PDA

View Full Version : test your acceleration versus AI



lowfighter
04-21-2007, 11:47 PM
Some time ago I started looking closer at AI performance, combat settings etc. I've come up with a little test to see how AI accelerate during level flight. I'm posting it now because there's periodical discussions about AI performance, also because I think it might be usefull to help improve our flight performance. So the tests mission has 3 airplanes A, B, and C. A and B are the same type, C might be different type. At the start of the mission A (player aircraft) is about 400m behind B, and B is about 4000m behind C. A and B are red and C is blue ( or A and B are blue and C is red). C has NO ammonition so it will just fly straight and act as a "bait" for B. Because C is pretty far ahead it will take time till the B AI will catch the C AI. All planes are at the same altitude, A and B have the same speed too. So when the mission starts you're just behind B and the aim is to either stay with B or even try to go ahead him and reach first C. What would be the outcome?
It's also interesting to run the mission on autopilot and write down the combat settings of the AI (rpm, MP etc).
Of course the test can be run at different altitudes, or having A and B different to test relative acceleration performance etc.
If you know the Full Mission Builder you can set such a test mission yourself. If you don't here's an example you can try with P47's. Copy the text below and paste into worldpad. Save it as (for example) 'speedtest.mis' in any single player folder for example in
C:\Program Files\ubisoft\IL-2 Sturmovik 1946\Missions\Single\US\P-47
Then you can play it as a single player mission.
Here's the text to be pasted:

[MAIN]
MAP Crimea/load.ini
TIME 12.0
CloudType 0
CloudHeight 1000.0
player usa0103
army 1
playerNum 0
[Wing]
usa0103
usa0102
g0103
[usa0103]
Planes 1
Skill 1
Class air.P_47D27
Fuel 100
weapons default
[usa0103_Way]
NORMFLY 30364.53 69813.04 100.00 300.00 &0
NORMFLY 100056.34 69819.59 100.00 300.00 &0
[usa0102]
Planes 1
Skill 1
Class air.P_47D27
Fuel 100
weapons default
[usa0102_Way]
NORMFLY 30963.78 69803.47 100.00 300.00 &0
NORMFLY 101107.86 69879.12 100.00 300.00 &0
[g0103]
Planes 1
Skill 1
Class air.P_47D27
Fuel 100
weapons none
[g0103_Way]
NORMFLY 34087.01 69723.39 100.00 650.00 &0
NORMFLY 102458.18 69803.05 100.00 650.00 &0
[NStationary]
[Buildings]
[Bridge]
[House]

FritzGryphon
04-22-2007, 12:27 AM
A nice, controlled test.

I raced the red plane at 110% power.

Initially, I was catching up, and closed to within 160m. However, at that point, the red plane pulled away, and continued to pull away at an increasing rate for the remainder of the test.

Unfortunately, the race doesn't last long enough for my engine to break down. But I suspect the red plane would keep using WEP indefinitely.

I tried the test with autopilot, but since the red plane climbs and my plane flies straight, no comparison is possible.

F19_Ob
04-22-2007, 01:09 AM
I was thinking about why they would boost the AI.
One thing that came to mind was the accelleration might be needed for the AI to be able to sustain turnfights and BnZ reasonably realisticly?
There are ofcourse the mentioned penalties but are there other ways to make it work?

They ofcourse are aware of the AI's problem, or impossibility to think ahead in battle, like a human would, so they must compensate for it.
Likely it's very difficult to find a balance that works for all possible situations.

I don't know, just searching possible solutions to 'why'.

badatit
04-22-2007, 10:58 AM
I'm rusty, but I gave it a try.
I copied the txt to a new file, and changed all a/c type to P51-20NA.
I stayed low, pure pursuit on a/c C. This put me well out ahead of a/c B.
Settings were full switch.
After crossing over landmass, I lost vis on a/c C. I continued initial flight hdg of 090.
I regained vis on a/c C a short time later at my 1:00, he had made a wide banking turn to the southeast.
I came around to 135, then approx 180, when a/c B cut the corner, and engaged C. B wasted two long burst, with no hits. By this time, I was in gun range. After a/c C had been destroyed, I rejoined with a/c B. Again...I am very rusty.

TRK file is here.
http://www.sendspace.com/file/dmxu0o

badatit
04-22-2007, 01:08 PM
Here is another one.
http://www.sendspace.com/file/o9aih0

The setup is, extreme NW corner of Crimea.
One (two ship) element of 77thFG P51-20NA's
One (two ship) element of 79thFG P51-20NA's

79thFG is Ace AI, 100 fuel, default loadout.
They have one waypoint, down in the SE corner of the map. Their speed was set for 750KPH (it defaulted to 703KPH). 703kph is 436mph...well above the max for a p51 at sealevel. So, they are at full speed to waypoint one.

The 77thFG element setup is identical, with one exception...lead is player.

At missions start, speed falls off to around approx. 350mph.

I catch, and form up on 79thFG lead.
If you look at the track, you'll see I have no trouble keeping in close formation with the AI, and could pull away, and out distance them, if I choose to do so.
I was off the ball some, heavy control imput at times, etc. Still could keep formation easily.

I dont get, why people complain about the AI.
If you cant out perform the AI (same a/c), your not getting full performance out of your a/c.
The AI fly close to the edge, but a human can get closer still.

badatit
04-22-2007, 04:15 PM
Same as above, but added waypoints to see how hard it'd be to keep up with their 50m to 10000m combat climb.
They are good. But, as before, they are not at the extreme edge.

TRK file
http://www.sendspace.com/file/z4w9jl

badatit
04-22-2007, 06:41 PM
I stand corrected.
After a bit more testing, I was warmed up.

I put a blue p51 out in front of my leader.
Once he saw it, I had a "real" tough time staying up with him. Aces and Rookies alike...They are faster, when in combat.

AFJ_rsm
04-22-2007, 08:31 PM
yes


this is precisely why oleg should give us the cem and pitch tricks of the 109 back

lowfighter
04-22-2007, 09:22 PM
One little thing about the start of the test. If you hit autopilot right at the beginning you see that the AI increases power gradually, taking some 4-5 seconds I think till he reaches full power. Therefore if I start the test I'll try to do the same. If I just quickly push the trottle at maximum I'll have a slight acceleration advantage at the beginning.

lowfighter
04-22-2007, 11:03 PM
Originally posted by F19_Ob:
I was thinking about why they would boost the AI.
One thing that came to mind was the accelleration might be needed for the AI to be able to sustain turnfights and BnZ reasonably realisticly?
There are ofcourse the mentioned penalties but are there other ways to make it work?

They ofcourse are aware of the AI's problem, or impossibility to think ahead in battle, like a human would, so they must compensate for it.
Likely it's very difficult to find a balance that works for all possible situations.

I don't know, just searching possible solutions to 'why'.

I'd guess the rather simple "intelligence" model is done to minimise the real-time computations. I think it's not that hard to improve certain AI flaws but the question is at what cost. Then it's also a question of the game engine, or in other words the programing principles which govern the game. The engine would allow or not different things happening, and sometimes to allow more things happening it would need to heavily modify the engine itself. Soetimes I write some programs with a certain task and then have the idea to include more features and then I realise that the best thing is to rewrite from beginning using different principles... That's not so hard with small programms but imagine trying to do that with a monster program like il2.

I did this acceleration test with different aircraft and there's big confusion in my head, I'll try to sort things out and post my thoughts...
Just one thing, radiators, I don't know what settings the AI use. I used in all tests radiator on position 2. Didn't fiddle with prop pitch.
Sometimes managed to reach first the C airplane sometimes not. Whenever I tried to stay with B I was left slightly behind but that after considerable time. When I ignored B and just try to reach as fast as possible the C I beat the B ...almost all the time.