colmo

One scenario where lag is a major issue is bases, with numerous craft in physics range. Even with one craft per thread, that'll be a huge boon for colonisers. Just to be safe, I'm getting a hexcore with hyperthreading...

Baha has stated AI is next up once radar and GPS is finalised.

I want to be able to run a load of stuff in the background (inc. video capture), and Haswell-E is quad channel, hence the 16gb v 32gb decision. More RAM is also handy for photo (big panos) and video editing. A huge surplus of RAM means I could mount KSP in a RAM disk to really speed repeat reloads.

I was speccing my new PC build (Haswell-E, Linux) with 16Gb. Now seriously considering 32Gb...

I envision a Top Gun AI Pro ruleset which embraces FAR and a few mods (especially procedural parts, and some engines and cockpits) to take plane design to the next level. It'll also make extensive use of radar - no more 360 degree pilot visibility, and might introduce air to ground as part of the setup conundrum - take out the radar and blind the enemy, or concentrate on air superiority regardless of radar? I'll need a new machine to run it, which might take a few weeks.

Loving the preview of the Panther engine - there is indeed a yawning gap between Wheesley and Whiplash. http://m.imgur.com/a/NkzMp

With GPS now implemented, the tools are coming into place for AI auto-landing. The AI would need: - a GPS coordinate for ideal landing spot, - heading so it approaches from the right direction (aligned with runway), - a tweakable glide slope, and - tweakable ideal landing speed (so it can cut throttle and/or deploy brakes to achieve desired approach speed). A button to lock current position and heading for the first two would be handy also. Other useful options: - desired timer from last enemy contact until landing begins. - cruise speed for circling behaviour when no enemy in contact (saves fuel) I've previously suggested ways to allow AI to fly VTOLs by swapping pitch and throttle response to control altitude and speed. Planes gain altitude by pitching up, VTOLs by increasing thrust or collective. Planes gain speed by increasing thrust, VTOLs by pitching down. VTOL mode would enable yaw steering control. Tweakable roll and pitch limiters would be useful for both helicopters and non-aerobatic planes. Landing AI for VTOLs would then be trivial - just reduce landing speed to a small number. Having VTOL / HTOL behaviour toggle on an action group bound to gear allows AI to switch between the two; vertical takeoff with horizontal flight (a jump jet) or horizontal take off and VTOL type flight (e.g. some kind of jet assisted helicopter) This way, we have to design our VTOLs around BDA instead of BDA having to cater to us.

The Gladius doesn't look so bad. Wing strakes are pretty much designed to clip a bit into awkward curving fuselage and add minimal lift to the equation. An eyeballed 50% is what I suggested, we don't want to get bogged down in scrutineering. I appreciate that making wings look nice in stock can be tricky. When you've got clipped wing parts which are clearly not for looks but redundancy and drastically shifting and increasing lift, that's just exploitation of the aero model. We don't want to go to FAR because that's a new thing to learn and an extra burden on the recording rig, FPS does suffer according to Scott Manley. I gave this idea for free use so I'm mindful I'm not sole proprietor. You kids will have to thrash it out for yourselves.

My initial draft.

My personal rules on wing spam and clipping are: - no sandwiching of wings - elevons can only be attached to a fixed wing's edges, either leading, trailing or tip. - control surfaces should not be significantly clipping though other parts in their normal range of movement. - fixed wings should not clip inside fuselage more than necessary for seamless fit and finish - over 50% of surface area concealed is a definite no-no. A specific exception is engine nacelles built into the wings. Edit: Engines shouldn't be clipped by any wings, and I try to avoid slicing through intakes with wings. Edit2: An elevon may be fitted to fuselage or wing surfaces if, and only if, it has all pitch, yaw and roll deactivated (I.e. it's used as an airbrake or fixed winglet)

Having watched my plane lose to a 'design' which outmanoeuvred it with one wing, I find myself thinking how "it's hard to legislate wing clipping" becomes "I'm going to completely take the .... with wing clipping and take a steaming dump on the spirit of the contest as originally envisaged"? Apologies for the bluntness, but I'm disgruntled at the abuse of the contest. It doesn't even remotely meet my idea of fair. I don't mind my plane losing, because I'm happy I didn't cheat to build it.

Chaps, Top Gun AI is about to get the Manley treatment: https://twitter.com/bahamutod/status/634104546130792450

I think my designs fared poorly because I was a little too attached to the Mk2 parts which are larger than necessary, making planes long and thus relatively poor in pitch authority. I suspect they'd translate to a FAR environment better than the 'peanut fighters' which have dominated this particular tourney. Planes with a wingspan greater than their length have done best in this one. I can imagine the popular mk1-2 adapter used to make them so small would be enormously draggy in FAR. No complaints, the original thread's poll showed a stock aero preference. The WIP plane isn't quite the same though...I don't want to spoil the surprise!

I don't believe G-Effects would impact play at all...though missing a crucial moment due to blacked out screen would be annoying. Make sure you use pilot Kerbals, they don't black out so easy. A WIP plane of mine can pull 15G (that's not a typo), which would make this mod rather interesting...

I tested this recently and while in stock my carrier sank to the bottom, in Better Buoyancy it actually floated, albeit with a very low bow. I suspect some tweaking of fuel to balance CoM with centre of buoyancy would fix it the rest of the way. The ramp was mysteriously invisible but there are other parts that could replace that.

Kudos. I did mean to write up the rules as we've mostly settled on but a new job took precedence! Every tournament we run we refine and learn. I have some quite elaborate plans which build on what we've developed.

Even discounting instakill, that's a good argument to reduce their points cost. One to remember for when guns work again. I'll head on over to YouTube and add the new videos to the #TopGunAI playlist. - - - Updated - - - A minute into Round 16 illustrates my point about crashes - the location of the VAB could have made the difference between time to crash of two planes that shot each other down. Should be counted a draw, not have you raking through the flight log to find which plane blew up first.

Iirc,the Vulcan is 5500 rpm, GAU 4000 rpm or so? GAU bullets are also slower. It's in the configs, I may be wrong. Guns are not ideal at the minute so fudging with instakill may be necessary. Knowing this in advance, though, I'd have spammed .50 cals.

There is judgement required - I'd give it about 10-15 seconds to see if the victor immediately crashes or is shot down by a missile or guns by a death-rattle shot. It's best to decide at the outset and stick with it.

Instakill is another rule change after the fact - why pack a GAU8 if a lighter, cheaper, faster Vulcan will do better. Once you publish rules and receive submissions, you can't be constantly tinkering with things that upset design balance.

Respectfully, that is a daft rule. In my original draft, I considered a plane dead when it was no longer under powered, controlled flight. It could still shoot, but it counted as dead, so in a 1v1, as the only team member, it loses. Planes can glide a long time without engines, and which crashes first is randomly down to what their vector was at the time. It's not fair to judge which did better by which one's wreckage blows up last. In a collision, the only fair result, if both were live before, is a draw.

I rather enjoyed the bout between my two planes - the 107 would have won if starting positions were more distant. It's not a true air superiority fighter, the 141 is, so no complaints!

Well it was in Kerbtown, as seen with the rather nice ICBM silo mod. The source code has been made available. http://forum.kerbalspaceprogram.com/threads/47065-0-21-ICBM-Launch-Silo The Kerbal Terrain System makes it possible to hollow out a nice steep pit in which to drop such a static model.

Yes, this sort-of gives us back what we lost from 0.8.3 - auto-aim for nearly accurate shots. It might be an interim solution, one to consider for the next contest. I've added all these videos to the Top Gun AI playlist on YouTube btw. They're actually a useful archive for even Bahamuto D to see. I'll send him the link.

I strongly recommend any future tournaments, if using a similar layout, use an anti parallel (opposite directions) layout. As it is, the first plane to bank 90 degrees and fire at opponent that is far too close nearly always wins. It doesn't make for many interesting contests. I'm also baffled by countermeasures - it seems you need to use a ridiculous amount of them to be effective, which I think is why my CK-141C lost - it did get first shot, but missed every time. Still, with the previous contest stalled we're getting close to our first completed contest, which Alphasus deserves credit for.