Hi all. Been waiting for VR for the last 2 years.... and the roll rate of AI is insane.. any fix ahead

[Buzzsaw]

On 2/24/2024 at 7:39 PM, Rei-sen said:

Yes, for sure!

 

 

Will look at this... may be another reason why this is happening.

[Dagwoodyt]

On 11/3/2023 at 2:22 PM, Buzzsaw said:

Yes, we are working on AI behaviour, including roll rate.

?

[BENKOE]

[Buzzsaw]

Yes, we have been working on the AI roll behaviour... and yes I said this awhile ago.

 

But as I also mentioned, there are other factors and 'bugs' in the game affecting why AI aircraft perform in a particular fashion.

 

It is not just the FM... which is the same for AI and human... but when you load another set of of processing equations on top of the FM, things get complicated.

[Dagwoodyt]

On 2/15/2024 at 10:19 PM, Buzzsaw said:

AI maneuver to avoid being hit based on their Awareness and Vision ratings.  (value between 1-10....  there are 10 skill levels for AI... most players leave the skill levels at generic values, i.e. Rookie/Average/Veteran/Ace, but they can be set wherever the mission builder wants)

 

If the mission builder sets awareness and vision high, then the AI will vigorously maneuver to avoid being shot and will react as the player brings his guns to bear.

 

If the mission builder sets these values low, then the AI will not notice a hostile aircraft approaching for a shot.

 

Up to the mission builder.

 

11 hours ago, Buzzsaw said:

Yes, we have been working on the AI roll behaviour... and yes I said this awhile ago.

 

But as I also mentioned, there are other factors and 'bugs' in the game affecting why AI aircraft perform in a particular fashion.

 

It is not just the FM... which is the same for AI and human... but when you load another set of of processing equations on top of the FM, things get complicated.

It sounds like there are "processing equations" that AI skill settings do not address. As it stands now AI settings can only be accessed through the FMB. If skill settings are really the issue they need to be accessible to the player before mission start. Is a 12 year old unofficial assessment of skill level setting effects the only available reference?

[Rei-sen]

If AI has the same FM, how come it is able to roll several times faster than player in the same aircraft at the same airspeed and altitude?

[Buzzsaw]

3 hours ago, Rei-sen said:

If AI has the same FM, how come it is able to roll several times faster than player in the same aircraft at the same airspeed and altitude?

All aircraft have different roll rates... depending on type, the speed of the aircraft, etc,

 

At lower speeds a Spitfire generally has a lower roll rate than a Bf-109.  (exception, clipped wing types...  with the different 109 types varying in their roll rate as well... the E model was better than the F at low speeds, but inferior at higher speeds)  When the F models were introduced, they had a better rollrate than the fabric aileron Spitfires at all speeds.  The Spits generally had a better rollrate across the spectrum of speed once they got the metal ailerons... especially at high speed.

 

Rollrate is a function of aileron design shape, with larger, deeper ailerons giving better performance at low speed, and shallower, smaller ailerons performing better at high speeds. Aileron placement on the wing also has an effect... ailerons closer to the tip are more effective.  Also a factor are the effects of compressibility due to airfoil type, as well as the relative rigidity of the aircraft's wing... stiffer wings are more conducive to high rates of roll due to lower flex.  And metal ailerons give better performance at high speeds as there is less 'ballooning' of the surface... although metal types are also inferior at low speeds due to the higher weight and lower response.  Power assisted ailerons such as were introduced with the late P-38J improve high speed performance.

[Volant_Eagle]

4 hours ago, Buzzsaw said:

All aircraft have different roll rates... depending on type, the speed of the aircraft, etc,

 

At lower speeds a Spitfire generally has a lower roll rate than a Bf-109.  (exception, clipped wing types...  with the different 109 types varying in their roll rate as well... the E model was better than the F at low speeds, but inferior at higher speeds)  When the F models were introduced, they had a better rollrate than the fabric aileron Spitfires at all speeds.  The Spits generally had a better rollrate across the spectrum of speed once they got the metal ailerons... especially at high speed.

 

Rollrate is a function of aileron design shape, with larger, deeper ailerons giving better performance at low speed, and shallower, smaller ailerons performing better at high speeds. Aileron placement on the wing also has an effect... ailerons closer to the tip are more effective.  Also a factor are the effects of compressibility due to airfoil type, as well as the relative rigidity of the aircraft's wing... stiffer wings are more conducive to high rates of roll due to lower flex.  And metal ailerons give better performance at high speeds as there is less 'ballooning' of the surface... although metal types are also inferior at low speeds due to the higher weight and lower response.  Power assisted ailerons such as were introduced with the late P-38J improve high speed performance.

All very true, but I think you missed what Rei-sen meant. He said same aircraft at same altitude and same airspeed. He’s not talking about the differences in roll rates between one type of plane and another.

 

Examples:

 

Player in Spitfire IIb at 5,000ft and 200mph vs AI in Spitfire IIb at 5,000ft and 200mph.

 

Player in Bf 109E-4 at 2,000m and 300kph vs AI in Bf 109E-4 at 2,000m and 300kph.

 

Player in Blenheim Mk IV at 5,000 and 200mph vs AI in Blenheim Mk IV at 5,000ft and 200mph.

 

If the flight physics are truly the same between players and AI, then it should be physically impossible to see an AI rolling at a rate any faster than a player could achieve with full aileron deflection in that same plane at the same speed and altitude. This is not the case in game however. The Blenheim is one of the most noticeable offenders I’ve noticed in this regard. Flying it as a player, the roll rate feels like driving a dump truck in comparison to most fighters (which seems realistic), but if I attack a formation of AI Blenheims, they all start jinking around and changing direction like a flight of swallows. They roll at rates that seem to be multiple times what I’ve ever been able achieve with full deflection in a Blenheim. Yes, I can maybe still out roll then in a 109, but just barely and that significantly changes how the fight goes in game compared to real life.

 

Best of luck in tracking down this issue. It’s been complained about forever and it’s hard not to notice. I’m sure a lot of people would be very pleased to finally have this bug fixed.
 

The problem being rooted in something on top of the flight physics does make sense to me. Especially with the Blenheim being one of the most noticeable problems. The Blenheim has a bit of an identity crisis between whether it’s a bomber or a fighter. I don’t notice this weird rolling behavior with bombers who know that they’re bombers. So it seems reasonable that some sort of protocol related to AI is overriding the flight physics.

 

 

[BENKOE]

[Dagwoodyt]

A related phenomenon that spoils immersion in SP is AI's unwillingness to stay in a fight. Many times I have seen this behavior in flying small fighter vs fighter engagments e.g. flying Spitfire 1a(100 octane) vs AI bf109 E-1. Given ample fuel, multiple waypoints and starting position close to AI's airfield the AI will perform three or four aggressive passes. If I am able to evade the attacks the AI then goes into a passive "drone" mode wherein said weird rolling behaviors are used to get the player to exhaust his ammunition supply. Many times I have then lost contact with the AI who then will use the opportunity to exit the fight by landing at his home airfield. I eventually started loitering near the AI airfield and would attack the AI during its' landing approach. The AI then  attempts to raise its' landing gear to take off again, often stalling and crashing in that process. This propensity to sneak away from a fight, coupled with the rolling behavior, entirely destroys immersion. The issue has been described numerous times in the forum yet I cannot recall ever seeing a TFS acknowledgement that a "target drone" AI behavior actually exists. 

[BENKOE]

On 2015-Apr-04 Team Fusion published following information [ref.1]:

 

Quote

CLIFFS OF DOVER has a large bug... in fact it is the biggest bug we have encountered, it affects power generated by engines, how effectively they are cooled, and how fast aircraft are at higher alts.

The bug is that in game, the calculation for density of the atmosphere decrease with altitude is off by a factor... density decreases with altitude faster than real life.

 

Those of us having the 1c:Maddox "vanilla game" remember that it has a standard atmosphere temperature model  with (273.15+17)° Kelvin (SI) for mean sea level. Where 290.15 are 17° celsius. The temperature fits nearly the average in August arround Dover [ref.2], at around 17.5 °C | 290,65 °K (SI). It also fitted nearly the average in August arround Calais[ref.3], at around  17.7 °C | 290,85 °K (SI).

 

The air density is a function of air pressure and air temperature, so its 'correctness' is dependent on that of pressure and temperature. Thus, if desired, density can also be considered as being a parameter with which any constant pressure occurs in the atmosphere.

 

When flying at 4000m in the CLIFFS OF DOVER 'vanilla game', it's important to know that when we use any equation that includes values of gravity, 'tape measure' altitude has to be converted into pressure altitude. So, flying at 4000m = 3997m pressure altitude and ~264°K ambient temperature, the warbird 'feels' like it is flying at 4062m density altitude, equal to 4065m geometric altitude. The temperature decreases with a uniform lapse rate of approx. 6.5°C/km up to 4000m {(T264-T290) : 3997}.

Of particular importance is the role that the increased temperature plays in reducing the air density, and hence the performance of your plane. Team Fusion comments as follows [ref. 1]:

 

Quote

THIS AFFECTS ALL AIRCRAFT NOT JUST YOUR FAVOURITE RIDE.

This is the reason in the vanilla game the British aircraft would overheat at any altitudes over about 18,000 ft. The 109's were not affected because the original developers couldn't find the bug, threw up their hands and decreed the 109's would NEVER overheat in any circumstance.

The bug has defied our ability to locate it. (its in the core.dlls somewhere)

For TF 4.312 we corrected the issue in regards to power generated by increasing by a factor the density of the atmosphere once it arrives in the manifold. The speed factor, ie. the fact the atmosphere is less dense, we have not been able to entirely fix, hence aircraft are a little faster than historical at higher altitudes. We have also corrected the overheat issue somewhat by increasing the cooling factor for a given radiator by making the radiator calculate its cooling effect as though the altitude was lower. However, we have not been able to match the mathematical curve of the density bug decrease with the mathematical curve of the altitude calculation mod exactly, so at certain altitudes, in particular the altitude where the supercharger is working the hardest, (rated altitude or in your example at around 4000 meters) the heat generated tends to overpower the cooling effects. We could change the mod curve, but then the aircraft would tend to NEVER overheat at lower altitudes, something which is already a factor.
So while we would obviously like a better solution to the bug problems of the original game, so far we have not found one... but the solutions we are currently employing are not in any way detremental to your favourite aircraft, but in fact apply to all the types in the game.

 

So, Team Fusion changed the 1C:Maddox atmosphere against an atmosphere model that comes close to what we know as the International Standard Atmosphere [4]. But, it is not identical. Team Fusion Blitz_v5.040 atmosphere refers to a virtual combination of altitude and temperature that obviously has a detrimental effect on AI and/or what is, perhaps in the core.dlls somewhere.

 

Reference:

[1] IL*2 Sturmovik Cliffs Of Dover 4.x - Bug #744

[2] Climate Data -Dover-

[3] Climate Data Calais

[4] 1976 International Standard Atmosphere (PDF)

 

Edited February 28, 2024 by BENKOE

[Volant_Eagle]

Bugs with the atmosphere may exist and be causing various performance inaccuracies, but I highly doubt this would have anything to do with the particular problem in question. Which is the difference between max roll rate achievable by a player vs that achievable by an AI in the same exact plane.

 

Changing the atmosphere could certainly affect roll rate, but all aircraft in simulation would still be flying through the same atmosphere as each other. So no matter how small or large the change to the atmosphere, it should only increase or decrease the roll rate for a given aircraft by exactly same amount for both players and AI (Assuming they do in fact both follow the same rules of physics).

 

Also, if atmosphere was the actual culprit causing these high roll rates; I would expect it to be causing other performance changes with exactly the same magnitude. Instead, most performance and handling characteristics seem either right on or only slightly off, and it’s only AI roll rate that is off by ridiculous proportions.

 

Because this AI roll bug stands out so uniquely and noticeably, I don’t think it’s likely that its cause has anything to do with the game’s overall flight physics or atmosphere models. My hunch is that either something has been overly simplified for AI and thus physics applied to them are skewed, or there is some sort of AI behavior protocol that has been given higher priority than physics and is overriding it.

 

 

[DD_Arthur]

It’s simply a bug deep within a flawed engine.

That TF ‘discovered’ what it considered the biggest bug in the game nine years ago is because TF has always prioritised the multiplayer experience over single player.

The fact that the a.i. seems to be flown by the Incredible Hulk is irrelevant to having a full server so they ignored it.

[Dagwoodyt]

Blitz has partial Tacview support via a mod available to registered ATAG users. I have not downloaded the mod as it is not available via this forum. Anyone who uses Tacview and is an ATAG member could presumably install the mod in effort to document player best case roll rates vs observed AI roll rates for given aircraft.

 

 

 

 

 

[Rei-sen]

Might as well just compare maximum AI roll rate in TacView with whatever test data devs use as a reference for the particular aircraft. I'm pretty sure the results will be hilarious. 

[Dagwoodyt]

I use Tacview pretty routinely in Great Battles and DCS and find it quite useful.

 

[Buzzsaw]

On 2/28/2024 at 11:43 AM, DD_Arthur said:

It’s simply a bug deep within a flawed engine.

That TF ‘discovered’ what it considered the biggest bug in the game nine years ago is because TF has always prioritised the multiplayer experience over single player.

The fact that the a.i. seems to be flown by the Incredible Hulk is irrelevant to having a full server so they ignored it.

 

The bug with high altitude performance was solved 7 years ago... anyone who takes the time to fly the game can see that.

 

We re-wrote the entire engine code to solve the issue.  (8 months of work)

 

Temperatures are now correct at altitude.  (based on the temperature being set correctly at sea level)

 

Aircraft now climb to their historical altitudes without overheating.

 

In fact, the CLIFFS Flight Model is, and remains one of the best out there... better than our competitors in our opinion.

 

As mentioned, we are looking at the roll rates for AI... the errors you might experience are nothing to do with the Flight Model and are something else which needs re-writing, and will be fixed.  Unless you are a programmer you wouldn't understand.

[DD_Arthur]

6 hours ago, Buzzsaw said:

 

The bug with high altitude performance was solved 7 years ago... anyone who takes the time to fly the game can see that.

 

We re-wrote the entire engine code to solve the issue.  (8 months of work)

 

Temperatures are now correct at altitude.  (based on the temperature being set correctly at sea level)

 

Aircraft now climb to their historical altitudes without overheating.

 

In fact, the CLIFFS Flight Model is, and remains one of the best out there... better than our competitors in our opinion.

 

As mentioned, we are looking at the roll rates for AI... the errors you might experience are nothing to do with the Flight Model and are something else which needs re-writing, and will be fixed.  Unless you are a programmer you wouldn't understand.

 

? 

Yep, 8 months of work to solve an issue that affected primarily online players yet the ATAG server always degenerated into a low-level bust up between 'French point' and 'English point.'

In the meantime that effort could of been put into the still unresolved A.I. behaviour that would have benefitted the vast majority of CLoD's users. Unfortunately, the vast majority of CLoD's users in that time have simply walked away....

Since we all know that TF 'leadership' are NOT programmers this last sentence seems to imply there's a long way to go yet. The Cliffs flight model remains one of the oldest out there being Oleg's beloved, twenty-five year old, wooden lookup tables whilst DCS, GBS, X-Plane and MSFS forge ahead with blade element theory and soft body physics..... 

 

[Buzzsaw]

4 hours ago, DD_Arthur said:

 

? 

Yep, 8 months of work to solve an issue that affected primarily online players yet the ATAG server always degenerated into a low-level bust up between 'French point' and 'English point.'

In the meantime that effort could of been put into the still unresolved A.I. behaviour that would have benefitted the vast majority of CLoD's users. Unfortunately, the vast majority of CLoD's users in that time have simply walked away....

Since we all know that TF 'leadership' are NOT programmers this last sentence seems to imply there's a long way to go yet. The Cliffs flight model remains one of the oldest out there being Oleg's beloved, twenty-five year old, wooden lookup tables whilst DCS, GBS, X-Plane and MSFS forge ahead with blade element theory and soft body physics..... 

 

Actually the work on revising the engine code affects all players... both multiplayer and single.

 

With the original game, it was impossible to fly at the altitudes where most of the Luftwaffe bombing missions occurred.

 

So BoB campaigns which are historically based, with both sides being required to climb to altitude, and fight at altitude were not possible.

 

In addition, with the changes, missions/campaigns involving B-17's and the other heavy American bombers would be even less possible, since these missions were flown and fought at up to 30,000 ft/9500 meters.

 

The changes we made were necessary for both accurate flight modeling and historical fidelity... and benefit anyone who plays the game.

 

And we have not neglected the AI behaviour.  We have introduced numerous changes in both BLITZ and TOBRUK which improved AI behaviour... and we will be introducing additional improvements, including a hoped for solution to the issue of AI rollrate for TF 6.0.

[BENKOE]

On 3/9/2024 at 6:30 AM, Buzzsaw said:

 

The bug with high altitude performance was solved 7 years ago... anyone who takes the time to fly the game can see that.

 

Thanks @Buzzsaw, please explain where the atmosphere bug is?

 

 

Edited March 10, 2024 by BENKOE

[jdu]

On 2/29/2024 at 4:27 AM, Dagwoodyt said:

Blitz has partial Tacview support via a mod available to registered ATAG users. I have not downloaded the mod as it is not available via this forum. Anyone who uses Tacview and is an ATAG member could presumably install the mod in effort to document player best case roll rates vs observed AI roll rates for given aircraft.

 

 

 

 

 

V2.0 is on the way.
 

https://www.youtube.com/watch?v=agcQxz1CHm0

 

[Feldgrun]

6 hours ago, Buzzsaw said:

... and we will be introducing additional improvements, including a hoped for solution to the issue of AI rollrate for TF 6.0.

 

The AI roll rate has been an issue for many years. From Stormbirds review of IL2 Desert Wings-Tobruk in August 2020:

 

"It’s also frustratingly annoying at times where the AI can out roll or out maneuver a human player flying the same aircraft because their flight model just isn’t the same. I have had some great dogfights against this AI system and then also had a few that I was genuinely annoyed at when the target ahead of me flick rolled away like no human could ever do."

 

The AI roll rate is the reason I avoid playing against AI fighters in this game. It's also bad when I play against AI Blenheims, which is a bit comical.

[Dagwoodyt]

6 hours ago, Buzzsaw said:

We have introduced numerous changes in both BLITZ and TOBRUK which improved AI behaviour... and we will be introducing additional improvements, including a hoped for solution to the issue of AI rollrate for TF 6.0.

Hopefully, eliminating AI "target drone" behavior is lumped among  unspecified "additional improvements" or perhaps buried too deep in the code to even acknowledge its' existence.

[Bussard_x]

The AI P40 and AI Dewoitine have the too high roll rates, both fully new aircraft in the Tobruk release. I have not seen this with other aircraft. If AI fine tuning is done for these aircraft it should be okay.

[Dawson]

4 hours ago, Bussard_x said:

The AI P40 and AI Dewoitine have the too high roll rates, both fully new aircraft in the Tobruk release. I have not seen this with other aircraft. If AI fine tuning is done for these aircraft it should be okay.

Fiat G50 is another offender.

[Feldgrun]

The roll rate generally needs to be dealt with, not just a few fighters. The AI Bf 110s & Blenheims can flick over like one-seated fighters. 

Edited March 11, 2024 by Feldgrun

[Gunfreak]

18 hours ago, Feldgrun said:

The roll rate generally needs to be dealt with, not just a few fighters. The AI Bf 110s & Blenheims can flick over like one-seated fighters. 

I find the 110 even harder to deal with than the 109. It's like trying to grab a oiled up weasel. It seems at least as manoeuvrable as the 109. Not the relatively easy prey for the Hurricanes and Spitfires the sources talk about.

[LLv34_Flanker]

S!

 

110 is usually regarded as an easy target, but again when flown to it's strengths it was a formidable opponent. Not when tied to close escort of bombers without speed. And basically this has given it rather a poor reputation.

 

It could leave Hurricane in the dust and even Spitfire has to work hard if the 110 fights smart. ZG76 had better K/D in 110 than many 109 squadrons against fighters. Of course later war changed this due higher performance fighters appearing. But again at night it was still the scourge of RAF Bomber Command.

 

TL;DR Keep E and altitude and a 110 can be a nasty opponent.

 

 

[No.54_Reddog]

23 minutes ago, LLv34_Flanker said:

S!

 

110 is usually regarded as an easy target, but again when flown to it's strengths it was a formidable opponent. Not when tied to close escort of bombers without speed. And basically this has given it rather a poor reputation.

 

It could leave Hurricane in the dust and even Spitfire has to work hard if the 110 fights smart. ZG76 had better K/D in 110 than many 109 squadrons against fighters. Of course later war changed this due higher performance fighters appearing. But again at night it was still the scourge of RAF Bomber Command.

 

TL;DR Keep E and altitude and a 110 can be a nasty opponent.

 

 

All true. Except that's not how the AI flys it and that's what this topic is about. 

 

The AI has been "being redone" since at least 2018 if not 2016 IIRC. Its now 2024. Like everything else with this game it's simply another example of missed opportunities and poor decision making.

[Dagwoodyt]

3 hours ago, LLv34_Flanker said:

S!

 

110 is usually regarded as an easy target, but again when flown to it's strengths it was a formidable opponent. Not when tied to close escort of bombers without speed. And basically this has given it rather a poor reputation.

 

It could leave Hurricane in the dust and even Spitfire has to work hard if the 110 fights smart. ZG76 had better K/D in 110 than many 109 squadrons against fighters. Of course later war changed this due higher performance fighters appearing. But again at night it was still the scourge of RAF Bomber Command.

 

TL;DR Keep E and altitude and a 110 can be a nasty opponent.

 

 

In Great Battles 5.202 VR I had so much trouble in SP 1v1's flying P47's against "Ace" 109 E7's that I tried flying against "Ace" AI Bf 110's instead. I've managed victories against the 110's, but only with great effort. Fights of more than 20 minutes are not uncommon. A damaged AI is still dangerous if I lose sight of it as the AI is in every fight for the long haul. So what's with CLoD AI? After 3-4 aggressive passes the AI is generally off to attend a previous engagement somewhere and treats my nuisance by rolling and rolling and doing split esses down to just above the wave tops to rid itself of the burden. I recall many DWT flights in Martlets chasing 109's running home to some forlorn little desert airstrip. I am left with an impression that this is an immutable facet of the SP game.

[LLv34_Flanker]

S!

 

Problem with AI in many games, regardless of genre, is that it is usually very predictable and either a super being or exceptionally dumb. Or any combination of that matter. CloD has it's issues as well and really hope those will be solved somehow.

[Deltahawkoz]

Well it's not just me then..as these posts attest. Been playing Desert Hawks alas the insane AI make it unplayable in my view.  Tie fighters Star Wars.  Perhaps tone down AI in some way with ME?

[Buzzsaw]

On 5/1/2024 at 2:52 AM, Deltahawkoz said:

Well it's not just me then..as these posts attest. Been playing Desert Hawks alas the insane AI make it unplayable in my view.  Tie fighters Star Wars.  Perhaps tone down AI in some way with ME?

We are working on this now.

[BENKOE]

What is this statement referring to?

Temperature at sea level as monthly average temperature?
Temperature at sea level as yearly average temperature?
Approximately temperature?
Undisturbed outside or static air temperature?
Increased total air temperature?
Temperature of dry air?
Temperature of moist air?

Temperature, which may be used for calculations following the perfect gas law?

 

On 3/9/2024 at 6:30 AM, Buzzsaw said:

 

... Temperatures are now correct at altitude.  (based on the temperature being set correctly at sea level) ...