For thoes that are interested LETS TALK AI PREFORMANCE

[bobhegf]

I am starting this post hoping to get a better understanding of how the AI work and to share some of the things I have found out about the AI and I hope others will do the same. I will start, I am sure that most people that fly CFS2 have found out that the AI takeoff in one direction and land in an other direction, you don`t land or takeoff down wind. I took care of that problem many years ago and posted it more than once but had very few try it. The problem I am having at the present time is getting the AI up over the trees in my Guadalcanal missions and stop after landing where I want them to. They seem to want to come in over a wide area not coming in over the trees but through them. I am working on correcting that problem or trying to.

I hope that all of you will help build this post so that it will give us all a better understanding about mission building and how the AI work in CFS2. This may get others interested building missions or trying correct some of the problems we have.

 

[Blood_Hawk23]

Pushing the Envelope...

A few things that I've messed with in my WWI AI work.

Too many aircraft designers ignore this portion of the DP file. This is from the stock Zero. Which is a mess.

[MISC_DATA]
unit_family=1
category=1
allegiance=4
max_group_size=8
min_speed=240
cruise_speed=415
max_speed=622
min_alt=50
cruise_alt=6667
max_alt=10000
entered_service=8/1/40
crew=1

Please note: All DP specs are Metric. This causes alot of confusion for many people. the enteries MUST remain in metric. Do Not set them to MPH or Feet.

Lets study the Flight envelope for a minute. As this is what these 6 specs are setting for the AI.

min_speed=240
This should be the stall speed. The zeros real stall speed was 110.
This is the point where the plane will either start to climb or stall. The AI seems to use it both ways. They will use it in combat as well. Now this may or may not be effected by weight. I have not noticed any Ill effect by lowering it. They seem to adjust rather well to load.

cruise_speed=415
For the zero an Ideal speed would be 333.
Some just Half the min and max. This I use as the best fuel consumpion speed even though the AI don't run out of fuel. It does effect combat a little. You want real world specs but this you could play with. You might notice different behavior in combat.

max_speed=622
Should be 533, The do not exceed max speed was 660. We want strait line speed.

min_alt=50
This has given me the most pause for thought. 50 meters min alt. Why? 164 feet min alt. I'm not sure if this is the hard deck that MS wanted to use or if it effected landing on the ships. That is something I never tested. I put my WWI ACs at 25. I noticed the AI would use more of their flight envelope. That along with the other three effect what they will do in combat.

cruise_alt=6667
I use the best max speed alt or the combat alt. Many of my WWI AC I would put this at 2000m. The Zero's max speed Alt was 4550. This set the mid point of the combat envelope.

max_alt=10000
This should be the service Ceiling. which Is correct here.

The AIs use this info to fly the plane. Its not just to tell us what the plane could do.

So If you want to either improve or limit your AI I would start there. Other areas can effect how well the AI uses this info. An accurate flight model aerodynamicly is the most important part of the AIs handling. These set the envelope of flight.

Oh, as an extra note. These specs also effect Mission Builder. For now I won't get into that part.

This just what I have observed in my WWI AI flight tuning.

Till Later,
John

 

[achim27619]

AIs starting and landing

Bob,

to the best of my knowledge the start and landing behavior of an AI is not at least determined by the entries in the airbases.dat.

[runway.121]
id=8226
name=Italy - Ferrara
base_lat=N44 46.93
base_lon=E11 36.23
base_alt=22
heading=102
allegiance=0
units_across=2
takeoff_start_pos=10,0
takeoff_stop_pos=1600,200
landing_start_pos=5000,350
landing_touchdown_pos=600,0
landing_stop_pos=50,0

base_lat and lon is the position at the end or better the beginning of the runway. Other airfield builder like MaskRider usually put it into the center.
takeoff_start_pos: in this case 10 meters down the runway (taking the position as a reference)
takeoff-stop_pos: I never checked that one but I assume that is 1600 meters down from the start-pos and 200 feet above ground level
landing_start_pos: That is 5000 meters out and 350 feet above ground level
When an AI is to land the last waypoint is always be 5 miles out, 1500 feet agl and it is lined up for the approach.
landing_touchdown: 5000 meters out and 350 feet above ground level the AI starts the landing
landing_touchdown_pos: the AI touches down 600 meters down the runway
landing_stop_pos the AI should stop 50 meters short of the lat_lon pos.

I have to admit that I am not 100 % sure about meters and feet but this is what I learned from testing.

So if you want the AIs to clear an obstacle you have to shorten the distance and raise the elevation in the takeoff_stop_pos.

I hope that helps.


Cheers

Achim

 

[Blood_Hawk23]

Bob,

to the best of my knowledge the start and landing behavior of an AI is not at least determined by the entries in the airbases.dat.

[runway.121]
id=8226
name=Italy - Ferrara
base_lat=N44 46.93
base_lon=E11 36.23
base_alt=22
heading=102
allegiance=0
units_across=2
takeoff_start_pos=10,0
takeoff_stop_pos=1600,200
landing_start_pos=5000,350
landing_touchdown_pos=600,0
landing_stop_pos=50,0

base_lat and lon is the position at the end or better the beginning of the runway. Other airfield builder like MaskRider usually put it into the center.
takeoff_start_pos: in this case 10 meters down the runway (taking the position as a reference)
takeoff-stop_pos: I never checked that one but I assume that is 1600 meters down from the start-pos and 200 feet above ground level
landing_start_pos: That is 5000 meters out and 350 feet above ground level
When an AI is to land the last waypoint is always be 5 miles out, 1500 feet agl and it is lined up for the approach.
landing_touchdown: 5000 meters out and 350 feet above ground level the AI starts the landing
landing_touchdown_pos: the AI touches down 600 meters down the runway
landing_stop_pos the AI should stop 50 meters short of the lat_lon pos.

I have to admit that I am not 100 % sure about meters and feet but this is what I learned from testing.

So if you want the AIs to clear an obstacle you have to shorten the distance and raise the elevation in the takeoff_stop_pos.

I hope that helps.


Cheers

Achim

Hey Achim,

I think that this is one of the rare occasions that feet is used.

So the second number is the height. Ohhh... you know I have always wondered about that.

 

[achim27619]

Hi John,

as Ialready said I am not 100 % sure that the hight is in feet and not metric. Some more testing is needed to verify it.

Cheers

Achim

 

[Blood_Hawk23]

Hi John,

as Ialready said I am not 100 % sure that the hight is in feet and not metric. Some more testing is needed to verify it.

Cheers

Achim

no argument here. I does make sense. I didn't realise it was height. It would be interesting to test. I'd take a jeep and set a pattern. then sit back and watch.

On a side note. how can we view a target. we have a bomb view. is there a target view?

 

[Rami]

Reply...

Hey guys,

I can add to this thread by saying that the 20th Fighter Group campaign has tested my ability to understand AI and get them to conform to the mission goals.

For instance, I realized that there is no way to have goals change for the AI during the course of a mission. Starting in late April, 20th Fighter Group pilots started strafing targets of opportunity on the deck on the way home following uneventful escort missions, which made me scratch my head on how to pull off. What I ended up doing is a "lone wolf" mission, where you fly as a single-ship element, in close formation with the remainder of your flight for the escort run. You perform escort duty until the given point, where you break off and the bombers return home.

As you head toward the strafing area, the first waypoint after you depart the bombers is where you rendezvous with the "other" fighters who will join you for the strafing run, with the added benefit of having bombs and rockets, because they didn't have to perform the escort duty. You then use your ammunition to strafe targets of opportunity before returning home.

 

[bobhegf]

The Cactus missions I am working on are all lone wolf missions. One reason was they almost always had an odd number of aircraft on hand and the second reason was the tactics used by the F4Fs to attack the bombers. The F4Fs would dive from 2000ft above and make a high head on pass aiming for the wing root or a high side pass aiming for the wing root of the enemy bombers passing through the enemy formation and coming out on the other side. Because of this it was hard for a wingman to stay with his lead. The rule was once the attack or attacks were completed return to base and while on the way home if you saw another F4F or groups of F4Fs join up and come home with them.

 

[Blood_Hawk23]

AI Landings... Again...

This is a topic we see a lot of. To this day Some planes will some won't.

I just did a simple test mission where I followed a plane from takeoff to landing. Well it didn't land. It went right of the run way at about 350ft above the ground. Slowed to almost its stall speed but kept right on going. it never sped up or tried to go around again.

Now I remember there were a number of factors. One of them was in the .air file. Something about the "1203-autopilot limits". I could have the wrong entry.

The aircraft I was using was AWai P-51d that Seagull V released with the RAAF Korea pack. Great packs BTW.
The mission has 6 WPs. takeoff to WP2 Alt at 2000' and 300kts Aspd. WP2 goes 5miles out. Turn left to WP3 Alt and Aspd the same for WP3. At 2 miles away from WP2 turn left. Fly to WP4 no changes, Fly 10miles. Turn lft at wp4 slow to 200knt and drop to 1000'. Fly to WP5. at 2miles turn left for landing. landing is 5miles out.

Should there be another WP to bring it down more and at a lower speed? Say WP6 alt 500' Aspd 150knts.

What are your thoughts?

Till Later,
John

 

[kelticheart]

It must be in metres

.....landing_start_pos: That is 5000 meters out and 350 feet above ground level
When an AI is to land the last waypoint is always be 5 miles out, 1500 feet agl and it is lined up for the approach.
landing_touchdown: 5000 meters out and 350 feet above ground level the AI starts the landing.....

Hi Achim,

I believe all of the above figures are in metres. 350' above ground level would make a scant 106.68 metres of altitude to hold for 5km (3.107 miles) before final. It's definitely not advisable flying "hanging from the prop" which that would mean, as any flight instructor can tell us.
Any engine failure, downdraft or too close proximity with stall speed and there's not enough height to recover from, a guaranteed crash.

Built-in simulator flight lessons teach as well it's always better having a high approach to the runway than a low one. Each one of us must have surely visited an airport, travelled by plane or seen an airshow, how many times we did witness such a low altitude approach?

Hence, that 350 above ground level number must be in metres, the conversion rate being # of metres/0.3048 = feet, that makes 1,148.29' of altitude. A more comfortable height to begin a landing approach.
Oh, boy! My manual about the AI subject is getting bigger with this thread!

Cheers!
KH

 

[achim27619]

altitude in meters

Hi Stefano,

here you have a good point. You are probably right so the altitude metric also.
If I find the time I will try to check that one.

Cheers

Achim

 

[Blood_Hawk23]

Yea it would have to be all one unit of measure. they wouldn't use both feet and meters in the same space. but tell me if this looks correct. Red numbers are the foot conversions

[runway.121]
id=8226
name=Italy - Ferrara
base_lat=N44 46.93
base_lon=E11 36.23
base_alt=22
heading=102
allegiance=0
units_across=2
takeoff_start_pos=10,0 (32,0)
takeoff_stop_pos=1600,200 (5249,656) thats about a mile
landing_start_pos=5000,350 (16404,1148) that is 3 miles out
landing_touchdown_pos=600,0 (1968,0) thats near 1/4 of a mile
landing_stop_pos=50,0 (164,0)

I'm not a pilot myself so I don't know the proper approach. If that looks correct, to any of you that do fly, then it must be.

I alway thought it was feet in reference to the runways center. I didn't think of it as the approach to the runway.

How long is this runway. is the any way to find out?

 

[achim27619]

Hi John,

this particular runway is 3350 feet long.

Cheers

Achim

 

[bobhegf]

In CFS2 the runway is measured in 10ths of an nm which is around 6000ft. You can use the MB to get the distance.

 

[Blood_Hawk23]

thank you guys.

I was playing with MKAFD, this for the most part is the standard.

takeoff_start_pos=100,0
takeoff_stop_pos=1600,200
landing_start_pos=5000,350
landing_touchdown_pos=600,0
landing_stop_pos=-250,0

All of this has me thinking. This is what I use for most of my WWI aerodromes.

takeoff_start_pos=-100,0
takeoff_stop_pos=1000,200
landing_start_pos=2000,350
landing_touchdown_pos=550,0
landing_stop_pos=-250,0

my runways are between 2000 to 3000 feet long. most are 2000'. the basic British airfield was 2000'x900'.
so if these are indeed metric, then I maybe setting my aerodromes incorrectly. I have to think about all of this.

 

[kelticheart]

thank you guys.

I was playing with MKAFD, this for the most part is the standard.

takeoff_start_pos=100,0
takeoff_stop_pos=1600,200
landing_start_pos=5000,350
landing_touchdown_pos=600,0
landing_stop_pos=-250,0

All of this has me thinking. This is what I use for most of my WWI aerodromes.

takeoff_start_pos=-100,0
takeoff_stop_pos=1000,200
landing_start_pos=2000,350
landing_touchdown_pos=550,0
landing_stop_pos=-250,0

my runways are between 2000 to 3000 feet long. most are 2000'. the basic British airfield was 2000'x900'.
so if these are indeed metric, then I maybe setting my aerodromes incorrectly. I have to think about all of this.

BH, here are the conversion tables:

millimetres / 25.4 = inches
centimetres / 2.54 = inches
metres / 0.3048 = feet
kilometres (1 km=1,000 metres) / 1.6093 = statute miles

Reverse the table, this time multiplying instead of dividing, in order to get metric measures converted to inches, feet and miles.

Cheers!
KH

 

[Roxane-21]

BH, here are the conversion tables:

millimetres / 25.4 = inches
centimetres / 2.54 = inches
metres / 0.3048 = feet
kilometres (1 km=1,000 metres) / 1.6093 = statute miles

Reverse the table, this time multiplying instead of dividing, in order to get metric measures converted to inches, feet and miles.

Cheers!
KH

You forgot :

kilometres / 1.852 = nautical miles

 

[Blood_Hawk23]

I get all of that. Actually I cheat. Either I use google or I have a spreadsheet that I use which also does speed conversions.

Kelti,
Does that look right for an approach pattern?

 

[codeseven]

Hi Bob, still after those AI answers I see You and I go way back on this subject.

I have always found the behavior of the AI to be the most fascinating and misunderstood aspect of CFS2. I still check in here every once in a while just to see what new has been figured out concerning the AI and have always thought that the person that finally fully figured AI behavior could change this sim in a big, big way!

Specifically, over the years I have been mostly interested in AI landing behavior. For instance, I could get the AI to start, takeoff and land on a carrier, even moving carriers, but never could get them to stop on deck after landing before rolling off into the sea. I know it is possible because I saw it once with my own eyes. I was messing around one day and had just the AI aircraft taking off and landing on a static carrier. I would mess around with different aircraft file parameters and try again and again to get them to stop on deck. I had spent several hours watching plane after plane land, slow down (never saw an AI catch a wire either, but I think the tail hook stuff if hardcoded out of them) and come to almost a stop but then suddenly turn (usually to the right, or starboard) and just roll off the deck near the bow. Then, one time, out of the blue a stock Wildcat used as AI took off, landed, slowed down and stopped at the end of the flight deck! what?!! I tried to get it to do it again but it never would Just a fluke I guess.

Interesting info in this thread, I hope it all gets figured out one day.


Codeseven

 

[Blood_Hawk23]

I wonder if you increased the brake strenght. made it so it would stop landing before 0.

I'm still tring to get them to land on the ground. nevermind on a carrier. I think it have to try and dig up some old info.