AI performance improvements

[nearmiss]

I have have a compilations of postings documents I've compiled over the years about fixing, tweaking and patching the AI performance.

I'm wondering, if anyone has done anything more substantial than my meager efforts of just extracting information from forums postings and creating my little AI document folder.

I have always been perplexed by the AI in CFS2. Needless to say, I should say "pain in the Wazoo". Planes crashing on takeoff, crashing on landing, diving into the ground, flying aimlessly, etc.

I'd use the search, but 2 letters searches go nowhere on these forums. I don't know what search terms to use effectively. AI = 2 characters so the results are nil.

Maybe we should give the AI a more fitting name, which would be searchable.

I like the term "Gomers", a term many of you remember from Gomer Pyle. All the Marines I know always hated that show and Gomer Pyle, he sure wasn't any John Wayne.(Sgt.Stryker)

So I'm referring to the AI as the "GOMERAI" then we can always know we are describing something inordinately dumb, but a necessary nuisance. LOL

---------------------------------------------------------------------

I have several pages of postings and docs about AI. I plan to put them up here as quote postings.

I don't claim any credit for the information, but I do think it has value or I'd never have made the documents.

I am renaming this thread to include "GOMERAI". Then you can search for it on the forums thereafter.

I'll edit and poke GOMERAI within each posting so the search will work.

Others can add to the thread as well, I mean I hope others will add to it.

I wasn't real careful about who wrote what, so if you want some credit by reading the docs let me know.

I'll edit each to make sure you get credit if you request.

Hope this helps as a starter reference document so we can share information about the AI to help each other.

Better still, what about including links to downloads that improve the AI performance

I've even found a few download files in some of my old files that aren't up on SOH.

I'll put them up and link from here as soon as the site is back to par.

[nearmiss]

Cees,

1 - What kind of bombers do you use ? Stock or add-on ? Level bomber, dive bomber, torpedo bombers ? What is the target : an infrastructure, a moving vehicule, a moving ship ? Did you put any weather in your mission : wind is to be avoided or kept slow ?

I you give me more details, I could probably help you.

Nevertheless, some things I know :
- don't force your bomber to follow an defined approach path by putting WPs near the target. Let them find their way. Dive bomber need to fly above an infrastructure target before making their run, same for torpedo bombers ...
- level bomber cannot attack a moving target. They drop at the predefined position. If the target has moved (like a ship), they either won't drop, or drop at the defined position of attack.
- altitude of the last WP can also have an influence, if that waypoint is too close to the attack point.
- Stock AIs are often more effective than add-ons, due to their flight dynamics (see §2). If they can stabilise their flight, they can't aim, and thus don't drop ! Tune your mission with a stock model. You can then make AI dedicated models by plugging the stock flight dynamics to the add-on ones (see Milne Bay)
- set the unit familly to 2.
- ...
Anyway, that's not an easy task and its a long process to tune and test a mission !

2 - AI flight behaviour
Most good (ie accurate) flight models are too hard to fly for the AI. They can almost never land. Some of them cannot take off. And if you set some wind, then each one crashes.
They also have difficulties to fly straight : they like flying like a dolphin, jumping everywhere ! The AI cannot stabilise the aircraft. The control loop isn't stable. This is particularly important if you want them to fight : they can't aim, they won't fire or drop their bombs. Try a dog fight against a 1% A6M and compare it with a stok one.

Again, my solution is to use stock flight dynamics for the AIs.
It is certainly not accurate, but you have certainly noticed already that the AI do not use the same flight model as the player's ... and for a bomber set to unit_family = 2 it doesn't matter as the aircraft follows the waypoints, whatever happens.


Guy

[nearmiss]

Since my return to the fold I've been working on updating my PTO install, starting with the fleet and dredged up an old project concerning how AI's attack ships.

If you've ever noticed when you set a formation of AI Only (Non-Player controlled AI's) to attack a certain WP that is not an individual ship only one AI A/C will attack a ship at a time. If you have the WP set to Attack with All Weapons that single AI will continue to attack until it is shot down or the target is destroyed. While this is happening all of the other AI's will fly off without engaging that ship or find other targets. I've figured out a way of changing this behavior where the lions share of the AI's will attack the most valuable target.

By adding a couple of lines to the DP I can now have multiple non-player AI's attack a single ship, even if they come from different formations. I have been working on a scale for Priority and Hardness values to be added to each ships DP. The Priority scale runs from 0 to 9 with CV's and BB's on the high end and PT Boats and such on the low end. But the Hardness value is the really important one. Right now I'm testing with values as high as 12. This means if two formations of 8 approach a CV with a Harness value of 12 that 12 of the 16 will go right after the CV. The other 4 will either find another target(s) like a DD or hover close by until one of the others has finished its attack. I put 1 CV and 4 DD's in a star formation with the CV in the Center. The WP was set to Attack with Bombs. I sent 16 AI's after them at one time. 12 AI's attacked the CV and 1 AI went after each DD, this happened all at once, not one at a time several minutes apart. I then threw in a CA and a CL. Both have a higher priority than the DD's. The AI's did go after one DD but all of the rest went after the CV and CA. The CL was in the lead futhest away from the AI's line of approach so they never go after it.

This has always been a part of the AI behavior that has frustrated me. I'm working on a set of ship DP's that not only incorporates this but lots of FX and especially AA gun corrections. I've done close to 100 DP's already covering VN, Usio, Mas, AF and a number of other ship builders collections.
I'll also include instructions how to incorporate this new information into your own DP's.
I actually figured this out from the DP of the stock Wharf Inf object. I noticed that it was the only ground object that more than one AI A/C would go after at once. From that I just expanded the idea to the ship realm.

As far as the 2TAF Airfields..... It's going to be awhile before I think about tackling a project that big. Little steps. With these ship DP's I can work on a couple a night and then play some. I got burned out bad I must confess with the Airfields. To complete a project of that scope you really need a team doing it in concert. One laying down the fields, someone creating buildings, someone doing research of buildings, the GSL.......

I think this is will bring more realism and help the AI's to replicate the real tactics of concentrating attacks on the CV's and Heavies. It'll still have some randomness to it but it'll be more in line with the way things really were.

B24 you're right on. I started this idea with those package but never was happy when I transposed it onto the ship DP's. I tried it with ships a long time ago but never could make it work right. The problem was that I only did it to selective ships, the others were left with stock DP's.
This time I did it to all of the ships, bingo! I'm working on a bunch of Usio's CL's right now and sent 16 AI A/C after 4 of them in a line astern formation. (CL's have a Hardness of 4). Peeling off by 4's the AI's marched right up the line attacking without any pause. Bang, Bang, Bang, Bang....


Sounds good Jagd, if you need some data on a field let me know and I'll see what I can dig up from the old stuff. The maps just have big black squares where the building were for the most part so you'll have to dig for photos to find out exactly what was what.
If you do not mind, may I ask your advice on AI Dive Bomber?

For "Dive Bombers", CFS2 sets Val and SBD as Category10, not Category3.

I tried both Categories on AI Dive Bomber, and found that Category10 Dive Bomber was more aggressive than Category3 one.
Sometimes, Category10 AI Dive Bomber never tried to attack even if the pilot skill was set Ace with High Aggressiveness.

Did you experience the similar case?
If so, should we always set AI Dive Bomber's Category as 10, not 3?
All of my divebombers are Cat. 3. I know CFS2 stock sets them to 10 but using Cat. 3 they act better. They don't try to act like fighters and they dive steeper. When attacked by fighters they try to evade and then proceed to their target if possible. This is for non-player AI's.

Player controlled AI's will always try to act like fighters, Even if they're B-24's.The Non-Player and Player Controlled AI's use different DLL's to control their behavior.
Last weekend I was doing some testing on some scenery and that led me to start a detailed study of the AI behavior when they are attacking static targets. I mostly wanted to understand the exact cause of AI's not attacking on the first pass. You know, where they pass over the target then turn 180* and come back. In the attached report you will find all of the answers. The When, Why and How of AI Strike Missions and how to set up things so you never have to worry about Non-Player AI's doing an Airshow Fly-by prior to their attack run. This Report covers Category 3 and 10 Aircraft and some very stong reasons for making all Dive Bombers Cat.3.

Enjoy the information, Cat.s 1 and 4 are in the works now.

No problem X, this should help all of the mission builders out there get the best out of the AI's. Helps make things just that more realistic for us all.

bob, my Cat 3's are the Dauntless, Helldiver, Val, and Judy. I don't have any Cat. 10 A/C. For testing I switched my SBD back to a 10.
another good thing about Cat.3 is the won't try to dog fight, they'll break formation and try to evade some but they keep pressing toward the target.
Lots of Observation Time and a little Deduction and there's is was.
Just waiting to be charted out. The numbers don't lie in this case.
The real big deal to this is the Cat.3 from 9000' up, they just go straight at it like clockwork, everytime. Powerglide Speed Run, Roll Over and Drop like a rock! It's really something to watch in Chase View.

Other A/C I'd throw into Cat3 would be the Stuka and any Attack Bomber with Dive Brakes. A/C like the Vindicator which if you powerdived like a SBD would rip both the skin and the wings off I'd leave a 10 or maybe a 4. Still testing the better of the two for this type of A/C.

Looks like moving the Attack WP back away from the target area is going to be the biggie for Cat.1. Just finished the low level tests (500' to 6000') and the further you move the Attack WP back the better they do. Preliminary Cat.4 tests showed the same thing.
Unit_Family
1=fighter
2=bomber
3=vehicle
4=Ground Unit or object
5=boat or ship
7=carrier

Category=
1 - aircraft fighter
2 - aircraft level bomber (B-24, Sally)
3 - aircraft dive bomber (Stuka, Dauntless, Val)
4 - aircraft torpedo bomber (Avenger, TBD)
5 - aircraft tactical bomber (B-25, Betty, I also have Kate in this Cat.)
6 - aircraft strike (I need to do some serious testing with this one)
7 - aircraft civil
8 - aircraft recon
9 - aircraft transport (C-47)
10 - aircraft fighter-bomber (any fighter equipped with Bombs or Torps)

Category codes for non-aircraft objects:
13 - ship aircraft carrier
14 - ship cruiser
15 - ship destroyer
16 - ship submarine
17 - ship battleship
18 - ship patrol
19 - ship auxiliary
21 - tracked vehicle
22 - wheeled vehicle
23 - train car
24 - amphibious vehicle
29 - artillery static
30 - mechanical equipment
31 - structure static
32 - factory
33 - non-target object
34 – runway
Something I noiticed with the TBD set to 5. The AI will do a level bomb run, but your wingman will do a glide bomb attack. All in the same mission.
No matter what the A/C's Category is, when used as a Player Formation AI they take on the behavior, or at least try to, of a Category 1 A/C. This is even true for Family 2 Bombers. Set up small mission where you lead a group of B-24s. Tell them to attack a ground target!!! LOL You'll see! MISC section entries for FAMILY/CATEGORY. FPRIVATE "TYPE=PICT;ALT=" Unit_Family 1 being Aircraft, and the Category entry being the type, of course. There are 1, 2, 4, 5, 9 & 10.......maybe more. 1 for all fighters; 2 for heavies like the B-24; 4 for the TBD; 5 for TBFs, B-25s, Bettys and other Japanese twins; 10 for Kates, Vals and SBDs. And these numbers will determine what type of attacks the aircraft will/will not perform. I've never made any sense of them. Like why is a TBD in a different category that than the SBD when they are both divebombers? Etc. What I have discovered is that if I want an AI aircraft to deliver ordinance that it refuses to deliver with it's normal type, I can get it to do the job by changing the type

nearmis--
Is there a difference between the way AI performs in combat based on the aircraft type?

Now that,s a loaded question !! you have to define combat first ,,,
Air to air---attack-ground targets---attack shipping,,,,
Yes would probably aply to all three ,,,

Does the Zero use the same tactics as the Corsair, George as the F6F, etc.?
Certainly not from player point of view ,,, but I think you directed the question to Ai,s ,,,, My own observation ,,, would say yes ,,, But you can inhance and modify each planes character in Ai mode as well as player ,,,,
Via the tune specs in CFG ,,, It becomes a sliding scale ,,that you can adjust to fit your own talent level ,,, For example you fly hard level in a wildcat and shoot down 8 zero,s ,,, Time to up the anti level ,,

Wildcat

[flight_tuning]
cruise_lift_scalar = 1.2------------More boom and zoom
parasite_drag_scalar = 1.0
induced_drag_scalar = 1.0
elevator_effectiveness = 0.6 -------Delayed elevater
aileron_effectiveness = 0.6--------------Delayed aileron
rudder_effectiveness = 0.6---------------Delayed rudder
pitch_stability = 1.0
roll_stability = 1.0
yaw_stability = 1.0
elevator_trim_effectiveness = 0.4
aileron_trim_effectiveness = 1.0
rudder_trim_effectiveness = 1.0

Zero--Ai
[flight_tuning]
cruise_lift_scalar = 1.4------------Much More boom and zoom
parasite_drag_scalar = 1.0
induced_drag_scalar = 1.0
elevator_effectiveness = 1.5------Faster ---elevater
aileron_effectiveness = 1.5-------faster--aileron
rudder_effectiveness = 1.5-------faster --rudder
pitch_stability = 1.0
roll_stability = 1.0
yaw_stability = 1.0
elevator_trim_effectiveness = 0.4
aileron_trim_effectiveness = 1.0
rudder_trim_effectiveness = 1.0


Now you,ve upgraded the E-Ai and slid down the player aircraft ,,, chances are you,ll knock down 4 Zeros ,,, but you,ll take damage doing it !!

My 109e can break straight up on Ai controls ,,,That my Hurricane cannot match and hold a target site on ,,,,so it puts the player in anticipation mode ,,

It really is an amazing sim ,,,that host,s all levels of talent ,,,,

[nearmiss]

Goopney 9-2-04

Pen32Win - I can only speak for myself - don't think that I misunderstood - lol - oh well - it's always possible - I'd never dare to separate aircraft.cfg and air file concerning these 'problems' and I'm quite sure that Farmboy hasn't misunderstood that, too!

again - I'd like to see somebody of you AI cracks nail down a simple to follow checklist on what to check / test to produce acceptable AI for 'designers' like myself - as far as I tried to follow this thread besides everything else, it comes down to something like ...

- take an air file from the closest stock plane as a base
- work up - automatic landings ..
- Angle of Incidence should be 1 degree (magic number)
- MOI's
- drag
- what else ? - what are the most important numbers that have to be as correct as possible?
- are there more 'magic numbers'?

remember - I'm a simple guy that tries to nail air files with the help of mudpond.us spreadsheets ...


Farmboy 9-03-04

One interesting point about the P-38 is that it's not a dive bomber and was prohibited from dive angles exceeding 30 degrees, so the fact that it does not succeed as a dive bomber in the AI role is not such a big issue. However it was mentioned that the Avenger and Helldiver may be problematic, and of course the Helldiver was a dive bomber so one would expect it to dive bomb properly. The Avenger, a torpedo bomber, may not have been expected to dive bomb, but I don't have that information at my fingertips and have never researched that particular aircraft.

Good point about the speed and handling of a P-38 with the weight and MOI values of a Wildcat. What you've got there is one very fast Wildcat, at least in terms of handling and performance!

pen32win 9-03-04

Hi FB,

You're right it does make it hot but what you have is a very light P-38. The airfile being untouched still leaves it with a lot of the P-38’s characteristics which just need to be tuned down to compensate for the lighter weight. Just watch how wide it swings when making continuous strafing attacks. It goes way out like the stock P-38 before lining up its next pass. The stock F4F has completed its second run long before the “P-38 Light” gets around to lining up a second time. Heck the stock P-38 as an AI has a hard time dropping bombs on static targets let alone moving ones and its staffing runs leave a lot to be desired too.

While dive bombing was not in the P-38 playbook, yes I’m well aware of the P-38 Compressibility problem, they did glide bomb and did lots of it in the CAS roll. Depending on the Category of the A/C, the Pre-Attack and Attack WP Altitude (or the current altitude in the case of an Player Controlled AI) and the type of target CFS2 will start one of four different AI Surface Attack sub routines. (These are just the ones I’ve been able to catalog, there may be more) Each is different in Angle of Attack and length. One is the DB Routine which you most often see used against moving ships or vehicles. The Player controlled AI’s have the hardest time following this routine. The AI will climb or descend to 6500' get on the ships tail and then steeply dive on the target. You see this with Category 1 and 10 A/C. There is another Surface Attack Routine used by Category 4 A/C which is much the same but is shallower in Angle of attack. The next routine is usually used by the AI when a static infrastructure target is selected and the assigned attack altitude is under 3000', It is a very flat Glide Bomb run. The last is a very long GB Routine used when the attack is initiated from above 3000’ against static targets. Anyway, the P-38 was just an example of what the effects of high weights and MOI’s have on the AI. Most of our experience with these heavy A/C has been with the Typhoon, Tempest, Mosquito and Avenger all big multi-roll A/C. Without exception 2 different A/C have been required to make things work right for both the player and the AI.

As far as the Avenger, Glide Bombing Runs usually started from around 5000’ and were the most common employment of the Avenger when it was packing bombs, which was quite often, much more often then they carried Torps. There are only a couple of times I’ve found where the Avenger was used in a horizontal bombing role, one being the CV Raids on Tokyo. Most of the time though the above Glide Bombing Attack profile was used for Anti-Ship, Airfield Strike and CAS missions. CFS2 has the A/C Category wrong it should be a 4, not a 5 to make the most realistic use of it. The AI Only Avenger I produced for the Marianas Campaign reflects this. BTW, The book “The Big-E” goes into great detail about these tactics used by the “Turkeys” and is an excellent read.


Gnoopy, Once Again it depends on the weight of the real A/C. For those that are under 9-10K most everything can be done in the airfile and a single A/C can cover both bases. The Aircraft.cfg gets the real weight of the A/C and the MOI's are produced using a set of formulas.

MOI = Empty Weight * (D^2/K)
For Pitch D=Length in Feet, K=810
For Roll D=Wingspan in Feet, K=1870
For Yaw D=0.5 * (Length + Wingspan), K=770

I've used these formulas with pretty good results on A/C <10K Lbs.

As far as the Heavies, as Philippe said, it's harder to get them to both stay within the A/C's realistic flight envelope and make them do their assigned tasks than making a Player flown A/C realistic. The Typhoon was the hardest, it had to attack in every way you can think of, Dive Bomb, Glide Bomb, Rocket and Strafe both static and moving targets. When push came to shove we had to make two. It was just not possible to satisfy both objectives with a single A/C. We’ve created hundreds of test missions with the sole purpose of testing an AI flown A/C's ability to perform tasks, both AI only formations and player controlled AI's. You are free to believe me or not when I tell you that A/C >10K Lbs with high MOI's will not perform certain tasks well or at all no matter what you do to the airfile. Please, by all means prove me wrong, I wish I was and that no one would ever have to have an AI Only A/C in their A/C Folder.

Until such time that someone can disprove all of the countless hours of testing and tweaking we’ve done on the heavies I’m sticking to my guns, you need two A/C. When someone creates an airfile that can do the following I will publicly concede the point, grovel at their feet, buy them a beer and raise a high toast to them.

Here’s the criteria the A/C must meet or exceed. (Good Candidates, TBF/M Avenger, Vultee Vengeance, P-38, P-47, B-25G/H/J Gunship Mitchell, A-20 Havoc, A-26 Invader, Mosquito, Beaufighter, Bf-110, Ju-88, Me-210/410, Ki-102b. Boarder Line Candidates that are a little under the weight spec but would fit the bill because of their high MOI’s: Typhoon/Tempest, Ki-45 KAIb, Ju-87, Hs-129, Il-2)

1. The A/C must retain a realistic flight envelope for its type.
2. Player and AI are the same A/C, same Airfile, same Aircraft.cfg file
3. Empty Weight >10,000lbs in Aircraft.cfg
4. E/W Pitch MOI > 18000 in Aircraft.cfg
5. E/W Roll MOI > 14000 in Aircraft.cfg
6. E/W Yaw MOI > 26000 in Aircraft.cfg
7. Must pass the "All Around Surface Attack Test Mission".

This mission is a battery of tests that includes.
a. Drop Bombs on Evading Ships. A/C must drop bombs within 3 passes.
DD Target
b. Firing Rockets at Evading Ships (If applicable to A/C). 3 Pass Rule Applies.
Barge Target
c. Strafing Evading Ships making consecutive low angle runs starting from <2000’. Barge Target must be destroyed.
d. Dropping Bombs on Moving Land Vehicles. A/C must drop bombs within 3 passes.
e. Firing Rockets at Moving Land Vehicles (If applicable to A/C). 3 Pass Rule Applies
f. Strafe Moving Land Vehicles making low angle runs starting from <2000’ Tank Target must be destroyed.
g. Destroy by Strafing 10 Ground Infrastructure Objects Ranging in size from Hut to Large Hangar Type A. Low angle attack rule applies.
h. Must be able to T/O and Land.
i. All above tasks must be completed as both an AI only formation and a player controlled AI.

There are no tricks to the mission like putting objects at the base of a mountain where you know the AI’s can’t get at them or other such silly nonsense. It is a fair and complete test of an AI’s ability to perform the various Surface Attack tasks that it would have been assigned during WWII. The A/C’s Category in the DP will be set up so that it can perform all of the assigned tasks. No tricks there either, if the real A/C wasn’t routinely used for the stated objective that portion of the test is thrown out. If you think the test is unrealistic first consider this. Heavy Fighters, especially US, British and German A/C were more than just Dogfighters. They were Multi-Roll Weapons platforms often being assigned everything from Air Superiority and Escort Missions to Anti-Shipping and Ground Support Strikes. They were asked to play many different rolls, we expect our CFS2 representations to do likewise even if it does take two of the same A/C to do it! LOL

[nearmiss]

Farmboy's AI Commentary:

The AI behavior as it relates to the flight model can be solved pretty easily, in terms of what modifications need to be done to flight model itself. The problem is that you have to begin with a good flight model and then make the needed changes. The problems related to taking off like rocketships or not taking off at all are all related to having a good flight model. The MOIs of course need to correct but these are easily calculated with a number of programs available out there, or can be estimated quite accurately by evaluating a similar aircraft. The lift and moment curves for the wing need to be correct, the horsepower needs to be right, and the thrust curves for the prop need to be right. If these things are whacked out the plane will fly accordingly, that is to say, whacky.

The best way to make a good flight model for a CFS2 plane, if you want it to be a good AI performer, is to begin with a stock flight model from a similar plane. The stock flight models, despite a lot of criticism from some folks, have a lot going for them. They have the basics down, and with some modifications will do very well as a basis for any similar aircraft. I did some testing on the stock Zero, it had both the engine horsepower and top speed at altitude nailed, right on the button. And the general flight charactistics are quite good. The same is generally true for most of the other planes although there are some odd things going on with a couple of them.

To do a good, quick and dirty .air file for your plane, just change a few parameters in the stock .air file that is most like the plane you are building and use that. Enter new information for the wing and tail feathers, engine locations, and engine/propeller, make sure you put good information into the aircraft.cfg file about the weight and balance, and generally get the basics squared away. This approach will work real well for most folks.

All that said, it is entirely possible to build a custom "boutique" flight model with great accuracy in all parameters, and still have it perform beautifully as an AI. I'm just finishing up a P-51D, a custom job, and it flies "by the book", matching up with all the published data for the plane's performance, including time to altitude as well as the initial climb rate, which is what most folks take to be the only meaningful climb parameter (not so!). Flat spins and accelerated stalls are also part of the package, as well as stall, roll, acceleration, and dive characteristics. But the plane works extremely well as an AI unit. I've been testing it extensively and thus far it just behaves beautifully; takeoffs are very realistic, it's rock stable as a wingman, and it attacks and destroys assigned targets as it should.

There is one parameter in the .air file -- and ONLY one, as far as I have determined -- that affects an AI aircraft's ability to attack and destroy its target. I'm sure some folks have come across an add-on aircraft or two that seem OK as AI, but they just dive at the target, failing to release their ordnance or attack properly. I did a lot of research in determining what field was the culprit, cutting/pasting/changing fields one by one until I landed on the guilty party. Turns out it's pretty simple. The main wing needs to have 1.0 degree of incidence. Nothing else had any effect. The .dp file, of course, needs to be set up properly as well, but that's easy enough to do, by just copying the misc data from a similar plane and only changing the parameters that need to be changed, such as the number of crew, speeds, altitudes, and such. The unit family and category are key here. The gun button needs to be set to "guns" and not "cannons", also.

There were some comments earlier about .air file modification that I want to remark on. First, I would say that there are few, if any, "hard" settings in a flight model. The simulation engine is flawed in certain respects, at least in terms of accurately duplicating aircraft behavior over a wide range of conditions. Thus, manipulation of parameters that would appear to be carved in stone is often necessary in order to arrive at accurate behavior and performance. One of these areas is the way power is developed by the engine as the aircraft climbs. At a steady power setting, say 50 inches and 2600 RPM for a given aircraft, the measured horsepower will gradually increase with altitude rather than remaining constant. So if you set your engine HP to the specified number at sea level (for example, 1,200 HP at 49" of boost for an R-1830), that engine will make something like 1,300 HP at 25,000 feet at the same power setting. This appears to be due to the fact that the air temperature decreases with altitude, creating a denser charge at a given manifold pressure. This makes sense in terms of pure physics but on the real aircraft, carburetor/inlet air temperature was kept within a narrow band, which meant essentially consistent performance regardless of altitude. This variation changes depending on the type of induction system you choose (either fuel injected, carbureted, or pressure carbureted) but it is there nonetheless. The problem with this is that the rate of climb does not fall off as it should, because the thinner air and decreasing prop efficiency is compensated for by the increasing horsepower. The only way around this is to decrease the propeller thrust accordingly in Section 511, a value that normally should be a "hard" number. This applies to just about everything in the flight model.

The other bit had to do with drag figures. Drag figures for quite a few WWII-era aircraft are available; for example, the P-51D has a zero-lift drag value of 0.016, while the B-24 comes in at 0.0406, and the B-29 at 0.0241. The Mustang is about the slipperiest airplane out there, in terms of prop fighters. In the absence of an actual value, there is a way to calculate Cd0, but it's just as effective to estimate drag based on the range of values and types of aircraft they apply to. Figure most fighters to be in the range of 0.025-0.035, and bombers to be from 0.035-0.055. Take a good look at the aircraft, see how many guns, blisters, and other protrusions it has, and find out what sort of wing it uses. Radial engines add a lot of drag as compared to inline liquid-cooled types, so figure another 20-25 percent drag for this type of engine compared to the inlines. Most of the stock aircraft use identical Cd0 figures and adjust the top speed using Section 430, not such a good way to do it but "quick and dirty" all the same.

08/28/04

[nearmiss]

This document is to aid mission builders in preventing unwanted behavior from Non-Player AI aircraft on Strike Missions against Static Targets. The most common of which is when the AI's pass by their intended target and then turn around 180* and come back to attack. This document covers Aircraft Categories 1 and 4. I used the F6F-3 for Cat.1 and the TBF as the Cat.4 as it was used in a glide bombing role.
MS also made the Kate a Cat.4 aircraft but I can find no historical record of it being used in this role, only as a Torpedo plane and for Level Bombing, aka Cat.5.

About the Test Mission;

1. All formations are of 8 aircraft in a Echelon Right Formation, light bomb load, Ace, High Agress.
2. The WP's leading to the attack are all at the same altitude in the table at 200Knts.
Start Ranges from 2 to 100 miles from the Pre-Attack WP.
Pre-Attack 10 miles from Attack WP.
Attack 3 attack WP's were tested.
1. On nearest Runway on the Line of Attack. Attack X Airfield (Cent)
2. At the Outer Edge of the Airfield just beyond the AA Positions (OE)
3. 1 Mile out from Center of the field. (1M)
All were set to Attack With Bombs.

Legend:
Alt - Altitude of all WP's from Start to Attack.

Action - Action of Formation from the point were they detect their target (about 6 Miles out) to their
Pushover Point. These Are:
C = Climb all the way to pushover.
C-L = Climb then level out prior to pushover.
C-D = Climb then Descend prior to pushover.
L = Level all the way to target
L-D = Level then descend to pushover. This is the perfect SBD Speed run for Cat3.

Cent, OE, 1M - The number of Aircraft out of 8 that attacked on the first pass by Attack WP Placement.

Attack - The Angle of the 1st pass Attack, these are:
GB = Glide Bomb - very shallow attack run.
MD = Medium Dive - Slightly steeper than a GB Run.
SD = Steep Dive - Steep Dive Bombing Angle. Only Cat3 uses this.

Table: CATEGORY 1
Note: I added an additional test for the higher Altitudes with the Attack WP 2 Miles from the Center of the Target Area.

Formation Alt Action Cent OE 1M 2M Attack

Able 500 C-L 8 8 8 N/A GB

Baker 1000 C-L 8 8 8 N/A GB

Condor 2500 C 5 8 8 N/A GB

Dog 3000 C 5 7 8 N/A GB

Eagle 4000 C 2 6 8 N/A GB

Falcon 5000 C 1 3 8 N/A GB

Goblin 6000 C 1 1 7 N/A GB

Hawk 7000 C-L 0 1 5 8 MD

Indy 8000 C-L 0 0 3 8 MD

Joker 9000 C-D 0 0 4 8 MD

Kilroy 10000 L-D 0 0 2 8 MD

Limbo 11000 L-D 0 0 2 8 MD

Mable 12000 L-D 0 0 0 5 MD

Neptune 15000 L-D 0 0 0 6 MD

Ottowa 18000 L-D 0 0 0 5 MD


Table: CATEGORY 4
Note: I added an additional test for the higher Altitudes with the Attack WP 5 Miles from the Center of the Target Area.

Formation Alt Action Cent OE 1M 5M Attack

Able 500 C-L 8 8 8 N/A GB

Baker 1000 C-L 8 8 8 N/A GB

Condor 2500 C-L 7 8 8 N/A GB

Dog 3000 C-L 6 8 8 N/A GB

Eagle 4000 C-L 5 7 8 N/A GB

Falcon 5000 C-L 2 5 8 N/A MD

Goblin 6000 C 1 3 8 N/A MD

Hawk 7000 C-L 0 2 6 8 MD

Indy 8000 C-L 0 1 5 8 MD

Joker 9000 C-D 0 0 3 8 MD

Kilroy 10000 C-D 0 0 2 8 MD

Limbo 11000 L-D 0 0 2 8 MD

Mable 12000 L-D 0 0 0 8 MD

Neptune 15000 L-D 0 0 0 8 MD

Ottowa 18000 L-D 0 0 0 8 MD



Key Points:
WP Distance is the Major key for these 2 Cat's as can be clearly seen with the Attack WP moved back to 2 and Miles out.
You will need to add an additional WP in line with the target but the results are worth it.

Start WP [---5Miles---] Post Target WP
S----------A------------T----------P
Attack WP Target Area (not a WP)

[nearmiss]

This document is to aid mission builders in preventing unwanted behavior from Non-Player AI aircraft on Strike Missions against Static Targets. The most common of which is when the AI's pass by their intended target and then turn around 180* and come back to attack. This document covers Aircraft
Categories 3 and 10. 10 is the category that MS assigned to the Dive bombers in the stock aircraft. As you will see category 3 does a much more realistic job of replicating the actions of real WWII
Dive bombers.

About the Test Mission;

1. All formations are of 8 aircraft in a Echelon Right Formation, light bomb load, Ace, High Agress.
2. The WP's leading to the attack are all at the same altitude in the table at 200Knts.
Start Ranges from 2 to 100 miles from the Pre-Attack WP.
Pre-Attack 10 miles from Attack WP.
Attack 3 attack WP's were tested.
1. On nearest Runway on the Line of Attack. Attack X Airfield (Cent)
2. At the Outer Edge of the Airfield just beyond the AA Positions (OE)
3. 1 Mile out from Center of the field. (1M)
All were set to Attack With Bombs.

Legend:
Alt - Altitude of all WP's from Start to Attack.

Action - Action of Formation from the point were they detect their target (about 6 Miles out) to their
Pushover Point. These Are:
C = Climb all the way to pushover.
C-L = Climb then level out prior to pushover.
C-D = Climb then Descend prior to pushover.
L = Level all the way to target
L-D = Level then descend to pushover. This is the perfect SBD Speed run for Cat3.

Cent, OE, 1M - The number of Aircraft out of 8 that attacked on the first pass by Attack WP Placement.

Attack - The Angle of the 1st pass Attack, these are:
GB = Glide Bomb - very shallow attack run.
MD = Medium Dive - Slightly steeper than a GB Run.
SD = Steep Dive - Steep Dive Bombing Angle. Only Cat3 uses this.

Table: CATEGORY 10

Formation Alt Action Cent OE 1M Attack

Able 500 C 8 8 8 GB

Baker 1000 C 8 8 8 GB

Condor 2500 C-L 7 8 8 GB

Dog 3000 C-L 6 8 8 GB

Eagle 4000 C-L 5 8 8 GB

Falcon 5000 C-L 2 6 8 GB

Goblin 6000 C-L 1 4 8 GB

Hawk 7000 C 0 2 6 GB

Indy 8000 C 0 0 5 MD

Joker 9000 C-D 0 0 2 MD

Kilroy 10000 L-D 0 0 4 MD

Limbo 11000 L-D 0 0 2 MD

Mable 12000 L-D 0 0 0 N/A

Neptune 15000 L-D 0 0 0 N/A

Ottowa 18000 L-D 0 0 0 N/A


Table: CATEGORY 3

Formation Alt Action Cent OE 1M Attack
Able 500 C 8 8 8 GB

Baker 1000 C 8 8 8 GB

Condor 2500 C-L 7 8 8 GB

Dog 3000 C-L 7 8 8 GB

Eagle 4000 C-L 4 7 8 GB

Falcon 5000 C-L 2 5 8 GB

Goblin 6000 C-L 1 3 8 GB

Hawk 7000 C 0 2 7 MD

Indy 8000 C 0 1 5 MD

Joker 9000 C-D 8 8 8 SD

Kilroy 10000 C-D 8 8 8 SD

Limbo 11000 L-D 8 8 8 SD

Mable 12000 L-D 8 8 8 SD

Neptune 15000 L-D 8 8 8 SD

Ottowa 18000 L-D 8 8 8 SD


As you can see Category 3 Acts much better at the real altitudes used by WWII Dive Bombers =>9000'.
Combined with the better Angle of Attack I have dumper Category 10 for all DB's.

Other Points:

There are four factors that determine an AI's ability to attack on the first pass.
1. The Category of the Aircraft.
2. Altitude of the formation leading up to and at the Attack WP.
3. Attack WP Distance from the objects being attacked. This affects Cat10 at any
altitude and Cat3 at lower altitudes. 9000'-18000' Cat3 always attacks.
The further you can move the attack WP away from the targets the better chance they have of a 1st Pass Attack.
4. Depth of the target, the deeper the target the better chance the AI's have of
acquiring a target and attacking on the first pass.

Cat 10 A/C really work better as Torpedo Bombers than they do as Dive Bombers. Though Cat4 does a
better job than Cat10 does at Glide Bombing like the TBF/TBM's did. More on this in the next report.
This document only relates to Static Targets. Moving objects, like Ships use a different attack routine
that is much less flexible which is "Get behind the target at 5000' or 6500' and attack in a medium
dive." Even Cat3 A/C will spiral down to 6500' before they attempt to attack. That great Steep Dive
only happens with static targets. So Far!!! LOL. The moving target reports will follow.

Next Report - Category 1 and 4 Aircraft against static targets.

[nearmiss]

CFS-2 Thoughts

--------------------------------------------------------------------------------

Improving CFS-2,,


Any body that bought CFS-2 in the last 2 years , got a box with 2 CD,s
and no manuel, And they post questions on why the empty aircraft folders were included
in the install,, Ai (Empty) pL (Empty) and player aircraft,, Most people including myself
said WT Hell are they in here for ,, Not in the old game Original install ,, and when we
found no aircraft inside the empty folders we deleted them..Since the powers that be M$s
forgot to include a readme as to why the change,, (must have known flightsimmers never
read those things) The normal response was Delete , Delete, Delete... Lets have a show of
Hands here , how many of us did just that , (Raises right hand)......
What if they M$s had a reason for doing the install with all them empty folders ,,
And they just plain figured we was smart enough to read between the lines..

Lets look at what we have 7 playable planes and 7 Ai planes,,
Why were the empty folders named the same as player aircraft-- ??
Simple explanation would be different sqd codes same plane... maybe..
What if they intended more complex differences --??
Airfiles,, DP,, Different Squadron,s (same plane)
Will it work ??? Yes you can do it...
Think about that answer -----

Step back to 7 player and 7 Ai planes
BUT 3 versions---- Player,, Player A1,, Enemy A1..
What are the differences in each catagory,,,
Normal response ,, Player you fly,,A1 Player you yell at for being stupid !!, Comp Controlled
And the A1 enemy you curse for shooting off your left wing,, Comp Controlled,,

1/Simple question ,, Does the same plane fly different under comp controll,,
Answer =Yes----Why---Answer = Programed Inputs,,,, or PI for short

2/Is the same DP handeled differently when used by the A1 ???
AWS= YES-----READ THAT AGAIN---Y E S-----------How come ??Aws= Programed Inputs


3/Is there a difference in A1 Player to A1 enemy flight control ???
Aws--Using the same plane , UnKnown ???? Call Gates !!

4/Is there a difference in A1 Player To A1 enemy in DP control
Aws= Yes----Please explain--!!
Aws= A1 Player (Fighter) will fire guns at 500 mters
Aws= A1 Enemy (Fighter) will fire guns at 850 mters
Why ?? Aws= Programed Inputs

Ok Ok , you in the 3 row with the hand up ?? Whats your question ??

"The Dp file has a section where you can set the distance the guns open up at......"
Teacher: Very good Johnny , your doing your homework,, Did you Know that only applys
to catagory 2,4,5,10 , in other words only bomber gunners can vary the engagment distance..
(Fighters) are catagory 1 and therefore excluded from engagment distance of the Dp file....
Aws= Programed Inputs,,,,


5/Anybody thats played CFS-2 longer than 3 weeks has put aircraft into it..
Some are done to perfection,, other,s need a few changes to suit our own personel taste..
That New P51b is nice but the sites too low ,,Gotta fix that,, You open up Dp edit,, (The
Aces use notepad by the way),,Real guns were converged to 100 ft/mters and centered,,
But you want it at 250 ft,, 2 clicks later and Dp edit fixes the problem..Gota test this
baby out,, Qcombat and an A1 enemy crosses your gunsite at 275 ft and gets smoked,, Cool,,
Piece of cake,, Make A mission with 3 P51bs and watch your handy work..Lock on the target
and sit back and watch your A1 players , tackel the job , 2 hrs and 45 minutes later you
realize that your New P51b A1 players cann,t hit the barn door,,You recheck the Dp file ,
Dp Edit confirms the fact that indeed you set convergence to 250 ft,, You delete the CDP
and refly the mission again,,Toss out the Barn door cause now you find out the A1 player
cann't even hit the barn.. Go read section 4= A1-player (Fighter) fire guns from
500 ft /mter (PI) ,, Anybody with a degree in Trig could figuer it out.. I failed Math,,
But I would guess your setting at 250 meters,, used by A1-player that opens up at 500 mtrs
would be the equivelant of shooting a mouse with a 12 gauge from half a mile..

6/New mission file gets posted , 4 P-47d aircraft Player formation take off from Leiston..
But the A1 player aircraft take off faster than the player aircraft can , a collusion occurs
Player gets killed End of mission,, Its not the mission setup its the A1 player airfile that
needs repair,,Go back to No 1,, The same plane flys different under A1 control,, Sure you
could change out the plane and try again,, open the mission file and add some distance
for the 3 A1 player,, I opted to fix the airfile but added it to the mission as a seperate
A1-player aircraft,,, And Keep the old airfile as only a player version,,

7/ By now your wondering where is this headed,,
Well I have an idea..
Why not setup three independant aircraft folders,,


Player Aircraft =P51d
A1 Player Aircraft =-P51d
A1 Enemy Aircraft =-A6M2_Zero

Advantages are Retaining A high fidelity detail model for player use,,
Differnt Code letters , Sqd ,for A1 player,
Custom 1 % airfiles or better can be used by player,,
Custom Dp ranges can be set 100 ft and not effect A1 player performance,,

A1 Player can use a seperate airfile and Dp file for optimum performance
A1 Player can use a lower Poly model for larger Formations , better FPS

Enemy planes could be flown on custom air files and DP
A1 Enemy would have optimum Airfiles and Dp files
A1 Enemy could use lower poly models For better FPs

Missions could be designed to add in an A1 player seperate aircraft or added yourself
to older missions,, The stock aircraft are pretty good ,,but its the add ons that
need adjustment ,, With the veriety of different aircraft now available , We have
room for improvement and Theaters ,, Ken Stallings Eastern Front ,, Malta,,
North Africa,, Italy,, Sicily,,Crete,, Not to mention all the Pacific possibilities,,

Aircraft designers are building A1 aircraft,,Merlins Spit Mk I ,, The Mossey Gmax series,,
Wozza,s special A1,s..Gmax Typhoon,, -----

So this is not a new Idea!!
But rather a post to inform New members about the possibilites inherent in this
Sim,,, Sarg Willy

As a mission builder, I absolutely confirm and agree with what Sarg Willy said. I observed the same and this led me to use modified stock airfiles for AI aircrafts in my Milne Bay campaign.

AI aircrafts need a dedicated air file, to be able to take off and land, even with windy conditions (1% airfile cannot be handled by the AI), to be able to stabilise the aircraft and shoot (most add-ons don't shoot or rarely shoot).

AI aircrafts need a dedicated DP, with guns harmonized at a different distance and elevation, and all linked to trigger 0 (i.e. machine guns, most AI don't use their cannons).

I do not totally agree with AI needing a different 3D model for better framerates, as a good multires model provides the same (try Bruno Duffort's or Paul Rebuffat's ...) ... and fly in close formation with a nicely detailled aircraft is beautifull. You see more often other aircrafts than yours when playing !

As far as I know, only one designer has provided AI dedicated aircrafts up to now : P. Burnage with his Mosquitos, Typhoons and Tempests.


Guy


In CFs2 When shot down, the plane doubles itself

--------------------------------------------------------------------------------

In response to a question about certain cfs2 aircraft..."When shot down, it doubles itself! What can be done about that?"

Shessi replied:

The prob is in the DP file.
1) Go to your aircraft file and delete the CDP file (this is a computer generated copy of the DP file) and will do no harm in deleting. If you do not when you change the DP file the old CDP will be loaded.

2) Open the DP file and scroll down to the EFFECTS section. At the end of each effect is a line which usually ends in eg break,,2.

3) Go through each EFFECT and delete the number from each one. Save the file when edited.

4) You will now have an aircraft which can be shot down and crash as a 'whole' aircraft.

Also according to Jagdflieger "...the problem stems from using a CFS 1 DP file as your base DP file for CFS 2. Apparently CFS 1 read that final number as a break point, while CFS 2 treats it as a command to create a whole new plane."

Great tip for Cfs2 thanks Shessi!

[nearmiss]

Sarge Willy,

Well Thankyou Dv and Bc I didn,t have the p-38 files ,, I,m sure they,ll come in handy,,-----------------------------------------------------------
I use a CFS-1 p47r air file ,, Takes off without running over the leader ,, Uses a P51 D prop so has less torque !! That allows KM,s P47d or R to have that long rollout with out fear of stupid Ai,s chewing your tail off ,, Before someone jumps all over that as not fast enough ,, Remember the Ai are tailored from the DP file Misc Section ,, You want faster than up the Max and Cruise values in the Dp file ,,, The Ai can handel the flight dynamics much better ,, With an accurate Dp gun file these can chew up 109,s very well ,,
To use on Km model ,, You must copy over the contact points from KM ,, Load the Ai aircraft and check the center of Gravity ,, adjust to 28 % Happy hunting ,,


Bearcat,

Thanks for that explanation of the dihedral effect. I haven't messed with the 1101 section in the airfile yet. I've generally kept my modifications as to roll rate limited to the aircraft config. file, that is increasing the roll effectiveness in the flight tuning. As to the MOI, I confess I've always found this something of a mysterious process. I know how to compute the values well enough, but I'm all at sea when it comes to modifiying them for something like an aileron boost. I've noticed the very behavior you describe with my own experimental AI P-38; they do quite a bit of banking before lining up for a shot but I didn't know why before now. I may have asked this before, but I lost quite of few of my tutorial files oveer the holidays -- is there any particular process you use to come up with these numbers? Is it trial and error?

JAMES


Bearcat241,

Three places basically: (1) in the cfg by increasing aileron effectiveness and (2) lowering empty_weight_roll_MOI. (3) In the *.air its in section 1101 > dihedral effect. Main wing dihedral has a very direct influence on roll rate. Higher dihedral translate to snappier rolling, but also raises instability and increases twitchiness at low speed, so for player a/c known as historical fast rollers, this should be balanced. Dihedral is the amount of upward tilt a wing has in relation to the fuselage. In a frontal view, the wings seem to be slightly bent upwards at the roots. Flatter dihedral makes for slower rolling but also enhances stability and lowers stall tendencies, making the plane more forgiving. Roll stability doesn't matter as much with AI, but too much roll acceleration will make it tougher for wingmen to get their firing solutions when up close. They'll waste extra time and effort trying to compensate in yanking/banking maneuvering. You'll see them roll one way on a bogey's tail but roll too far and then roll back opposite but again too much. And sometimes this can go on forever if their in rookie or veteran levels...aces naturally do better.

It was once more like trial and error, but after so many *.air and cfg mods you get an instinctive feeling for what will work on any particular type and weight of a/c. You develop a knack for visualizing all the known variables of a particular configuration at once in your head and can just sort of "know" how far to go before you start creating a UFO .

But for modest and truer dihedral effect on lightweight (3500-7000 lb) planes remembered as fast-rollers, try not to exceed 180 in tweaking. This will keep things balanced. For the middle and heavier weight fighters (8000-17000), max 200 should be more than enough. For added realism, you should offset these tweaks by adding or copying from another file record 546 - Roll rate per G. If you do the math right, this parameter creates a roll dampening effect as speed builds, so that the roll rate gradually slows down in the higher speed ranges, like in reality. This record contains 5 table pairs or settings which represent 5 speed ranges. a setting of 1.00000 in the 1st range means the plane will retain 100% of its roll effectiveness at lowest speeds. For the highest speeds (lower down in the list of ranges), you'll see dramatic decreases, sometimes as low as 0.20000 or an 80% reduction in aileron roll effectiveness.

Sarge,

Thank you very much for posting the P47 zip. I've installed my airfile, and inserted the KM contact points in the aircraft config file. My CG is 25.4. What is the significance of the 28 CG, and what values to I change to get it?

Many Thanks,

JAMES

Look inside aircraft CFG ,, for--

station_load.0=200, -2.0, 0.0, 3.0//Pilot Weight (lbs),

adjust the ----station load .0=200, -3.5 ,0.0, 3.0 ------to change the C of G from 26 to 28 ,,,,, reason shifting weight further back helps keep aircraft from nose over on Ai Landing ,,,

[nearmiss]

Sorry this one had more to do with Mission Building

Maybe that could be a next information project worth pursuing.

[nearmiss]

Sorry this one had more to do with Mission Building

[achim27619]

nearmiss

thanks for putting all those ideas into one post. I already have the odd info but in your collections there are tons of nuggets to be found. I just copied the lot and read it.

If you have more of those, please, keep posting

Thanks again

Achim

[sc7500]

BRAVO, NearMiss !

More gems to add to my "Must Have" reference library !

Thank you
SC
:kilroy:

[nearmiss]

I hope for everyone's sake this is just the tip of the iceberg.

I would love to read new things about improving the CFS2 AI performance.

In fact, just reading something that is worded differently can often make a difference in how you deal with things.

There are plenty of postings about AI, but the problem for researching has always been... you can not search on anything with just 2 characters.

I'm just sure there are plenty more postings with heaps of good information.

I'd be the first to agree the GOMERAI thing maybe kinda dumb, but it is easy to remember and it will definitely make searching easier.

This board like so many others has been hacked so many times, tons of information has been lost over the years. Don't think I'm ragging on the admins for this site, I am NOT. I've admin a couple of sites and you just can't outsmart everyone all the time, a ferret of unusual ilk is always out there capable to crap the works.

When I get more time I plan to really do some searches on CFS type forums to see what I can turn up. Right now, it was my hope to stimulate others to share and help with this.

AI performance in CFS2 is a BIG issue, everything from the FM of AI aircraft to how they react to commands, how the work in given situations all can add up to a miserable or good experience with CFS2.

:wavey:

If miami eagle gets interested in this project I'm sure we'll have tons of links and references very soon. Miamieagle is a world class searcher, and I'm grateful for it.

[DeathFromAbove]

Nearmiss

I want to thank you for having the information you've achieved, it's a shame that so much information was lost on the forum.

Since I became a member I've always posted and researched information that was related to A.I.

I'm sure this is a big NO NO here but since looking through the CFS2 background program files I can see how improvements can be made if it was easier to edit.


P.S.

I don't think the GOMERAI thing is a bad idea, maybe something that rolls off the tongue more lol.
I remember I tried to make an A.I. Consolidation Thread with all the links and some of the info from the forum, it would be great to have something similar again as a reference point. Like as you've done here.

[merlin2]

This discussion is a great example of the usefulness we would have for a CFS2 wiki. It would be a place where all the details could be stored, compiled, organized and added to. Sounds like a formidable task, but it doesn't have to be. It could begin on a small scale.

If I knew how to build and run such a wiki . . .

any possibility of developing something like that at SOH?

merlin2

[dvslats]

Pen32win


Part 1


Well it comes down to two things.

1. As with the stock A/C they work best if the guns are harmonized to 300 Meters (984.25 Ft.) I use AimIt to set this. This point is not absolutely critical as I've got some A/C that are set out to 500 Meters for the convergence point. But for the most part 300 Meters works best.



2. The vertical component is the part that is really critical. Many A/C designers aim the guns of their A/C in accordance with the sight they've built into the VC. I've seen Pitch Numbers as high as +/- 1.2~. The thing is the AI's don't like it to be much more than +/- 0.4 and the closer to +/-0.10 it is the better it tends to act. What I do is get the Convergence set with Aim it and then set the pitch to 0.00. Then I start the test flights. From there it's just a matter of seeing where the AI's are shooting. Remember, to get the guns to shoot higher you use a lower number and vise versa. I start out adjusting in increments of 0.10. You end up doing a lot of chasing AI’s around and going into Chase View to watch the target they’re trying to hit. Trial and Error.

Part 2


FW-190A-5:
[GUNSTATIONS]
; Machine Guns
gunstation.0=0,45,1,0.055,885,2,0.01,500,1,25,1d1* 9,-0.12,0.77,0.9,0.05575,0,0.02299,0,0,0,0,0.39
; Machine Guns
gunstation.1=0,45,1,0.055,885,2,0.01,500,1,25,1d1* 9,0.12,0.77,0.9,0.05575,0,-0.02299,0,0,0,0,0.39
; Left Cannon
gunstation.2=1,47,2,0.093,762,2,0.05,500,4,30,1d1* 44,-0.931,-0.301,1.1,-0.21342,0,0.17846,0,0,0,0,3.5
; Left Guns
gunstation.3=1,24,2,0.115,600,2,0.04,500,4,30,1d1* 37,-2.39,-0.148,0.803,-0.3412,0,0.45767,0,0,0,0,4.73
; Right Cannon
gunstation.4=1,48,2,0.093,762,2,0.05,500,4,30,1d1* 44,0.935,-0.301,1.1,-0.21342,0,-0.17923,0,0,0,0,3.5
; Right Guns
gunstation.5=1,31,2,0.115,600,2,0.04,500,4,30,1d1* 37,2.39,-0.148,0.803,-0.3412,0,-0.45767,0,0,0,0,4.73
; Bomb Release
gunstation.6=3,20,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0
[guns.0]
gun.0=900,0,0,0,0,0,1,900
[guns.1]
gun.0=900,0,0,0,0,0,1,900
[guns.2]
gun.0=250,0,0,0,0,0,1,250
[guns.3]
gun.0=90,0,0,0,0,0,1,90
[guns.4]
gun.0=250,0,0,0,0,0,1,250
[guns.5]
gun.0=90,0,0,0,0,0,1,90
[guns.6]
gun.0=22,0,0,0,0,0,0,0

FW-190A-6:
[GUNSTATIONS]
; Machine Guns
gunstation.0=0,45,1,0.055,885,2,0.01,500,1,25,1d1* 9,-0.12,0.77,0.9,0.05575,0,0.02299,0,0,0,0,0.39
; Machine Guns
gunstation.1=0,45,1,0.055,885,2,0.01,500,1,25,1d1* 9,0.12,0.77,0.9,0.05575,0,-0.02299,0,0,0,0,0.39
; Left Cannon
gunstation.2=1,47,2,0.093,762,2,0.05,500,4,30,1d1* 44,-0.931,-0.301,1.1,-0.21342,0,0.17846,0,0,0,0,4.73
; Left Guns
gunstation.3=1,24,2,0.085,762,2,0.05,500,4,30,1d1* 44,-2.39,-0.148,1.1,-0.34120,0,0.45755,0,0,0,0,4.73
; Right Cannon
gunstation.4=1,48,2,0.093,762,2,0.05,500,4,30,1d1* 44,0.935,-0.301,1.1,-0.21342,0,-0.17923,0,0,0,0,4.73
; Right Guns
gunstation.5=1,31,2,0.085,762,2,0.05,500,4,30,1d1* 44,2.39,-0.148,1.1,-0.34120,0,-0.45767,0,0,0,0,4.73
; Bomb Release
gunstation.6=3,20,8,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0
[guns.0]
gun.0=900,0,0,0,0,0,1,900
[guns.1]
gun.0=900,0,0,0,0,0,1,900
[guns.2]
gun.0=250,0,0,0,0,0,1,250
[guns.3]
gun.0=140,0,0,0,0,0,1,140
[guns.4]
gun.0=250,0,0,0,0,0,1,250
[guns.5]
gun.0=140,0,0,0,0,0,1,140
[guns.6]
gun.0=22,0,0,0,0,0,0,0


:applause:

[DeathFromAbove]

I recommend downloading and reading the FS2000 SDK.

http://files.fsnordic.net/Flight_Simulator/Miscellaneous/SDK/

" FS2000 Aircraft Software Development Kit "


Gives you some more information on editing the A/C, not all of it exactly pretains to CFS but most does.

[bobhegf]

Fellows I think that this is a very important post. I also feel that it will help the new fellows out.:bump:

[miamieagle]

I just read this post today.

Thank you for your compliment!

I will help you with what I can!:icon_lol:

PS great post!

[rotagen]

I know it's old but so am I....This thread should be posted on the top (stickied) somewhere... lots of good info , I saved the whole thing.

Just getting back into cfs2 here... the game industry has gone down a black hole of multiplayer conformity ... conformity in that the games I enjoy, generally combat sims with vehicles..(I find the first person stuff silly, just the way I am)... the games are pretty much fast-food "this is what you get so be satisfied and don't screw with it" stuff.

The equivalent of UC Davis grad school (conveyer belt regurgitate the textbook stuff) which I left and went to a real school where the first tests included "all right smart ass, show us your grasp of the subject, show some independent thought and design an experiment to determine the answers".... I can Digress with the best of them JEEZE !

Anyhew..... The HUGE weakness in CFS2 (CFS3 was even worse, but some desperate folks made new sims based on it)... is the GOMER-AI. I'm talking wingmen and enemies wandering around on a sunday drive with no destination.

If only the hardcore programming folks were allowed to tweak the heck out of the engine... Look, Jane's WW2F (say what you will but they did everything right... play it with ACD, 30 flights in near proximity, enjoy the adrenaline/endorphins/death grip on the joystick mayhem with good long lasting ground fires, good loud "splosions" etc... It has AI that will chop you to shreds and they're not messing around.....It's why soccer is the best sport the perfect mix of anarchy and order (but now it's politically correct with crap refs calling everything, and no slide tackles... wimps)...

There I go digressing again. But cfs2 has the best most beautiful damage modelling on the planes... good terrain... huge add-ons , etc... IL-2 got me interested for quite some time but the terrain is a joke as is the DM - planes explode into a million pieces, never ground crash, and their greedy business model... I have a right to gripe because I spent hours downloading the Red core stuff to fly WW1 and golden age biplanes etc... only to find their flight models did not function worth a darn...

Here's something interesting, I'm also an RC plane geek... I have made probably 30 scratch-built planes and one thing I learned (mentioned earlier in this thread) is that the wing incidence angle (and relation to horizontal stab) is CRITICAL .. and yes indeed 1 degree positive is the sweet spot.. I once made a negative incidence plane after reading glider guys like it... it was fine until it turned then all hell broke loose and smasherooo.

Moving the cg back yes, makes a plane faster and quicker turning as well as fast climb rate but relatively unstable, reverse for moving forward...

Greater dihedral makes a plane stable in the roll direction (as does a higher wing due to pendulum effect) but also slower turning, but perhaps cfs2 didn't model this perfectly?

Larger wing area = more stability at low speed , there's a very popular easy to make RC plane called the nutball, which is a literally a circle of thin foam with some dihedralled tips and a vertical stab with rudder no ailerons... no kidding.. that thing doesn't care what pitch you give it, or anything else you do, it ain't stalling, and will fly just like a real plane... the bernoulli idea is dead wrong as my Cal Poly classmate proved with actual wind tunnel testing and pressure sensors, the truth is simple but humans hate that... the closest analogy is skipping stones on a pond, with a little of that moving air downward effect...MAN I DIGRESS !

Anyways thanks for this thread perhaps I can start to approach slightly less aimless and drunk AI enemies.