... Sigh, that's when you have to get the scalpel out!

Q4) Well, I'm not the most experienced producer around, but I'd have to guess that consistency is a big challenge. Getting planes to act their age together. You may model a P36 and believe it's spot on, but then you get it in the air with a A6M2 that you also think is accurate, and you say "hmmm, is this what it was REALLY like?" There's simply no-way to get all the nuances of what prop combat is like, you have to make choices and hope that they have no impact, or at best a mild impact on a model that's 20 planes down the list.

Our philosophy is simply this: Give the players the benefit of the doubt in regards to their ability. Sure there are guys out there who can't open a can of beans and talk at the same time, but they won't get flight sims no matter HOW easy you make them. The guys who love flight sims, end up staking a lot on the "feeling" that they are good combat pilots. Well, delusions aside, they ARE good combat pilots...INSIDE THE BOX. As a producer you have to give them challenging models that feel right. So they can work hard to gain experience and get better.

The level of ability of WW2 sim pilots has risen incredibly in the last 5 years. And it's a direct result of more challenging models being produced. The work that Dale Addink and Doug Balmos did on WarBirds had a huge impact on the WW2 prop sim market, and the learning curve of the WW2 flight sim IMO. Marketers and suits will say that it needs to be easy-to-fly in order to make money, and I feel that that is a mistake.

The WW2 flight sim community as a whole is smarter than they were a few years ago. These guys are looking at the NACA reports too, and don't think for a minute that they are not. Especially when it comes to "their baby", whatever plane their grandpa flew in 1944 etc. So to make a short answer long, the real challenges are consistency, and giving your audience a challenge.

Wayward Design Q3&4

Q3) Throughout development, we have uncovered extensive research materials into the specifications and flight characteristics of all the aircraft we have modeled. This includes, in our case, detailed test-pilot reports of both allied and captured axis aircraft. During Beta test, pilots who have recently flown the actual aircraft will put the aircraft through their paces. Where this is not possible, for instance with the Me163, we will use the combined wisdom of pilots from other aircraft of similar design and performance.

Q4) Ground physics. An "impact" is by its very nature something that involves very large forces for very small amounts of time. In flight, most physics engines recalculate the physics about every 33ms. This is nowhere near being a high enough frequency for realistic ground behavior. I mean taxiing over bumpy grass, belly landing, crashing, dipping your wingtip onto the grass. We deal with this by using asynchronous and variable frequency impact physics.

Parsoft / Inertia Q3&4

Q3) There are a number of sources that we use when researching aircraft. These include independent researchers who do the work professionally, and book and web resources.

Q4) All in all, it's the conflicting information that's available. There were a great number of variations made for many of the aircraft from that time period. These include line and field modifications, which are not always noted in official and unofficial documents. When researching a particular aircraft we have to carefully compare the documentation with personal accounts from the war in order to provide the best simulation possible.

Response

I'm hesitant to say this, but based on the above statements simulation developers are faced with a dilemma. Obviously older technologies like simple "data-lookup" tables are far easier to use to convey specific flight-performance specifications.

But simulation consumers are demanding higher fidelity flight models, which by their very nature require vast amounts of "specific" data and specifications, many of which are extremely scarce if available at all. That coupled with the light-speed advance in technology makes the developers task a daunting one.

With today's high-fidelity flight models, it's no longer sufficient to simply model the aircraft's flight performance, now developers have to model the "aircraft." This includes having hard data for:

• Surface area = Not only for the wings, but also for the horizontal stabs, rudder, ailerons, elevators and the fuselage.
• Airfoil data = Including chord thickness, moment of inertia and airfoil data for any leading and trailing edge devices.
• Specific object weights = How much DOES the left aileron of a P51D weigh?
• Center of Gravity = Not only for the aircraft as a whole, but also for specific object components.
• Degrees of throw for all control surfaces.

As you can see … simply knowing the maximums for alt, airspeed and roll just don't cut it anymore.

Speaking of technology

Maybe with the advent of newer technology, there is a greater possibility ahead? While on the subject, let's pin them down again!

Q5) Taking into consideration the recent advancements in processor speed and 3D rendering technology, what do you think may be possible in the near future that wasn't possible in the past or present with regard to flight simulations?

Q6) In a perfect world … What would you like to provide to consumers toward the "best" simulation possible?

Wayward Design Q5&6

Q5) The great leaps in floating point performance, coupled with the next generation video cards having all the transform and lighting stuff in hardware will allow us to have much higher fidelity flight models, higher frequency physics calculations and a higher frame rate.

Q6) "B17-II The Mighty Eighth" and a free, zero-lag Internet connection. Oh, and fear. If I could condense fear into tablet form, there would be a bottle in every box. "Take one before every mission...

Parsoft / Inertia Q5&6

Q5) As the processors become faster we can devote more time to physics processing and bring the fidelity of the simulation up to an unheard of level. Combined with the advances in 3D acceleration, the bar on graphics will also continue to rise until we reach a level where we may not be able to tell the difference between what we see on-screen and a scene captured with a video camera.

Q6) Essentially a whole-war simulation, fully simulated squadrons, troops, ships, you name it. Basically, a simulation where the player can see and experience every aspect of WWII in one-encompassing package.

Looking Glass Q5&6

Q5) Well, graphics are obvious here; 32-bit texturing, texture compression, photo-realism, the list goes on and on. On the physics side of things, dual processor machines, once they become the norm, will be the biggie. Games currently dedicate about 10% to physics, 10% to AI, 10% to sound and the rest (70%) to rendering. Imagine having a dedicated CPU running just the physics! Some of the fallout would be better fidelity of not just your plane but of AI planes as well.

Q6) Purely from a physics point-of-view, it would be great if we could give the consumer a flight model that rivals the fidelity of the simulators used to train airline pilots. Those simulators model the airflow around the wings, fuselage and other surfaces of the plane using the techniques and equations of computational fluid dynamics (CFD); or at least they use much more plentiful and accurate low-level data on the planes they are modeling.

Playnet Q5&6

Q5) Well the biggest news will be in physics. Sure graphics will be constantly improved, but the ability to model fragmentation damage, ricochets, high fidelity structural damage, THAT'S the stuff that will be the true raise of the bar. Flight sims are already looking great, and will keep looking better, but the lifting of limitations on CPU speed will really be rewarding when we can increase the ability to model more physics.

Q6) A total-combat-experience, not just planes, and not just shooting. Air, Land, and Sea vehicles, planning, teamwork, accomplishments and consequences all happening in real time with real people. We're calling it WW2 Online…

WOW … Was that a ride or what?

I guess the question is: "Did I get my question answered?" If you will recall, I was asking, "What is up with the flight models today and what makes them the center of such debate?"

Unfortunately, my question stands. It's obvious the programmers are using every possible resource at their disposal in an attempt to bring their respective flight models to absolute fidelity. So it's hardly the technology, and that leaves two possible answers to my question.

1.) The hard data problem. As a result of this article, it's obvious there it a wealth of "common data." As the demand for high fidelity flight models increases, so does the need for very specific and isolated data specifications, something that evidently is a rare commodity considering the age and availability of the subjects.

2.) "Personal interpretation." Let's face it, in our own "mind's eyes" we all have our own personal ideas of how these classical birds flew. The developer that best captures our imagination that will run away with flight simulation's "brass ring," but that developer may or may not have best captured the model. In the end, it will be a subjective assessment by the player which will likely involve many factors other than merely the flight model.

These two factors probably explain the common complaints on flight models. In any case, it seems certain that the fm is a key component in simulation design and will continue to be a determining factor used by many virtual pilots in their buying decisions.

For more on these coming prop simulations in development see our Air Combat Previews index.