Dev Blog #34: Massive Bomber Raids

[LukeFF]

Hey everyone! In today's blog entry, we discuss how large bomber raids will be possible in Korea. IL2 Series. You can read all about it here: https://il2-korea.com/news/dd_34

 

[Juri_JS]

What's the performance impact of all the AI gunners in such a large formation?  Will it be necessary to simplify their behaviour too?

 

Moreover I wonder if features will be added in the editor to make building large formations easier, for example by arranging several flights into larger formations. At the moment the difficulties start once such a formation changes it course, because the individual flights won't automatically hold their position within the formation. 

[[CPT]Crunch]

Shouldn't need surgical gunners, sheer numbers is what made them dangerous.  Loosen up the programming and sniper tracking, it really isn't needed and doesn't sell well.  Just the fact a hundred guns are firing in a boxed area within a given distance I'd be fine with a formula giving odds of a hit.

 

Whatever so long as it isn't a slide show or destroys every living thing in the air.

[354thFG_Leifr]

Looks great, some of the first pieces of media that I am genuinely interested in and excited by - I hope the work pays off! Wouldn't it be a dream if some of this technology trickled down a little in to BoX? I'd kill to see some simple AI bomber formations, even if it's just the measly B25 and B26 aircraft we already have. 

 

Great work.

[=MERCS=JenkemJunkie]

I think the gunners should be snipers for people who are flying predictably or parking behind them, but they should reward proper approaches and jinking by not sniping people using proper tactics.

[Aapje]

2 hours ago, Juri_JS said:

Moreover I wonder if features will be added in the editor to make building large formations easier, for example by arranging several flights into larger formations. At the moment the difficulties start once such a formation changes it course, because the individual flights won't automatically hold their position within the formation. 

 

It makes most sense to me that this would be implemented as a single entity in the editor with the number of planes as a parameter, and then the entire formation having a single course. So then each plane would have a fixed spot in that formation.

 

To make it look realistic, it would then need some 'jitter' to cause variance between each plane, including in turning maneuvers, so the planes turn with slightly different timings.

 

@LukeFF

 

Do you know how the game deals with the inability to take off/land? Will these planes always fly in from the edge of the map, or will they just spawn in the air?

[LukeFF]

That would be more of a question for @Han to answer. 🙂

[[CPT]Crunch]

3 hours ago, =MERCS=JenkemJunkie said:

I think the gunners should be snipers for people who are flying predictably or parking behind them, but they should reward proper approaches and jinking by not sniping people using proper tactics.

Except where is the proper approach to a well defined box of sixty bombers, there is none from any angle, not if it's a proper defensive box formation.  Good luck with that.

[Han]

1 hour ago, LukeFF said:

That would be more of a question for @Han to answer. 🙂

Somehow that it wouldn't hurt the immersion. )

[Avimimus]

It'd be interesting to hear how the simplified aircraft compare in detail to the Il-2 1946 aircraft? How detailed are the flight models and graphics (recognising that Il-2 1946 was once the gold standard for realism).

[Avimimus]

  

4 hours ago, =MERCS=JenkemJunkie said:

I think the gunners should be snipers for people who are flying predictably or parking behind them, but they should reward proper approaches and jinking by not sniping people using proper tactics.

 

There are some foundational issues with the gunner calculations. IMHO, it would be very good to build in ranging errors (similar to what is used in Tank Crew), in addition to the other types of errors. Gunners should be 'accurate' at where they do aim but 'misjudge their aim'... but gradually improve in accuracy as they make corrections.

 

In my assessment, the current system takes a perfect shot, and combines all sources of inaccuracy into single equation which is then added to it. This has an issue in that there are some combinations of range and velocity where these calculations cancel themselves out and the gunner makes a perfect shot. If you know how to avoid these you are invulnerable, if you don't - then the gunners feel like all powerful snipers.

 

It was really difficult to balance when I made the Rise of Flight gunner mods - as the conventional system either made it so the gunners always missed or so they always hit... there was no middle ground. I ended up reducing the default angular inaccuracy and increasing the contributions of velocity to inaccuracy and it seemed to do much better...  (basically changing the equation through changing the weights).

 

If I recall correctly, I tended to reduce the "coefficients for the dependence of the final shooting area on the AI level" (e.g. 'CoefLow') to make the gunners more accurate by default and increased the "coefficient of conversion of the difference in the speed of the object on which the turret is located and the target, in the aiming plane, into the error angles" (i.e. 'ProjSpeedCoef'). It seemed to work a bit better.

 

Anyway, it would be good if the system was reworked from first principles at some point (and had error in ranging separated from error due to other factors, like difficulty operating the turret).

[=MERCS=JenkemJunkie]

1 hour ago, [CPT]Crunch said:

Except where is the proper approach to a well defined box of sixty bombers, there is none from any angle, not if it's a proper defensive box formation.  Good luck with that.

In the B-29s case, that things a beast and you drew the short straw if you're attacking a sixty bomber formation of them, but diving on them in a frontal attack while jinking would be less terrible than cruising behind them straight and level blasting away. If it's a RNG chance to get hit, and it doesn't track my movements, then why not game the system and just blast away while cruising level to have an easier time aiming? The B-29 didn't have a normal gunner system, it had a computerized system that tracked planes and calculated the shot, so if any plane should have gunners that track planes, it should be the B-29. 

3 minutes ago, Avimimus said:

  

 

There are some foundational issues with the gunner calculations. IMHO, it would be very good to build in ranging errors (similar to what is used in Tank Crew), in addition to the other types of errors. Gunners should be 'accurate' at where they do aim but 'misjudge their aim'... but gradually improve in accuracy as they make corrections.

 

In my assessment, the current system takes a perfect shot, and combines all sources of inaccuracy into single equation which is then added to it. This has an issue in that there are some combinations of range and velocity where these calculations cancel themselves out and the gunner makes a perfect shot. If you know how to avoid these you are invulnerable, if you don't - then the gunners feel like all powerful snipers.

 

It was really difficult to balance when I made the Rise of Flight gunner mods - as the conventional system either made it so the gunners always missed or so they always hit... there was no middle ground. I ended up reducing the default angular inaccuracy and increasing the contributions of velocity to inaccuracy and it seemed to do much better...  (basically changing the equation through changing the weights).

 

If I recall correctly, I tended to reduce the "coefficients for the dependence of the final shooting area on the AI level" (e.g. 'CoefLow') to make the gunners more accurate by default and increased the "coefficient of conversion of the difference in the speed of the object on which the turret is located and the target, in the aiming plane, into the error angles" (i.e. 'ProjSpeedCoef'). It seemed to work a bit better.

 

Anyway, it would be good if the system was reworked from first principles at some point (and had error in ranging separated from error due to other factors, like difficulty operating the turret).

True with the normal gunners we're all used to, but the B-29 is a special case in that it has a computer system that tracks the planes and provides targeting info, so the B-29's chance of making shots on non-maneuvering planes should be pretty high. If there's different AI skill levels though, that would be good for the lower B-29 AI gunners though.

[BlitzPig_EL]

Didn't the B-29 gunners have trouble tracking fast moving jets, as the turret speed was too slow, as it was developed for use against WW2 prop aircraft?

[=MERCS=JenkemJunkie]

If they're using proper tactics sure, but if they're making themselves easy targets there shouldn't just be some X% chance they get hit. 

[LukeFF]

43 minutes ago, BlitzPig_EL said:

Didn't the B-29 gunners have trouble tracking fast moving jets, as the turret speed was too slow, as it was developed for use against WW2 prop aircraft?

 

That is what I have read, yes

[Stonehouse]

It's probably too early for the dev team to comment definitively but it would be really interesting to hear about other AI improvements that are hoped for in Korea. Things like flak avoidance maneuvers and hopefully moving away from the current hardcoded ranges for different skills of fighter pilots to open fire to something that simulates higher levels of experience "judging" when better to fire and better at taking deflection and bullet drop into account. Also give fighters different behaviours based on target type would be a help. Realistic formations and pairs fighting and ground attack profiles is another example.

Edited June 7 by Stonehouse

[CzechTexan]

I see Flames over Tokyo in the future.

[Avimimus]

8 hours ago, LukeFF said:

That is what I have read, yes

 

That is interesting! I recall the B-29 having a very fast turret speed (faster than human operated turrets).

 

The firing time would also have been very short (the Mig-15 would be diving from above, rather than struggling to climb) and the weight of fire from the Mig is tremendous. The formations used in Korea generally weren't that large, but the evidence suggests that the Mig-15 had little trouble breaking through them. It'll be interesting (and quite dramatic) to try.

 

I do wonder if the AI in the new sim will allow for realistic tactics (much like late WWII bomber tactics, the Migs seemed to have preferred a 'one pass and out' approach - so they really shouldn't be sticking around to make a second pass at the bombers in most cases).