Advanced Air Superiority Fighter Competition

[Doke]

So we're back at it. I'm not sure that I relish the prospect of facing the Hummingbird again, but I'll give it a shot.

I really like the idea of weight classes. Given that the current 'meta' has been tending toward 'lowest possible wing loading', some kind of external constraints to make planes a bit more well-rounded seem like a good idea. I was seriously thinking about a sort of 'intercept challenge' or 'bombing challenge' to mix things up, but the AI and physics limits seem to render it moot.

Anyway, here's one I had hanging around that works perfectly for the heavy category. The Dok-7 Currawong. A triumph of thrust over aerodynamics. I haven't even bothered to set up DD for it, because it seems beside the point. All stock except for the BD Armory stuff.

I've also got a nifty little medium which is a great plane to fly, but it might need some work to make it more competitive under AI control: the Dok-9 Crow. Needs B9 procedural parts and adjustable landing gear.

 

[gridghost]

Tried to make a somewhat supermanouverable heavy fighter to compete with Dok-7 and came up with GG HF-2X Proto (right click...). It's pretty stable post stall and able to turn pretty hard. All stock except adjustable landing gear.

[PTGFlyer]

When it says B9 procedural wings, can we use other B9 stuff too?

[gridghost]

It says procedural parts in the OP so i suppose so.

[Van Disaster]

6 hours ago, PTGFlyer said:

When it says B9 procedural wings, can we use other B9 stuff too?

If it's not in the OP, no; that's been the usual rule. You can use the other procedural parts though, given that *is*.

Doke; that was why I started giving myself time in flight limits; the Arrow frame has way more fuel than you'd use in one of these dogfights, if you turn the burner off it'll fly halfway round Kerbin ( and the last Hummingbird I put up is about the same I think ). Let's not forget we're building basically subsonic interceptors and not multirole aircraft though, so low wing loading is a perfectly realistic aim - and to be honest the only time the lowest wing loading you can manage *isn't* something you want is a low level strike aircraft. The current design philosophy is mostly limited by the tuning tools we have, really; if Baha ever releases the BDA he wanted to do with much more designer input into the AI, that'd be a very different state.

GG's price brackets are another quite good limiter especially if you combined them with duration, so you'd have to pay the fuel cost out of the budget. I did this for the heavy bracket but haven't got round to tuning it yet;

 

One other possibility is to allow kOS, which would let you actively change engine parameters amongst other things.

Edited January 30, 2016 by Van Disaster

[Doke]

The HF-2X is a very capable plane, but you might need to tweak dynamic deflection a bit. The AI has a tendency to push the envelope a little too far and get into a bad stall or over-g and rip a bit of wing off.

Ideally, we could just set AoA and G limits, but I have no idea how to go about implementing that.

@Van Disaster

I think kOS might be going a bit far, unless you want a programming contest. I guess I can't shake the feeling that these fighters ought to be supersonic, but that's nothing to do with this contest. I like to go fast. I've been toying with the idea of some kind of successor to the 5th Gen Fighter Contest -- a 'renaissance fighter contest' with combat as a scoring mechanism but a few other requirements (fuel fraction, speed, paylaod, etc.) for qualification. Meh.

[gridghost]

7 hours ago, Doke said:

The HF-2X is a very capable plane, but you might need to tweak dynamic deflection a bit. The AI has a tendency to push the envelope a little too far and get into a bad stall or over-g and rip a bit of wing off.

Ideally, we could just set AoA and G limits, but I have no idea how to go about implementing that.

Now, that would be a very nice addition  

[Van Disaster]

AoA & G-limiting can be done by picking the right max control deflection & DD for our flight envelope; it's pretty rare these planes go above 5-6km, so you don't really have to worry that much about high speed & low Q. What we need ( I've said this before ) is the engine gimballing treated as a control surface or two as well. Don't try and make a dynamically unstable supermaneuverable plane for the AI, it's not good enough to fly it. Dynamically stable is fine anyway.

Take your plane up & manually fly it at the envelope limits ( that's a lot easier than flying it at half the limits... ) - that's basically fly straight until it either hits 400m/s or top speed - bank 90 degrees & pull back. If it doesn't fall apart, add more control deflection until it does!. Then just fly straight & back off the throttle, and adjust DD to keep it's AoA in check. Do the first test again but mess with entry speeds and play with the throttle mid turn, keep an eye on the AoA you're pulling & tweak DD again.

In essence that's all I do, afterwards is just running lots of test fights & eyeing the FAR info panel while they're going on.

Try adjusting the allowed top speed upwards for a few fights & see if it makes any difference. I suspect in the end it'll still come down to a lowish speed dogfight.

Edited January 30, 2016 by Van Disaster

[Doke]

How are people finding the new BDA? Lots of interesting tweaks and features for air duels. Missiles seem much more effective, especially AIM-120s (will we start seeing jammers?), and the AI more competent at flying in general and post-stall shenanigans in particular. I just saw the the Currawong and the HF-2X engage in some gun jousting, which was absolutely insane. Thrust vectoring is even more viable now: the AI is much better at stall recovery, and knowing which axis to use for post-stall maneuvre.. 180 degree turn on the yaw axis at 50 m/s for an AIM-120 kill? Sure, why not. I'm really exited to see the AI flying my creation the way it was intended.

[gridghost]

I have to download and test it when i get home  Seems like it's going to be hilarious and fun to watch the AI do some aerial acrobatics.

[Van Disaster]

Huh wow, this is a bit improved, eh. Can't see a way of disabling panther wet mode anywhere which worries me that my lighter craft are going to break up, but the latest Hummingbird seems to love it.

Edit: having said that the AI still pulls some really dumb moves, just a different set of dumb moves...

Edited February 2, 2016 by Van Disaster

[gridghost]

Aerial acrobatics: check

Controlled stall: check

Gun jousting: check

Incredible stupid moves by the AI as in killing the engine and just hanging just infront of your enemy just wainting to be shot down: check.

Amazingly cool moves by the AI as in hanging on thrust alone, yawing with vectors to get missile lock and kill: check

And the Currawong does not dissapoint

But i have to redesign my crafts a bit... *sigh*

[Van Disaster]

Hum, this is pretty annoying actually, the AI seems to assume you're using thrust vectoring given what I've seen it try & do with something that isn't; not only that it's even worse at bonehead moves to put itself in front of opponent's guns. Seems to have no concept of either "extend" ( except when it doesn't need to ) or lag pursuit, would rather turn complete circles inside & both bleed all it's energy off and given an ideal firing solution for the opponent while it does. If you want to see some really ugly piloting try my Wheesley powered craft ( the F4 version of the 1.0.4 winner ). The control PID seems even worse now too.

Also seems to crash the game a lot more

Edited February 2, 2016 by Van Disaster

[Doke]

Yeah, the AI seems to be much more aggressive with the controls now. I've found you need to tweak the settings a lot to make it more gentle for some planes. Unfortunately, I don't think it's ever had any concept of extending. Maybe speed controls?

I see BD just released a hotfix with some AI changes.

[Van Disaster]

It did try and extend before, only it wasn't very good at it - remember when planes were losing a turning fight & rolled out the other way, and then flew nice & straight to give a good target? all they really needed to do was dive to accelerate & just loosen the turn radius a bit, that'd at least increase seperation given the other plane would still be in a hard turn trying to get a firing solution. There's also the concept of corner velocity ( which is an idea for energy fighting rather than BD's preferred supermaneuverability ), which might make more sense as a slider speed than the minimum one. That is something we've been setting with DD, though.

There were some actual bugs, like gun distance not working & min speed not really being effective - will see what got fixed. Also the planes seem really reluctant to use chaff, which might explain the AIM-120 effectiveness...

Edit: well, can't really tell if any boneheadedness was fixed - the default response to a mistake now seems to be a flat circle though, which is a really strange response to a vertical overshoot for instance...

Edited February 3, 2016 by Van Disaster

[Doke]

Ran a couple of test battles this evening. Conclusions:

(1) The AI is allergic to immelmans.

(2) Supermaneuverability is very much a double edged sword. Running mirror matches, what usually happens is that one or both planes is destroyed by an AMRAAM in the merge. Even with fairly conservative settings, the AI just loves to burn energy. I'll give ECMs a shot later and see if it makes a difference, but otherwise energy fighting might be essential.

[gridghost]

I did some testing and got a good video to boot  

No sound though. But it's a showcase of energy- against supermaneuverability-fighting.

 

Spoiler

https://www.youtube.com/watch?v=YpwEuMDe06U

 

 

[Van Disaster]

The AI is almost allergic to the vertical but it's been barrel-rolling to avoid overshoot when it's catching another craft in a horizontal chase - often triggers something akin to rolling scissors - so I've no idea why that behaviour isn't triggered for other situations ( like the same move in a dive... ). And yeah, flat U-turns vs an Immelmann is insane, unless you're doing an airshow.

The easy answer to fighting a supermaneuverable plane would be to sucker it into slowing down & just extending & lobbing missiles at it, but good luck getting anything to do that :p. Mind you a real pilot is not going to stall out in combat unless it's extremely safe to do so, rather than convenient.

Want to try your HF-2X vs the P9.3?

Edited February 4, 2016 by Van Disaster

[gridghost]

That last fight was a new design: GG LHF-x12 Proto (right click...) 

Still prototyped, but good enough  

Have fun.

Though i guess the P9.3 will beat it outright, seeing as it weights approx. half of the LHF and has the same turnratio  Damn, it's a hard one...

Edited February 4, 2016 by gridghost

[Van Disaster]

I'll try it. I livestreamed a bunch of fights when I was attempting to get your guy's supermaneuverables to actually do their stuff, but it wasn't until about the 6th run with the Currowong that it started; did showcase a lot of AI behaviour though ( especially how bad it is on the defensive still :S ).

 

 

 

 

Edit: hrm, how does one do spoiler tags now...

 

Edited February 5, 2016 by Van Disaster

[gridghost]

I noticed that the Currawang have one engine on 80% thrust and the other on 100%, btw. Makes it somewhat more unstable.

[Van Disaster]

That's pretty nice! streamed a bunch of fights with it.

 

The Hummingbird managed to shoot itself down with one of it's own sidewinders, that's a new thing. It's a pity it can't shoot straight...

Ugh I detest this forum software more every day.

Edited February 5, 2016 by Van Disaster

[gridghost]

I see it actually managed to get 1 (!) kill out of, what, 15 fights

But I need to down the partcount and have less vectoring and more DD (default doesn't cut it really) and see to making it preserve more energy. But it looks cool

[Van Disaster]

Seems extremely hard to stop the AI pulling into a stall even with a dynamically stable craft; the minimum combat speed doesn't seem to do much. The reason the P9.3 ( 9.4 in the video is just a change of AI parameters ) works is because somehow I managed to hit a gigantic sweet spot; I tried swapping the nacelle for a 400U tank & some intakes and the extra mass - even though it's in the same spot and the CG is still in the middle of the tank - means the AI just bleeds speed like mad in turns in a way the original doesn't even come close to doing ( it's only 250U of fuel heavier ). I'm not really sure how to tune that out, because if you reduce control authority that means it still tries to do the same thing but lacks the tools to get out of it :p.

I tend not to count head-on missile kills when I'm tuning, becuase the AI can just not notice missiles sometimes even with the TI pinging away & the thing coming from near head on. Not that it doesn't count as a kill, there's just nothing you can do about it.

Edited February 5, 2016 by Van Disaster

[gridghost]

Ya, that's what i found playing around with my rules, too. That's why (and for endurance reasons, too) i opted for at least 200 units of fuel and an engine restriction of 75% thrust on my Light Fighter class. Otherwise it would outmanuever the other classes by miles. It still does, though :). I'm trying to get to a restrictionlevel that seems fair to all of the classes, and this is what i came up with after some trial and error. I like long and intresting bouts where both actually have a chance of survival and success.

And to clarify, i usually dont count head-on missile kills either. But you have to give it something