WWII BAD-T V: The AI Strikes Back - BD AI Dogfight Tournament

[SuicidalInsanity]

Without more info, best guess is your elevators are stalling out - try reducing pitch authority a bit and see if that helps.

[Guest]

"Several Marks Later"

Currently it just about out performs anything i can throw at it in terms of speed acceleration and climb rate, and it has quite good maneuverability. AI aiming is a bit off and needs some tweaking before submission.

 

Edited October 19, 2019 by Guest

[Pds314]

1 hour ago, dundun92 said:

"Several Marks Later"

Currently it just about out performs anything i can throw at it in terms of speed acceleration and climb rate, and it has quite good maneuverability. AI aiming is a bit off and needs some tweaking before submission.

 

That's a beautiful engine configuration.

5 hours ago, Jeb-head-mug kerman said:

Help! My plane keeps oscelating up and down whenever I use the elevator!

AI Damping and anti-stall features. Also use far to see if it is in fact stalling.

[Box of Stardust]

Looks a bit like a Dora, hits like a Dora. Also comparatively slow to its peers like a Dora...

I'm starting to feel like I maybe should have done a heavy fighter (because engine power and speed and climb rate), but I spent 8 straight hours on this getting it to a spot where I think it's pretty good, starting at hour 0 with no real experience in building combat planes in FAR, and thus, no real idea what a good baseline was for flight performance other than observing aircraft in this thread.

Edited October 19, 2019 by Box of Stardust

[Jeb-head-mug kerman]

4 hours ago, Pds314 said:

That's a beautiful engine configuration.

AI Damping and anti-stall features. Also use far to see if it is in fact stalling.

Well, I've found it actually does stall whenever I pull the elevator. Anything I can do to fix that?

[Pds314]

1 hour ago, Jeb-head-mug kerman said:

Well, I've found it actually does stall whenever I pull the elevator. Anything I can do to fix that?

If the wing is stalling, one option is leading edge flaps with about -40 to -80 AOA% control so that they can catch the airflow and redirect it onto the top of the wing at a less sharp angle.

Another is just to decrease the aspect ratio of the wing. Although there are limits to how big a gain this can really be. Long story short, it means you can pull higher AOA but end up with similar lift coefficient at stall. It does not improve wing efficiency or lift coefficient but raw wing area and only if wingspan is not shrunk in kind.

Another option is to give it trailing edge flaps that increase the lift at a lower angle of attack. Almost all WWII fighters had some kind of combat flaps they could deploy for this purpose, although they were typically at a fixed setting, not dynamically adjusted by AOA. I've seen everything from 25-200% AOA for these.

My entry, and my test vehicles, do all of the above.

Edited October 19, 2019 by Pds314

[Jeb-head-mug kerman]

1 minute ago, Pds314 said:

If the wing is stalling, one option is leading edge flaps with about -40 to -80 AOA% control.

Sorry if I'm being stupid, but there are flaps in FAR?

[Pds314]

Just now, Jeb-head-mug kerman said:

Sorry if I'm being stupid, but there are flaps in FAR?

Well what I mean is control surfaces set to negative AOA%. There are also flaps but don't actually use them.

Edited October 19, 2019 by Pds314

[Jeb-head-mug kerman]

7 minutes ago, Pds314 said:

Well what I mean is control surfaces set to negative AOA%. There are also flaps but don't actually use them.

My thanks! My plane now flies as it should, and all it needs is some more manouverability! I'll post it tommorow.!

[Pds314]

4 minutes ago, Jeb-head-mug kerman said:

My thanks! My plane now flies as it should, and all it needs is some more manouverability! I'll post it tommorow.!

My understanding of why this works, summarized in MS Paint.

[Pds314]

Messing with the airfoil geometry mid-flight really improves lift by a huge amount. Obviously this is for real life, but FAR tries to model real life and does a fairly good job most if the time at modeling nuanced effects like this.

 

From an article called "maximum lift coefficient, an overview."

 

"The maximum lift of aerofoils which have rear separation depends on the geometry of the rear part of the section; the camber is also a significant parameter for both types as we saw in the previous section. Thin (τ < 0.08), smooth, symmetrical aerofoils, which inevitably have small leading edge radii, have C_Lmax values of 0.9 or less. The highest values are achieved by aerofoils with fairly large thicknesses and leading edge radii and so have rear separations. Values of the order of 1.6 for conventional sections and up to about 2.0 for modern aerofoils which have been specifically designed for high lift can be achieved."

Point is, flat boards facing into the wind are quite suboptimal as wings. At least if high lift at high AOA is the goal.

Edited October 19, 2019 by Pds314

[GDJ]

I confirmed that the AI is completely incapable of hitting another airplane that does not change it's speed, altitude, or heading.

At least I can fly for 45 minutes in a battle sim. I'll just outlast everybody and force them to hit the ground.

 

Edited October 19, 2019 by GDJ

[Pds314]

3 hours ago, GDJ said:

I confirmed that the AI is completely incapable of hitting another airplane that does not change it's speed, altitude, or heading.

At least I can fly for 45 minutes in a battle sim. I'll just outlast everybody and force them to hit the ground.

 

Unless your non-maneuvering target is much faster or at a wierd altitude I would have to disagree... look at the video I posted earlier of one aircraft following another. The AI can deliver accurate fire from that sort of distance. Not maneuvering would make its job much easier.

[SuicidalInsanity]

Is it a case of the AI consistently hitting a non-existent target floating a dozen meters off the target's flank? Mess around with the steerKi setting, see if you can clamp down on the steering error, or reduce the max engagement distance. Are your guns aligned properly? (hit F2 in the SPH + rotate gizmo). Switching out for less accurate guns like the MG131 or ShVAK might help, if nothing else, bullet spread might land a hit or two and force the target to change speed/alt/heading.
 

Edited October 19, 2019 by SuicidalInsanity

[Box of Stardust]

I just noticed the Advanced AI settings were hidden.

welp, time to go adjust those too

 

Craft aren't downloaded until tomorrow, right?

Edited October 19, 2019 by Box of Stardust

[GDJ]

1 hour ago, Pds314 said:

Unless your non-maneuvering target is much faster or at a wierd altitude I would have to disagree... look at the video I posted earlier of one aircraft following another. The AI can deliver accurate fire from that sort of distance. Not maneuvering would make its job much easier.

Awesome reply, but the target and the aircraft being built are within 100 m of altitude , and at most 10 m/s speed differential, which fluctuates slightly from either catching up or being pulled away.

I dunno.......I'll finish my plane and submit it and hope for the best. All I can do.

1 hour ago, SuicidalInsanity said:

Is it a case of the AI consistently hitting a non-existent target floating a dozen meters off the target's flank? Mess around with the steerKi setting, see if you can clamp down on the steering error, or reduce the max engagement distance. Are your guns aligned properly? (hit F2 in the SPH + rotate gizmo). Switching out for less accurate guns like the MG131 or ShVAK might help, if nothing else, bullet spread might land a hit or two and force the target to change speed/alt/heading.
 

1) Anywhere from a few mm's to about 2 dozen metres, yeah.
2) I've adjusted the SteerKI until I'm blue in the face, moving it at 0.5 increments.
3) Tried adjusting guns to both being dead straight forward to converging at around 1500m.
4) I think I tried those guns. I picked the guns with the fastest muzzle velocity (over 1000 m/s).
5) Also tried adjusting the steer dampener to smooth out the jerky motions in the plane.

Ahh whatever. I'll port the AA antenna settings to the BAD-T config file version of the cockpit and submit it tonight. I worked hard enough on the aerodynamics and the wing shape that it can nail 270 m/s in a shallow dive, pull out and line up with the target that was pursuing it. I'll hope for the best. I'm good.

Maybe I'm just comparing the craft with FAR to my last tourney that was using stock physics. That's probably it.

Edited October 19, 2019 by GDJ

[SuicidalInsanity]

44 minutes ago, Box of Stardust said:

Craft aren't downloaded until tomorrow, right?

Monday morning. So you have all of today and basically all of tomorrow to finalize things.

[Guest]

@GDJdo you have any steer ki on your AI? if not that may be the issue.

[GDJ]

1 hour ago, dundun92 said:

@GDJdo you have any steer ki on your AI? if not that may be the issue.

Yes I do. I have adjusted it.

[Pds314]

49 minutes ago, GDJ said:

Yes I do. I have adjusted it.

What are steering and damping factors? Above default and in a roughly 2:1 ratio or not?

[Guest]

37 minutes ago, Pds314 said:

What are steering and damping factors? Above default and in a roughly 2:1 ratio or not?

2:1 is a bit harsh for damping, 3-4:1 is much better imo, just have to use sub 2 damping values

[Jeb-head-mug kerman]

Craft almost done. Now testing against the Dummy.

[GDJ]

2 hours ago, Pds314 said:

What are steering and damping factors? Above default and in a roughly 2:1 ratio or not?

 

1 hour ago, dundun92 said:

2:1 is a bit harsh for damping, 3-4:1 is much better imo, just have to use sub 2 damping values

Good point you two. I'll check it out.

I also haven't been using the Dummy to test the craft. I'm still using the Kraken Mk1, which is STILL undefeated........ever.......anywhere.......just ask @Triop
 

 

Edited October 20, 2019 by GDJ

[Pds314]

1 hour ago, dundun92 said:

2:1 is a bit harsh for damping, 3-4:1 is much better imo, just have to use sub 2 damping values

I think it really depends on the characteristics. I use 2:1 because I like using 25:12.5 as the exact values so that it has really good reaction time in sync with the physics timestep. And any lower damping will result in control bouncing.

Edited October 20, 2019 by Pds314

[Jeb-head-mug kerman]

Finally got the plane ready!

My entry the ENDRE ER SX 13 'Hornet'

It is in the Heavy fighter class, and should be quiet manouverable, and is also very stable. It is also quite fast, capable of doing 234 m/s in level flight!

It is armed with 2 Hispana's and 2 more 20mm guns (of the name of which I can't remember:).  I based the design of the Mosquito bomber.

The poor Dummy tried its best but it was not fast or manouverable enough!

Here it is!https://kerbalx.com/JebHeadMugKerman/Endre-Er-Sx-13-Hornet-II

UPDATED VERSION!!!!:https://kerbalx.com/JebHeadMugKerman/Endre-Er-Sx-13-Hornet-III

http://

http://

http://

fighting the dummy

I am quite happy with it, considering this is the first successful craft I have made in FAR.

 

Edited October 20, 2019 by Jeb-head-mug kerman