Help and Support Forums => Aces High Bug Reports => Topic started by: CptTrips on January 24, 2019, 12:45:00 AM
Title: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 24, 2019, 12:45:00 AM
I might be wrong on this, but something doesn't seem right.
To trouble shoot I flew a mission manually while using .psave to drop files of my pos while recording film so I can go back and see what my speeds were at the time I save pos. I was then creating a Staged Mission basically mimicking what I had actually flown.
I started a take-off in a Camel and lifted off ~64 mph and climbed for ~20 sec at ~67-70mph from ~71-74 ft ASL to ~445 ft ASL. That was a climb of ~374ft in 20 sec. That would be a climb rate of ~1122 fpm. Not exact measurements but in the ball-park.
The Camel according to Wiki has a max climb rate of 1085 fpm so within the slop of my measurements, I'd call that good, or at least not surprising.
However, when I plot the same flight profile in the Mission Editor, at the alt I flew it at the speeds I recorded, the segment is telling me my best climb is 472 fpm. WTF? So then it complains that the segment is unflyable. It can't be. I flew it! What is it basing that number off of? The flight model seems right, I think the Mission Editor is FUBAR! :bhead
(See attachment.)
[Note: Double checking...changing the fuel load-out didn't change the reported best climb. So that isn't it.]
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 24, 2019, 08:44:08 AM
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 24, 2019, 09:26:19 PM
Another data point....
So I decide to just ignore the warnings and create the mission with the data I actually flew so knew that was a good climb rate. It appears that the AI is limited to that bogus climb rate. Instead of climbing at the defined speed of 70mph for instance, they are wobbling along at 50mph. Like they are constantly on edge of stalling trying to climb beyond their max climb rate. Meanwhile, I can climb right past them at the proper climb rate.
So I believe the Editor is truthfully reporting the climb rate of the AI, but that climb rate isn't the proper climb rate that I am capable of (which is very close to the documented rate).
That makes it kinda sucky to have to follow them in take-off an climb out as they have less than half my climb rate.
As another test, I create a simple air spawn and level flight at 100mph and they had no problem maintaining that speed in level flight as defined in the mission file. So it is a climb issue.
:bhead
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 30, 2019, 03:06:25 PM
What is the climb rate of a camel supposed to be? And if you ignore the best climb rate. It will do exactly what you are speaking of.
HiTech
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 30, 2019, 04:30:00 PM
What is the climb rate of a camel supposed to be? And if you ignore the best climb rate. It will do exactly what you are speaking of.
HiTech
I'm not sure I understood your comment.
Let me phrase it more carefully.
1. Under as close of conditions as possible, I manually flew the camel at ~70mph for 20 sec and climbed ~374ft = ~1122fpm.
2. I tried as closely as possible to plot the same climb in the editor but it stated only a climb rate of 472fpm is possible. If I ignored that warning and plot what I flew, and then build the mission and follow the AI as they try and fly it, it did not appear as if they could attain that climb rate and they were wallowing at ~50-55mph and not 70mph and did not seem to attain the target height in that time. I got the impression that the AI was truly limited to the climb rate the editor had reported, even though seem to have a much higher climb rate flying manually.
Are you disputing #2? Are you saying the AI should have climbed as I plotted it regardless of the warning if that is what I could fly manually? It didn't appear to be the case as I flew along side the AI as they tried. I can redo a test. I'll try to create a side by side climb with the AI. The AI should be able to stay right along with me. I'll try to get direct evidence if it can't, like a film.
I'm not debating what the max climb rate should be, I'm just saying it seems it should be the same for the AI as a manual pilot under the same conditions. What ever that number is.
(Also, the warning said in some future version, violating those climb numbers would break the build. So if you start enforcing them, they should definitely be realistic. But we can argue about that later. :))
:salute
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 30, 2019, 05:01:16 PM
Wait. I guess before I bother testing more I need to make sure I'm making the proper basic assumption.
Do the AI and player use the same flight model?
Should I expect them to basically exhibit the same performance?
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 30, 2019, 05:21:15 PM
Different model. The number shown is Max.
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 30, 2019, 05:45:57 PM
OK. That is my problem. I was assuming the AI could fly what I flew with the same plane.
Basically I was creating a utility to let me just fly the route the way I wanted and record the data, then create a route and segments out of the recorded data programmatically. Then I was going to have the AI fly the first route while I fly another manually recording the next set of data interacting with the first AI route. I could then continue that like again like laying down multiple audio tracks. I thought that might be easier to create complex interacting behaviors than futzing with the editor.
Got all that working, but the AI can't fly what I just flew with the same plane. It is using a significantly different flight model with only 42% of the climb rate. Doesn't that seem like a pretty big discrepancy? Could we tighten that up? Could anyone record various flight profiles to help you update the data?
Hmmmm.
Could the Editor could have an optional switch to ignore it's normal constraints and just fly the freakin profile as defined? Like an advanced option. It was true recorded flight data. If anyone ever submitted a mission it should be an easy scan to see if that override had been checked if you have concerns.
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 30, 2019, 05:51:01 PM
On the bright-side, I now have a good strategy against the AI. Don't try and out turn them, out CLIMB them! You have twice the climb rate! :t
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 30, 2019, 10:55:19 PM
The other bad thing here is when you are the flight leader and you set yourself on auto-climb auto-pilot, you just leave your AI wingmen in the dust because they can't even come close to keeping up with you with the exact same plane and load-out. :cry
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: Bizman on January 31, 2019, 02:29:41 AM
The other bad thing here is when you are the flight leader and you set yourself on auto-climb auto-pilot, you just leave your AI wingmen in the dust because they can't even come close to keeping up with you with the exact same plane and load-out. :cry
That's the problem with real players as well. If the leader has full throttle, there's no chance for the wingmen to stay in formation, no matter how skillful they are. There's always minor delays between command, receiving and action. Plus the inaccuracy of gauges which leads to minor adjustments which eat speed and increase distance between planes. The leader should always reduce speed to allow the wingmen to catch up without having to use WEP.
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 09:13:28 AM
That's the problem with real players as well. If the leader has full throttle, there's no chance for the wingmen to stay in formation, no matter how skillful they are. There's always minor delays between command, receiving and action. Plus the inaccuracy of gauges which leads to minor adjustments which eat speed and increase distance between planes. The leader should always reduce speed to allow the wingmen to catch up without having to use WEP.
I'm not talking about a momentary gap. It is a continual increasing gap. It is continually increasing because they only have 42% of your climb rate. Every second you climb, they fall further behind and below you.
If they are the lead and you take-off with them and climb out with them to a combat altitude, you are going to have to take over twice as long to get to that desired altitude as you should have to.
Are you suggesting the FM just be left that way?
This isn't minor. It's not even in the ball park. They have less than half the available climb rate as the player's plane. As numerically reported in the Editor and verified by Hitech. (at least for the Camel) This isn't my imagination, or an incorrect interpretation of what happened.
Rather than just excuse it, I'd prefer to just fix it. I'm even offering to assist collecting the flight data if that is useful.
:salute
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: Bizman on January 31, 2019, 10:27:23 AM
Although we're comparing AI to real players, the issue is the same. If the leader continuously flies at full throttle, the followers can easily fall beyond icon range. That's not momentary nor is it minor, but with live players it's understandable. The followers can't read the leader's mind in order to simultaneously do the same turns and other actions no to mention other elements that can distract them.
I understand your concern, though. If the AI wingmen only have 42% of the leader's climb rate, that would equal a late night highly boozed scenario mission with the best pilot leading the group. That doesn't sound like making the AI act like real players, only us Finns. For what I've learned, there's many players who really practice to be good in this game, flying sober in daylight.
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 11:01:01 AM
Although we're comparing AI to real players, the issue is the same. If the leader continuously flies at full throttle, the followers can easily fall beyond icon range. That's not momentary nor is it minor, but with live players it's understandable. The followers can't read the leader's mind in order to simultaneously do the same turns and other actions no to mention other elements that can distract them.
I understand your concern, though. If the AI wingmen only have 42% of the leader's climb rate, that would equal a late night highly boozed scenario mission with the best pilot leading the group. That doesn't sound like making the AI act like real players, only us Finns. For what I've learned, there's many players who really practice to be good in this game, flying sober in daylight.
Oh I get what you are saying. Especially in turns between the inside corner planes and the outside corner planes. That is not what we are dealing with here.
I get there is a different flight model between the AI and player. Given you would want to keep the AI as lightweight as possible as far as processing, it seems perfectly reasonable that you wouldn't try and use a full fidelity force based flight model for each AI plane. I would suspect there is probably some type of interpolation lookup table of values. It seems there is either bad data or a bug in the interpolation.
Or if it is an unavoidable side effect of a consciously design trade-off, then maybe an override flag to just fly the points as designed. At least for non-attack segments. When the AI breaks off the route to go fight something, let him use his normal FM, but on route fly what i put in.
But ideally, we could just adjust the built in model to get closer to reasonable.
Oh, and if I lived in Finland I would drink heavily in Winter too. ;o)
:salute
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 31, 2019, 11:04:28 AM
The large difference may also just be with WWI planes. I never tested the WWI ai.
HiTech
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 11:25:34 AM
The large difference may also just be with WWI planes. I never tested the WWI ai.
HiTech
Entirely likely. That is why I added the caveat (At Least for the Camel).
I can tell you from my experience there is a big difference in how insane the AI's maneuvering ability between the WWII aircraft I have tested Staged Missions with and the WWI aircraft.
The WWII AI is pretty scary, but the WWI AI seems off the chart. See my previous challenge here: https://bbs.hitechcreations.com/smf/index.php/topic,394820.0.html (https://bbs.hitechcreations.com/smf/index.php/topic,394820.0.html)
I'd pay you a U.S. dollar to fight that guy and post me a film. :D Short of a HO, or a maneuver kill, there is no physical way to out fly him. Maybe an extremely luck snap shot. But you definitely get the feeling the AI has a silly turn rate. :)
Luckily he can't shoot straight, so you can delay the inevitable, but he will continue to close on you with a slower plane until he rams you from behind. :O
And yet, they can't climb out on route at a reasonable rate. :rofl I've said this AI is like retarded Terminators. :D
It there any data I can help collect for you to assist your analysis? Would it help to manually fly various climbs at various speeds at various alts with the WWI planes and collect that data? I'm willing to help any way I can.
:salute
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 31, 2019, 12:21:51 PM
Ok did some testing and looking.
The AI climb is very close to the real flight model. The easy mode flight model uses the AI flight model so was simple for me to test.
The editor show bests climb rate is at 75% power. If you switch to military it shows climb rate at 95% power.
Also the calc for it is an quick approximation, and will show lower then what the AI can really do.
And finally , if your segment is even 1 foot per min greater then what the model can do, the AI plane will go into a stall condition and never catch up. I.E. don't ever exceed what the max climb rate shows with military power.
Also at some point I'll take a look at changing the AI slightly so if on a climb segment, they never fly slower then Best Climb Speed.
HiTech
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: FLS on January 31, 2019, 12:38:13 PM
The WW1 AI don't seem g limited.
The best strategy if the AI is in a rotary engine a/c is to get them to spiral climb to the right.
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 12:46:53 PM
The AI climb is very close to the real flight model. The easy mode flight model uses the AI flight model so was simple for me to test.
The editor show bests climb rate is at 75% power. If you switch to military it shows climb rate at 95% power.
Also the calc for it is an quick approximation, and will show lower then what the AI can really do.
And finally , if your segment is even 1 foot per min greater then what the model can do, the AI plane will go into a stall condition and never catch up. I.E. don't ever exceed what the max climb rate shows with military power.
Also at some point I'll take a look at changing the AI slightly so if on a climb segment, they never fly slower then Best Climb Speed.
HiTech
OK, my bad.
So I was flying 70mph climb at full throttle. At military throttle the Editor is reporting 783 fpm. Still a lot lower than the ~1100 fpm I flew.
I guess that last 5% throttle is enough to make that difference? PLus it is a conservative estimate.
Could we have options for standard, military, and full throttle climb?
People are impatient, and no one around here pays for engine maintenance. :)
At a 100% throttle setting, it should have pretty much matched what I flew?
:salute
[edit] hmmmmmm You just gave me an idea. When I fly and record my routes, I could set my settings to easy mode so I can just fly like the AI? So I'll just fly like the AI instead of trying to get the AI to fly like me. That might help me avoid problems like this.
Does that mean I get to make 60g turns too? :t
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 01:04:45 PM
That is great flying, but even he seemed tame to what I was seeing.
Mind sanity checking me with this one? Waffle Terrain. https://bbs.hitechcreations.com/smf/index.php/topic,394820.0.html (https://bbs.hitechcreations.com/smf/index.php/topic,394820.0.html)
That video is old enough to make me wonder if things have been exacerbated since then.
(If you did try that one, and you get a shot at him, could you confirm if the AI seems very fragile to hits? That seems to have changed lately, but maybe I am misinterpreting that too.)
Hmmm maybe I ought to try that one again too. That was last Oct and easy mode might have been tweaked. Everything is a moving target. ;)
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: FLS on January 31, 2019, 01:06:59 PM
The Wiki climb data on the Camel comes from this document which I posted here some years ago. I don't know if we have the same engine modeled.
Well tried that 1v1 again and it was pretty much what I remembered. I did hold him off in a stalemate for a long time. I just kept trying to out climb him. :)
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 31, 2019, 03:13:27 PM
Well tried that 1v1 again and it was pretty much what I remembered. I did hold him off in a stalemate for a long time. I just kept trying to out climb him. :)
With WOP we have added AI ability settings. I just have not put them in the Mission Editor yet. Lets you set things like max g range. And other numbers which modify the performance of the AI.
HiTech
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: CptTrips on January 31, 2019, 03:21:15 PM
With WOP we have added AI ability settings. I just have not put them in the Mission Editor yet. Lets you set things like max g range. And other numbers which modify the performance of the AI.
HiTech
That will be freakin awesome!
:salute
Think this stuff could ever lead to a stand alone single player game? Mo money, mo money....
Title: Re: Staged Mission Editor: Best Climb Seems FUBAR
Post by: hitech on January 31, 2019, 03:38:25 PM
I have considered selling missions, game free, and mission creators get a cut. But the mission arena never seemed to attacked any interest.
HiTech
The problem with mission arena is that you are pitching your current players who are conditioned to only go to the MA and expect to be playing other players. There are many players out there who don't want to play other people, they just want something they can play by themselves when they have a spare hour. There are a lot of people who would be less intimidated just fighting AI. That would target a completely different player base. The non-multi-player crowd.
Rather than a Mission Arena, I was thinking more of a separate install, packaged offline stand-alone single-player game. Released on Steam or purchased off your website. A couple of terrains with a sets of missions taking the player through a series of career progression missions. "Aces High: Air Combat Anthology" :)
Terrain 1: Battle of Britain 20 missions career progression; RAF pilot starting with Hurricane's and the being transferred to Spits. Big finally, massive dogfights over London!
Terrain 2: Stuka Pilot 20 missions career progression; Barbarossa to Kursk. Attacking tanks, airfields, shipping, refugee columns. :)
Terrain 3: Cadillac of the Sky 20 missions P-51 career progression ETO; Bomber escorts. Fighter sweeps. Airfield strafing.
Etc.
Sorry. Apparently I've had way too much coffee this afternoon. :lol