FAR Fighter Challenge: BD Armory AI

[SpaceplaneAddict]

Awesome suggestions, thanks. By the way, I'm too OCD to make flat wings:D. That's also where i hide LOTS of ammunition. The thing has nose elevators on, for pitch and AoA, which is set to 200%. As for Dogfighting speed, Anything above 250 m/s works fine. Here's the new file: https://drive.google.com/file/d/0B588M8xKv0QKcGE4NUVSS3poNXc/view?usp=sharing

MAJOR EDIT:

LOL, I gave the wrong craft file. If you already downloaded, could you set the AV-R8's to 200% AoA? Thanks.

Edit 2: I also forgot to turn off pitch and yaw from the airbrake(s?) Sorry I messed the download up.

Edited August 28, 2015 by SpaceplaneAddict
Mistake(s)

[Wanderfound]

Awesome suggestions, thanks. By the way, I'm too OCD to make flat wings:D. That's also where i hide LOTS of ammunition. The thing has nose elevators on, for pitch and AoA, which is set to 200%. As for Dogfighting speed, Anything above 250 m/s works fine. Here's the new file: https://drive.google.com/file/d/0B588M8xKv0QKcGE4NUVSS3poNXc/view?usp=sharing

MAJOR EDIT:

LOL, I gave the wrong craft file. If you already downloaded, could you set the AV-R8's to 200% AoA? Thanks.

Edit 2: I also forgot to turn off pitch and yaw from the airbrake(s?) Sorry I messed the download up.

Sorted.

BTW: AoA settings work in opposite ways depending whether the control surface is in front or behind the CoM. Behind CoM (e.g. on the tail), positive AoA settings promote stability and reduce stalling, while negative AoA settings increase control authority. Ahead of CoM (e.g. canards), negative AoA settings promote stability/reduce stalling, while positive AoA settings increase control authority. If you run a FAR stability analysis in the SPH, you should be able to see how the control surfaces deflect in response to AoA.

Did you intend for me to set the canards with +200% AoA? Thanks to its otherwise low pitch authority, the K-16 flies successfully with +200%, but it does mean that the canards are pretty much permanently stalled until you reach supersonic speeds. If it was me, I'd be more inclined to set them with -100% AoA and crank up the maximum deflection on both canards and elevators.

I'm not sure that the autopilot is smart enough to make use of the airbrake, BTW.

[SpaceplaneAddict]

Lol, I'm so sloppy, I need to revise the design.

"MOM!", "Get me my cola, I'll be Kerballing for a few *days*"

Thanks for the explanaition tho.

Note to self: I suck at Ferram Aerospace...

Edit: I am a noob, simple

Edited August 29, 2015 by SpaceplaneAddict

[Wanderfound]

Some MiK-3 vs K-16 testing:

* The high weight of the plane and the drag from the stalled canards means that the K-16 takes a long time to get up to speed, and is limited to about 4G until it reaches transonic speeds and the canards cease stalling.

* Once it goes supersonic it can comfortably pull 6-8G at altitude.

* However, if it gets lured into a dive, it has a tendency to break up when pulling high Gs below 1,000m.

The MiK-3 vs K-16 fight began with the MiK reaching altitude first and launching a pair of missiles. However, the K-16 managed to get off one of its own in return. The K-16 successfully dodged both missiles, and gradually built up to supersonic speed. Meanwhile, the MiK turned to avoid the K-16's missile. The K-16 turned to follow and launched another AMRAAM, which the MiK narrowly dodged by diving towards the ocean, pulling out 500m off the water with a 10G turn. The K-16 attempted to follow this maneuver...but when it tried to pull out of the dive, its tailplane broke off, followed shortly afterwards by the disintegration of the entire airframe and impact with the ocean.

[SpaceplaneAddict]

What would you think about the thrust vectoring engine in the BahamutoD part pack? Could that possibly be allowed until 1.1 brings the new "Panther" jet?

Edit: Never mind, forgot to look at the rules

Edited August 29, 2015 by SpaceplaneAddict

[Wanderfound]

What would you think about the thrust vectoring engine in the BahamutoD part pack? Could that possibly be allowed until 1.1 brings the new "Panther" jet?

The point about limiting parts is mostly to minimise the number of mods on my not-very-good laptop. But there's nothing to stop you from building with whatever parts you want and flying it vs a MiK or whatever on your own machine.

[SpaceplaneAddict]

No problem, just checking. I got a status update on the K-16, it just destroyed your MiK-3 while flying like a drunk Irishman. (No offense to Irish people). Ran a few simulations, all of them had my K-16 stallin', and flyin' slow, yet it's bloodthirsty pilot fired missile after missle, eventually downing the MiK a few consecutive times.

Enough of that, now for the nitty-gritty details. I found that using swept wings for elevators work substantially better. Please keep it down below 400 m/s, and above 240 m/s.

Have a craft file: https://drive.google.com/file/d/0B588M8xKv0QKcmFzZ2lFZ1hrMlk/view?usp=sharing

P.S: Is your MiK-4 the new champ? Can't tell

[Wanderfound]

Nice; getting seriously competitive. Still some airframe durability issues, though...

I haven't really started the leaderboards yet; I've been waiting for competition to clarify. The MiK-4 is definitely the best of my lot at the moment, though.

Edited August 30, 2015 by Wanderfound

[Wanderfound]

PS: cranking the K-16 wing/tail strength/mass to 1.25 solves the airframe disintegration problem, but the extra mass aggravates the "diving into the ocean" problem. We could set the fights at a higher altitude, or you could try to find an extra G or two of turning ability (either a bit less mass, or a bit more control authority, or a bit more lift).

It's very much getting there, though; the MiK is spending more time dodging missiles than firing them.

[colmo]

Splendid, another AI dogfighter variant. I'm not au fait with FAR so the tips (inc. about Dynamic Deflection) are useful. I'll add this to the list of AI tourney threads on my discussion thread (see my sig).

[SpaceplaneAddict]

Could the AV-R8's be the problem???? Holy crap, I'm pulling 5G during takeoff, almost no AoA stall, and 7G during flight. Without the AV-R8s?????? Double-you-tea-eff:confused:.

[Wanderfound]

Could the AV-R8's be the problem???? Holy crap, I'm pulling 5G during takeoff, almost no AoA stall, and 7G during flight. Without the AV-R8s?????? Double-you-tea-eff:confused:.

Depends on what you mean by "the problem". You've got three major issues:

1) You need enough turning ability to dogfight effectively, which includes the ability to pull out of dives before hitting the ground. This means high-G supersonic maneuvering in thick, low altitude air.

2) You need enough durability to survive the consequences of those maneuvers. Wing mass/strength left at default is more than strong enough for any sane piloting, but low-altitude supersonic dogfighting is pretty much the definition of not-sane flying.

3) The plane needs to be docile enough that the autopilot doesn't flip it out of control, particularly while subsonic (once you've gone supersonic, it becomes much harder to stall). Ideally, you want the autopilot to hold it just at the edge of stall without falling all the way into it.

The first issue calls for as much control authority as possible, and as low a wing loading (mass divided by lift, basically) as possible. OTOH, the second issue demands strong wings, which adds mass, which increases wing loading. And the third issue calls for as little control authority as possible, directly conflicting with the first point.

Watch the video of the MiKs; they're also pulling highish G and momentarily entering a minor stall [1] on takeoff. It's normal for fighters to do that in KSP/FAR. You'll also see that the MiK-5 tends to stall a little itself when turning hard at subsonic speeds; it's not necessarily a major problem, although it is best avoided when possible (because even a minor stall adds drag and reduces lift, reducing acceleration, climb rate and turning ability)..

The change to the AoA settings on the canards has already fixed most of your stalling problem; toughen up the airframe a bit and gain just a bit more turning ability and you'll have a seriously effective fighter.

[1] The information around the navball comes from Kerbal Flight Data; the stall warning turns yellow during a minor stall, red in a major stall. Right-click on parts during flight to check if they're stalled and how much; this will let you know which bits of the wings need tuning. To reduce the stalling tendency of a control surface, either reduce its max deflection or increase its AoA setting (negative if in front of CoM, positive if behind CoM). If the main wing is stalling, either increase the chord of that wing or shift it rearwards a bit.

- - - Updated - - -

Just realised that I never posted here the MiK-5 video that I refer to above. It's this one:

Edited August 31, 2015 by Wanderfound

[Phearlock]

Here's an initial trial aircraft, I don't yet know quite what'll be optimal for this challenge, so I simply made an all-rounder. Decent low speed, subsonic and supersonic maneuverability, though I had to dump the roll rate to make it BDAI friendly. It's a bit of a large aircraft compared to what everyone else has posted so far, but most of it is structural fuselages and wings, so it should still retain acceptable acceleration, might revise it in the future to make it lighter.

Made in FAR v 15.5 (Will fly bad in 15.1) Also Wanderfound, do you have a download for the MiK4 somewhere? Just so everyone has access to the aircraft that's classified as the current leader.

Edit: I'll also suggest probably limiting defensive chaff/flares. As right now you can easily stick on 4 of each and since BDarmory AI missiles are both slow, crap at turning, have no range, and can't maintain a lock at all on anything that has some minor defensive things.

https://www.dropbox.com/s/wb2f5mibw9e4411/F-3C%20Pidgey.craft?dl=0

Picture!

Edited August 31, 2015 by Phearlock

[CommanderCoye]

HERE YOU GO

https://drive.google.com/file/d/0BznEolcLuLIlbHozN1FfRWpKUVk/view?usp=sharing

Sorry for being retarded!

EDIT: Just remembered I have some other stuff in the description, ignore it!

Edited August 31, 2015 by CommanderCoye

[Phearlock]

Um, @CommanderCoye, are you sure you submitted this in the right thread? This is for one of the FAR dogfight challenges, your craft looks very stock aero and doesn't have any of the FAR control surface configurations.

[Wanderfound]

Here's an initial trial aircraft, I don't yet know quite what'll be optimal for this challenge, so I simply made an all-rounder. Decent low speed, subsonic and supersonic maneuverability, though I had to dump the roll rate to make it BDAI friendly. It's a bit of a large aircraft compared to what everyone else has posted so far, but most of it is structural fuselages and wings, so it should still retain acceptable acceleration, might revise it in the future to make it lighter.

Shiny; I'll give it a test fly shortly.

Made in FAR v 15.5 (Will fly bad in 15.1) Also Wanderfound, do you have a download for the MiK4 somewhere? Just so everyone has access to the aircraft that's classified as the current leader.

Whoops, here it is: https://www.dropbox.com/s/9jqvn1dv8hxv7kv/Kerbofighter%20MiK-4.craft?dl=0

Edit: I'll also suggest probably limiting defensive chaff/flares. As right now you can easily stick on 4 of each and since BDarmory AI missiles are both slow, crap at turning, have no range, and can't maintain a lock at all on anything that has some minor defensive things.

Early testing without chaff or flares tended to turn the fights into just a contest of who fired the first missile; brief and not much fun. While the missiles don't hit that often when chaff/flares are deployed, what they do is force the AI into defensive maneuvering, which in turn allows the attacker a chance to get on the tail and bring the guns to bear.

I considered limiting it to no more than one chaff/flare dispenser per aircraft, but it didn't seem to have that big an impact; several chaff dispensers aren't that much better than one. I'm happy to change my mind on this if there's majority demand for it, though.

Edited September 1, 2015 by Wanderfound

[Wanderfound]

I'm now uploading the first MiK-4 vs F-3C fight to Youtube; rather epic, worth watching. They match up very well together, despite having differing strengths and weaknesses.

BTW: I've been running the fights with speeds matched as well as altitudes, in order to keep the planes together during the fight. They're all capable of going fast enough to blow themselves up, anyway; a live pilot would easily be able to match speeds if necessary, and matching them in the autopilot seems the best approximation of that.

[Phearlock]

BTW: I've been running the fights with speeds matched as well as altitudes, in order to keep the planes together during the fight. They're all capable of going fast enough to blow themselves up, anyway; a live pilot would easily be able to match speeds if necessary, and matching them in the autopilot seems the best approximation of that.

Are you adjusting the speeds up to fastest one, or down to the slowest one? Or are you meeting them in the middle (fastest goes down by as much as the slowest goes up). This is a huuuuge deal in terms of how to configure the AI. You need to add this to the OP ruleset in terms of you do fights. Or set a fixed max top speed that everyone should keep to.

Second. You either need to revise the MiK 4 or revise your ruleset, as right now it says "no reaction wheels" however the Mik-4 Does not have the cockpit reaction wheel disabled, this is a significant control authority boost for small aircraft, please follow your own rules.

[Wanderfound]

Are you adjusting the speeds up to fastest one, or down to the slowest one? Or are you meeting them in the middle (fastest goes down by as much as the slowest goes up). This is a huuuuge deal in terms of how to configure the AI. You need to add this to the OP ruleset in terms of you do fights. Or set a fixed max top speed that everyone should keep to.

Second. You either need to revise the MiK 4 or revise your ruleset, as right now it says "no reaction wheels" however the Mik-4 Does not have the cockpit reaction wheel disabled, this is a significant control authority boost for small aircraft, please follow your own rules.

So far, I've been running everything at 399m/s max; fast enough that subsonic stalling doesn't become an issue, but slow enough that the fights don't turn into extended long-range turning contests (which is what happened when I was fighting them at 500m/s; the planes just circled each other about 5km apart). Too fast makes for very dull video and inconclusive fights, because the autopilots aren't smart enough to throttle down when they need a tight turn. Mismatched speeds ruin gunfights; if the tailing plane is slower, it almost never gets within gun range; if it's faster, it tends to overshoot the target before having enough time to knock it down.

Even at 399m/s, close-matched fights run long enough that fuel consumption becomes an issue. Real-world time is also a concern; with video recording and multiple craft flying about, my game runs at 1/4 pace, so a fifteen minute dogfight actually requires about an hour of camerawork to record.

I realise that I've been fiddling with the rules, but it's with the intent of finding the best combination to create interesting fights and videos (I only started using BD Armory a few weeks ago, so I'm still finding out what works best), and because I haven't been thinking of the contest as "officially" started yet. I was waiting for a few viable fighters to appear, and taking the opportunity to fine-tune the scenario while doing so.

Once I start really tracking who wins what (which would appear to be around about now), the rules will be locked down. If you want, consider the upcoming first F-3C vs MiK-4 video as just the pre-fight warmup session, and we'll make sure everybody is clear on the rules and happy with them before the first "official" bout. I'd be happy with the rule of "all planes max speed set to 399m/s" if the other competitors are fine with that.

I'm not very competitive; I'm happy to see my planes blown up repeatedly so long as it makes for an interesting Youtube clip. I haven't been attempting to fiddle things in my favour; you'll note upthread that I was offering to stage the fights at higher altitude when another pilot was having dive-into-the-ground issues.

With the reaction wheel rule, I meant it more literally than you read it: no reaction wheel parts (small, medium or large). I didn't mandate disabling of capsule torque because I wanted to keep it accessible to relatively novice builders who will already be having enough trouble making something that the autopilot can fly smoothly. That's another rule that I'm happy to change my mind on if necessary, though.

[Phearlock]

Now I realize the 500 m/s thing, and I did configure the Pidgey to never get up to those speeds unless the opponent was extending hard (and once it's above 4000 meters even at 500 m/s it'll bleed speed pretty hard when it turns due to how I set up its control authority). The AI does seem smart enough to use speedbrakes to try avoid overshooting even though it doesn't have throttle control (hence the massive split brakes on the Pidgeys wings), if another aircraft has issues with maneuvering at those speeds due to failure to set good dynamic deflection settings or being too fragile, that's their problem. I do think enforcing a set max speed for all aircraft would work well though.

For the reaction wheel thing: I realize you may not be that competitive, but that's a poor excuse when you host a challenge like this that encourages competition. You need to be consistent. If you do want to keep cockpit reaction wheels allowed: You need to say so, none of the other FAR challenges currently running (5th gen fighter design thread, WWII dogfighting AI thread) allow them, I assumed no reaction wheels meant none. If you do want to keep cockpit reaction wheels as something that is allowed, I will design an aircraft around that (likely a small 2seater) to take full advantage of it, and it will indeed influence the later stages of this challenge to be biased heavily toward small aircraft that get the most out of the cockpit torque. Personally, I think they should be disallowed entirely to help keep the viable designs more diverse.

Edit: also wanderfound, the MiK-4 flies nothing like on the video you posted. Are you sure you have the same FAR version as I do? There was a pretty sizable update only about a week ago (v15.5 @ 25 august) that changed a bunch of lift and weight settings for wings, meaning a lot of craft required some minor/major re-tuning. I suspect you may have an older version as the MiK-4 seems to pull way too much AoA and spiral out of control really easily at low speed.

Edited September 1, 2015 by Phearlock

[CommanderCoye]

Um, @CommanderCoye, are you sure you submitted this in the right thread? This is for one of the FAR dogfight challenges, your craft looks very stock aero and doesn't have any of the FAR control surface configurations.

I find your lack of faith disturbing. Did you test it?

[Phearlock]

I find your lack of faith disturbing. Did you test it?

Sort of, it exploded from part clipping while still on the runway 3 times in a row, and the FAR stability analysis had it as super stable (aka, probably wouldn't be able to turn very well). Along with none of the control surfaces having anything but the default values (so ailerons, rudders and stabilators were all set to both pitch/yaw and roll.)

Edited September 1, 2015 by Phearlock

[CommanderCoye]

Sort of, it exploded from part clipping while still on the runway 3 times in a row, and the FAR stability analysis had it as super stable (aka, probably wouldn't be able to turn very well). Along with none of the control surfaces having anything but the default values (so ailerons, rudders and stabilators were all set to both pitch/yaw and roll.)

Ooooooo that's bad. I'll make a redesign then.

[Wanderfound]

F-3C vs MiK-4 testflight; very close, with a fair amount of luck involved on both sides. I wouldn't want to bet on the outcome going the same way in a rematch. Good fight, though:

- - - Updated - - -

Edit: also wanderfound, the MiK-4 flies nothing like on the video you posted. Are you sure you have the same FAR version as I do? There was a pretty sizable update only about a week ago (v15.5 @ 25 august) that changed a bunch of lift and weight settings for wings, meaning a lot of craft required some minor/major re-tuning. I suspect you may have an older version as the MiK-4 seems to pull way too much AoA and spiral out of control really easily at low speed.

My last FAR update was a couple of weeks ago; I'll grab the new version today.

I'm happy to go with a "cockpit torque must be disabled" rule; I'll update the OP accordingly, along with a "compulsory 399m/s max speed" requirement.

[Wanderfound]

Edit: also wanderfound, the MiK-4 flies nothing like on the video you posted. Are you sure you have the same FAR version as I do? There was a pretty sizable update only about a week ago (v15.5 @ 25 august) that changed a bunch of lift and weight settings for wings, meaning a lot of craft required some minor/major re-tuning. I suspect you may have an older version as the MiK-4 seems to pull way too much AoA and spiral out of control really easily at low speed.

Just had a test with an updated FAR; as far as I can tell, the MiK is flying the same as before. Did you have Dynamic Deflection installed on your test?