Oleg, very curious on the AI in Battle of Britain.
 

"Every AI pilot will have his own profile including favorite tactics and combat methods." (announcement SOW BOB)

Oleg, will the profile of every AI be something like this?:

-Experience
-Tactics
-Methods

And what about deflection shooting from AI pilots?
This is something that AI does not do in IL2 series. If you turn away
AI will stop firing. Really hope AI does this in BOB.

Will AI pilots avoid AA guns?
This is also something that AI does not do in IL2 series.
Currently flies over enemy airfield as if it's all is OKAY down there.

Will AI have same limits as human flown aircraft?
In IL2 series AI outruns human flown aircraft.
This is def. a show stopper for off-liner as I am.

Will AI detect enemy already when they are still little spots?
Like it is now in IL2 series, hope not.

Will AI pick their battles?
Will it head for home when plane has slight damage or in case of
overwhelming enemy situation?

Good questions all, and every one more important than most eye candy (save realistic clouds).

One suggestion. Make the AI code have SOME variables that mission builders can get at. If the skills are some sort of scripting, let us have a version of the script as text we can mess with. We don't need to know all the nuts and bolts if it is propietary, just some variables we can control. The variation in AI you wil see as a result will be very interesting.

Things like the max spotting distance for them, perhaps a range in degrees of a blind spot we can chose. For the skills, if you program 20 different 'AI skills" allow the mission builder in the mission file set which each pilot gets, etc, etc.

I too am curious about how the "AI" will be implemented in "BoB SoW".

You have a few interesting questions but I'm not sure I understand your statement about the "damaged fighters running for home". Is this a good thing or a bad thing? I have seen damaged "AI" fighters in "IL2 1946" running for home trailing smoke...do you like this?

I have posted some questions about the new "human damage model"(its in the Maddox's Room thread). I'm curious about how complex the new damage systems will be in "BoB SoW"... One thing I'm hoping for is that the same "DM" that will be use for the "AI" pilots in "BoB" will be used to limit/effect the "players" input.

I know that Oleg has said that the "line of sight" targeting issues have been addressed. ...now they (AI) will have the proper "blind spots" (like clouds and under/behind the aeroplane). I "think" he said that AA gunners will be governed by the same rules...

I understand your dissatisfaction with the "AI" pilots apparent lack of concern at being bombarded by the AA over enemy airfields but this is not without some historical foundation. ...bombers would have to fly strait and level (through the worst of flak and AA) when they where on their bomb run...:)

Take the BOB II WOV, and update with the patches. Then you'll experience some outstanding AI performance. Player doesn't have any real ability to adjust the AI responses, but the BDG team is working on it.

The AI in the BOB II WOV is without doubt the best of any CFS at this time.

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

Oleg has his work cut out for him with the AI performance, if BOB II WOV is a benchmark for performance.

Because the Online game is such a big part of IL2 it may be the same in BOB SOW. There might not be as much need for exceptional AI performance. AI performance is critical in Offline play.

I'm hoping Oleg will produce an outstanding AI performance. It will be a huge improvement in the sim.

Oleg once said he will use simplified AI for distant aircraft -- what I fear may be like the AI Bubbles used in some sims like Rowan's BoB and MiG Alley. Its possible the individual AI pilot personalities will be confined to the Player Squadron only.

Good questions indeed.

I remember that, too, but I rather think of it like it was realized in the Falcon-series. There was a war with literally thousands of units in the air and on the ground + several fleets for both sides. While they were outside the players visibility, they were merely calculated by chance and a simplified strategy system. When they were close enough for the player to become important, they were simulated as real units, with armor values, radar and effective gun ranges, etc.

Compared to WW2 planes, for example, there is little interest in simulating fear and fatigue as it was noted in that statement from Oleg in every AI gunner of a B17. Instead I imagine a formation will have sort of a swarm-intelligence, as they will stick together and react to communications and radar as a single unit. When they get closer, it's probably different values for each pilot being simulated, e.g. awareness of each pilot, while other things like CEM would be pretty much simplified. When entering combat, they'd probably simulate each pilot with the stated values and complex A.I.

I was glad to hear about the blind spots for BOB...
As the AI have been the worst part of IL2...
The ridiculous (and heavily used) negative G maneuvers...
and the head hunting AI on ACE level..
seeing through clouds and weather...
as well as the AI bomber gunners...

These are glaring problems of the game as a whole... and I know some of them will be addressed in BOB..
Id also like to see that their friend/foe identification is limited to a very close range. As it is now.. they seem to know you are enemy from the start...
rather than having to close and visually ID like we would without labels.
Making it so they have to get within say 2km or so before determining if you are friend or enemy would be nice. Or whatever is a an acceptable distance..
I dont know how the draw distance is yet in BOB :)

Id also like to see a much broader range of skills. Right now how you set the AI... you get to much of one thing and not enough of the other.. and setting it on ace.. well they are all aces...
I'd like to see even the ace level settings.. only allowing a "few" aces..

For squadrons with varying skill levels of pilots.. that makes flying as a group more engaging...

you never know what your going to come up against... but you dont have to worry about them all being aces. You dont want your less experienced guys getting discouraged cause all they ever do is take off and get shot in the head by all the ace AI.


But for the many of us that enjoy COOP type flying vs the air quake rooms... it will be nice to see the AI on a more even playing field and using the same flight models

nearmiss said something that it wouldn't be that important in online play... but I seriously beg to differ...
online coop has great appeal to many. And that requires some very good AI.

Disregarding AI.. I rate IL2 as an 11 out of 10. The best WWII sim ever.
But when you factor in the AI... that drops it to a 7
Its the only thing I hear people complain constantly about.

did you see that?!?!? God I hate those #$#* negative G manuvers!! Hey look at this.. this Ki-43 is out running my mustang!?! and etc.

So im really hoping they do a vast improvement on the AI.

A setting for level of aggressiveness would be essential too. Aces that can fight defensively and rookies that go charging in etc. Hyper-aggressive types would even target you in your parachute...

Feuerfalke (regarding Falcon sim):: Interesting. I would hope its a bit more than beyond the player's visibility -- 10km or 20km is not very far, especially for jets. Oleg did say more recently that external camera view may be restricted to aircraft close to the player, or some sort of restriction that implied an AI Bubble. Its possible that the distant aircraft will be simulated abstractly, which might be workable for a very large strategic level air warfare simulation. But to be honest, I really think that AI BOMBERS can be simulated directly, with very simple flight models, since their primary function is to fly in formation as their primary offensive and defensive combat tactic, unlike fighters using energy and manuever for offensive and defensive combat tactics.

For example, I'm trying to build a map for the StrikeFighters where the USSR can be surrounded by SAC airbases, with the map extending from Pakistan to Japan and north to Norway and Alaska. Any sim that tried to cover that might benefit from having aircraft and combat simulation beyond say, 1000km from the Player Plane made more abstract. Although environmental effects like aurora should be visible to several thousand kilometers, and some combat effects can be seen from thousands of kilometers, depending on yield...

In F4 the bubble was 120km IIRC.

120km. Not bad. Thanks.

Depends on aircraft performance. For something with the performance of B-70s and MiG-25s, maybe 1000km, considering the tracking of ground control which I think Oleg is making for BoB And Beyond should have, although for much lower performing aircraft.

Must say that I just installed 4.08m, AI is better in this version.
More close to human flown aircraft.

We see a lot of pictures of the development, but nothing said about AI development in the game. Has there been any news the last 2 years? Really curious about this development.

Diving out and running for home if damaged or in a tight position like being outnumbered was commonplace for both sides. As senior officers were fond of saying: aircraft are easy to replace but trained aircrew are not! So a fuel leak or a glycol leak, or damage to control surfaces, called for as rapid an exit as possible. This is written in most of the many books that I've read about the Battle of Britain. Fire or dense engine smoke was the signal for flipping the plane over and getting out, before the explosion or the steep dive to oblivion.

It's only in gaming situations that people continue to fly about shooting when the plane is crippled. In real life they hit the silk as soon as possible, or aimed their plane for the nearest dry land.

B

Yes, correct I hope that AI wil do this in Battle of Britain. Running for home when aircraft is damaged or when clearly outnumbered. That will make sim more realistic.
Fighting continuing till all aircraft from one side are all annihilated is definitely not realistic.

Nothing worse than an AI plane smoking and leaking coolant getting on your 6 and hammering you 5-6 mins after you had given him a taste of your lead, flying like theres no problem for him:(

Some of the Ki's can be on fire and then turn to attack you belching thick smoke..............

Ah then the flak............. thats another story :)

What is even more interesting, is that the reason it was used so much is that it was effective. The other side simply didn't give enough chase a lot of times!

Why? Simply because these guys fought in reality and acted with survival in mind, it's pure math. An enemy fighter that decides to split-S out of the fight is one less fighter your flight has to deal with. That's it, you stop dealing with him the moment he's not interested in putting guns on you and focus on the rest of his flight.

Of course this is dependant on mission type. It's more probable to let escaping enemies go during an escort mission, or any other type of multi-flight operation that requires the pilots to maintain certain positioning. On the other hand, it would be the least possible to see this in a freijagd/fighter sweep mission, then i think it's safe to assume that the attacking pilot would give chase to a damaged and escaping enemy much more readily.

This may be one of the reasons we don't see AI that breaks for home so often. A virtual PC pilot will 99% of the time dive after that damaged enemy fighter no matter what. Older sims had AI that used to do this and it resulted in an unrealistic number of kills scored by the human player during the course of an offline career. Sure, the AI is a bit too reluctant to save their own skins, but on the other hand we're guilty of exploiting AI that does make the effort to disengage ;)

Double edged sword there. I know When doing something like an SEOW Pilots are much more economical with their virtual lives because you typically have limited, planes and lives, and getting the mission done is much more important than getting a kill (well, for most guy's anyway) because the mission is usually more important to the actual campaign than the kill count.

If a similar way of enticing the live pilot to act more realistically could be found then the AI would be less of a problem.

I think that A.I in most games tends to favour the lot's of stupid targets instead of fewer clever ones.
I can't think of many games where the A.I just bugs out. Devs never seem to account for instinct over intelligence. They need to write some A.I with a sense of self preservation.

AI bombers?
 

Will AI bombers go for targets of opportunity? I never liked how the (for instance) AI Kamikaze planes would stooge about if the primary target was hit and sank. Other "targets of opportunity" got off easy instead of AI going after them. Will that change in SoW?
Flyby out

About the detecting of enemies: It would be nice if your leader would say (example) : "Attention! Planes at 10 o'clock. Lets go look who they are."

Yes, would be better and you have the element of surprise in this case.

One way to discourage this type of (unhistorical) behaviour could be to punish the player point-wise. Imagine negative points for leaving your squadron against orders, no points for shooting down planes that can't be confirmed (mates more than X kilometers off, enemy falling over sea or over enemy territory). Hunting that lonely plane that is buggering off for home won't be that attractive any more.

Personally I'm hoping this IS the case - I would rather have CPU cycles devoted to good AI for aircraft I'm fighting against than a lesser AI spread across the entire map.
Also it would presumably then be easily possible to have a map only high speed mode (as in the old Pacific Air War), so campaigns can be fought by those who have a life outside of sims.

This seems unlikely, surely it makes more sense to give the opposition better AI than the players own squadron.