Old Rhinebeck Aerodrome AI Traffic Package

[mick]

Progress!!!

I think I figured out why I was having such trouble with the AI planes! ;D

I had them just taking off and landing back at ORA, as they do at the real ORA. Apparently they never got far enough away from the field for FS9 to ignore them, and it always thought they were about to land. So any plane that tried to take off while there was another plane in the air would get held at the hold short node until it disappeared. And since there was usually a plane waiting at the hold short node, planes on approach would go around instead of landing until they too disappeared.

The solution, it appears, is to have the planes go somewhere and return. By overlapping flights it's possible to have a plane take off and appear to return and land in twenty minutes or so, but really it's flown away and a copy of it has arrived to land in its place.

I've only tested this with a very abbreviated traffic file, but it seems to work. I'm now in the process of completely rewriting the traffic files. Then we'll see if the whole scheme works. If it does I can keep to the plan of having a takeoff every fifteen minutes and a landing in between each takeoff.

Where do the planes go? Well, they go south. I found a little grass strip in northern New Jersey about 100 miles from ORA and that's the origin and destination airport. It has no parking, so the planes won't clutter up someone's home airport. They'll just land and disappear. On the return flight they won't start at that airport, they'll just pop into existence in the air when their flight plans put them 40 miles short of ORA.

It works with three planes. Now I have to write the traffic files and see if it works with fifteen planes.

Meanwhile, I came up with a great cannon shell effect for the tank. I modified it from one that came with the Microsoft Effects SDK. The firepower is awesome!

Stuart is back from his vacation, but he's having confuter trouble. Software trouble, I should say - his FSDS won't work! That will delay the availability of the special low-poly AI Caudron and SPAD, but it gives me more time to write and tweak the traffic files. And to start working on the ReadMe (groan...)

[Slide]

Respect! Applause!! I did not know it is that difficult to work with Artificial stupIdity. Happy you worked it out. And that´ll be quite a little Readme ...

[pooch]

If the final testing will confirm the results you got with the abbreviated traffic files, this would be really a major finding ! Because what will work in ORA will be true for all other airports when populating them with major AI-traffic.

I am following your progress with much interest.

Thanks and compliments.

[mick]

Well, I wrote the traffic files for fifteen planes and what worked for three doesn't work for fifteen! After the first one or two take-offs, planes start to get stuck at the hold-short node.

I set up a take-off every 20 minutes and a landing in between each take-off. The plane that's due to arrive next isn't in the vicinity of the airport yet, and the plane that just landed has parked and shut down, but the plane about to take off still gets stuck.

I even tried making the take-offs a half hour apart, but after a couple of take-offs planes would still get stuck.

I've just about run out of ideas!

If you refresh the Time & Season when a plane gets stuck it will pop back to its parking spot, taxi out and take off, but then the next plane or the one behind it will get stuck. And if you refresh the scenery when there's a plane almost ready to arrive, it will appear on the ground, parked and shut down, so you miss seeing it land.

This is especially frustrating because I have more activity than this at several busy airports in my "Flight Simulator 1954 - A Half Century of Flight" installation and there are no problems like this. My local Air Force Base has 16 planes present at all times and a take-off or a landing every five minutes, and it all runs like clockwork. I thought the difference was because of the speeds of the planes. Those early jets and high performance props spend a lot less time on approach than the planes of ORA. But at ORA, even with take-offs spaced so that the last plane down is parked before the plane taking off starts up, and the next approaching plane is still twenty minutes from touchdown, planes taking off still got stuck. So I went back to a 20 minute take-off interval because extending it to half an hour didn't help at all.

If I make the interval between take-offs less than 20 minutes, planes on approach execute missed approaches and go around until they disappear. With a 20 minute interval at least the approaching planes get to land.

Unless someone has some new ideas, I think I'm going to have to release the package with planes getting stuck listed as a "Known Issue" (meaning, "I know about this but I can't fix it") and make a note that you should refresh the Time & Season when a plane gets stuck.

I wonder if folks with more powerful confutors will have this problem. Maybe the whole set-up is more than my system can handle. Golden Wings is resource intensive, and ORA is very resource intensive, so maybe that's why I can run a more intense traffic scheme at my other airports in my other sim... ?

If anyone has any other ideas, please let me know!

[Slide]

[pooch]

Too bad...
What do you think of placing the question in one of the specific AI forums like Project AI or Aardvark AI ?
Maybe some of those specialists know about the interaction between landing and departing planes: which is the exact moment of a landing plane that makes a departing plane stop at the hold point. Is it the landing plane's position ( distance to the runway ) or is it a time factor.

[spad54]

Can you set it up to fly only 3 default and the rest as static? If not let it rip as is. If ppl get snarley because its not perfect, well they don't have to fly it or use it.

You've done yoeman work on this project and I for one appreciate your time and effort. No matter how it comes out, it's better than having no ORA at all.

Let her go Mick... holding short awaiting clearance!

[mick]

Well, there doesn't seem to be much consistency about how far out, either in time or distance, the next approaching plane has to be before the plane taking off next gets stuck at the hold short point. Sometimes planes could take off when the next arrival was due in five or ten minutes, and other times they'd get stuck when the next arrival wasn't due for twenty minutes.

It seems to have more to do with how long I'd been at the airport and how many AI planes were present.

It might even be dependent on system resources. Those with more powerful confutors may find that it runs better for them than it does for me. I hope so!

Anyway, it's working fairly well and it seems to be the best I can do. I wish I could get it perfect but I guess that's not gonna happen.

I could leave most of the planes inactive and just have a few of them fly, but if getting stuck is the result of having so many planes present, it would still happen. Also, that would require rewriting the traffic files yet again, and I don't think I could bear to do that!

Not much remains to be done, though finding the time to do it is still something of a problem. I can only work on it a little at a time.

Stuart is going to give us a couple more special low-poly AI versions of his models, which are kind of frame rate intensive in their flyable form. But first he has to get his FSDS installation fixed. I hope that won't take long, but the planes will be worth waiting for.

While we wait for those, I have to write the ReadMe, which is a project in itself.

The only tinkering that remains on the actual project is to copy and save all the AI planes' textures in DXT format with mips. Some are in 32-bit format and many are without mips, which makes them load slowly and adds to the frame rate hit. I think it's worth the effort. This thing is a real frame rate killer, so anything I can do to reduce that will be worthwhile.

It should be ready soon. I hope to have my stuff finished by the time Stuart's AI models are ready. I can't predict a date, but I hope and expect we're looking at days rather than weeks.

[Roger]

Excellent news...thanks forthe update Mick!

[mick]

Well, the traffic files are finished. I've accepted the fact that I'm not going to get it to run like clockwork all the time and sometimes planes are going to get stuck. That keeps the flight activity from being perfect, but it doesn't keep it from being fun.

I hope to get the textures converted and the ReadMe done this weekend, real life permitting.

I still don't know how long it will be for Stuart's AI model files. That depends on how long it takes Abacus' tech support folks to get his FSDS working. Not very long, I hope. I'm going to try to have my part ready by the time his part is.

This has been much more demanding and time comsuming than I expected! When I said I'd zip up my own installation and share it I had no idea how incomplete my installation was!

But I'm glad I promised to share it, because real life has kept me so busy that the only reason I've gotten so much done on ORA is that I made the committment. Otherwise I'd just be putting it off and who knows if or when it would've been finished.

And of course, if it wasn't being prepared for public release we wouldn't be getting all the specially dedicated models and repaints and extras that so many folks have contributed for it.

Stay tuned...

[pooch]

Thanks for all your hard work. I will stay tuned...

[alain]

Mike.
To day, I would like to say just to which point I liked your work (Old Rhinebeck Aerodrome AI Traffic Package) and which I will make some observations later (give me where to send email). I observed twice the trafiic during several hours (2 and 3-4 hours saturday and sunday during summer) and taken notes on what I saw... As a whole I took much pleasure to see the AI planes taxiing, take off, landing and sometimes waiting until an other plane lands. I saw only two or three times a plane to disappear. I will make a summary minute after minute of what I saw.
But I was disappointed because on my 11 planes, only 6 or 7 actually moved (those on 1 to 6 parking but never the D25 or the planes on the 9, 11 or 12 places). For this time, and while waiting for the next realase you promise to us and that I will look with much interest, I would like just to ask you a question :
What do you mean when you write : "Flight Simulator 1954 - A Half Century of Flight" ?
Thank for this ad on... I am wainting for news !
Alain
P.S. please don't worry for my bad english

[mick]

Thanks Alain,

You can send me a private message here at The Old Hangar with your notes.

Those first traffic files had a basic flaw. They had so many take offs that once they were in the air, the planes didn't have a chance to land! That's why I trashed them and started from the beginning with an entirely new set of traffic files. Now there will be far fewer take offs, but the planes will be able to land.

With the original files on my computer all the planes would taxi out and take off. I watched them all to make sure of it. Perhaps it's just as well that some of your planes didn't take off. If they had, you never would've seen any landings! And as you must have seen, at that little airport with its steep approaches over the trees and its short runway, the landings can be interesting!

Anyway, those files are history. The new ones aren't perfect, but they're much better.

Since the new project will include the planes, with no assembly required, everything should work the same way for everyone this time.

Progress continues. I've converted all the textures to DXT format with mips, and I saw a definite improvement in my frame rates. I have started to gather all the components together for packaging and I've at least begun to think seriously about starting to assemble the documentation. I have dozens of e-mails to go through to remind myself about the documentation requirements of the many contributors. Little by little it's getting closer to release!

[mick]

Alain, I forgot to answer one of your questions.

"Flight Simulator 1954 - A Half Century of Flight" is my "stock" FS9 installation. It's mainly an AI scheme. I've replaced all the jetliners in the stock AI scheme with big propliners, all the turboprops with DC-3 and DC-4 variants, and all the modern GA planes with types that were active in the early and mid-1950s. In my local region, the northeastern US, I've also re-militarized all the former Air Force and Navy bases that are now either closed, joint use or all-civil airports. I remilitarized them by deleting all the civil flights from the stock traffic files and creating new traffic files using military planes from the late 1940s and early 1950s.

[mick]

WOO-HOO!!!!!! ;D

My part is DONE!

I finished the "Owner's Manual" last evening.

Now all we're waiting for is Stuart's low-poly AI models of the Caudron and SPAD. Those should be along as soon as Abacus tech support gets his FSDS working.

I hope that will be soon! The AI models will be worth waiting for because this package is a real frame rate killer, and it needs all the help it can get.