On Sunday 04 May 2008 15:11, Heiko Schulz wrote: > Hi, > > Some notes from me, because I'm working on traffic > files which affect the traffic in Germany (EDDF > etc...) > > I think Durk can explain a lot of things better than > me. > > --- Greg Hawkes <[EMAIL PROTECTED]> schrieb: > > The aeroplane was airborne at about one-third along > > the length of the > > runway. A real 737 needs most of the runway to > > reach its rotation point. > > Right, the perfomance is yet only given very rough and > no differenced between the different types of aircraft > (like 747 vs 737). But Durk told, that he and Thomas > Förster are on that.
Yes that's correct. In addition, the takeoff animation is currently still very rough, being based on a number of predetermined points on the runway. A few weeks ago, I have started refining some parts of the traffic system. I'll try to work my way through this from start to finish, tackling push back taxi procedures first, and ending by the final park at gates, adding interaction with ATC in the process. This will probably take a few years of spare time before that is all finished. ... > > > The landing gear doesn't retract. This is a minor > > quibble, but I noticed > > that the AI 737-300 aircraft model includes separate > > components for the > > landing gear. I was hoping that the AI aircraft > > could be animated to > > retract the landing gear after take-off. Likewise, > > the engine's turbines > > don't spin. Are there any plans to animate AI > > aircraft? > > Hmmm... I had the same question, but in the moment it > seems not to be possible. I wanted to add lights for > better visibility, but I had to recognize that there > is an issue with the property paths. Because of that > it isn't possible in the moment. If I did understand > Durk right, then it could be a bug reagarding property > paths in AI-Interactive Traffic. I've been looking at this a bit recently. It seems that the property nodes do remain valid, but due to it's dynamic nature, this is particularly hard to access these properties using an external interface. As said, I've seen something similar occuring while using the MIBS camera tool. I didn't have time to investigate yet, but I have some ideas as to what might be happening here. As a more intermediate term goal, I would very much like to add support for these types of animations, since these are intrinsic to proper aircraft operation, this could very well become an intrinsic part of the C++ code base, where it would be a lot easier to access the AI model objects directly. > > > YMML runway 27 is on heading 274 degrees. The AI > > aircraft maintained > > this bearing until about three minutes into the > > flight, at 9,000 feet. > > Real-life Melbourne-to-Sydney aeroplanes turn right > > before the landing > > gear has fully retracted. (Heck, I've been on > > flights where turn starts > > before the landing gear has left the tarmac! You get > > a really good view > > of some horses in the paddock outside the airport > > boundary.) Are > > FlightGear's AI pilots more conservative than > > real-life ones? > > The aeroplane climbed slowly to about 3,300 feet, > > then started to climb > > rapidly to about 9,000 feet. It turned to bearing > > 065, which is the > > route to Sydney, and continued to climb. I tried > > this several times, and > > once the AI aircraft performed a 210-degree /left/ > > turn instead of a > > 150-degree /right/ turn toward Sydney. Why would it > > do that? > > I think it is really difficult to program routes for > all 20.000 airports we have and some hundred thousand > waypoints too. Then in reallife the routes depends on > the wind too- another starting direction- another SID. Yes, in many cases, this is indeed a matter of choosing fairly arbitrary defaults. Eventually, I would like to make use of published departure / arrival procedures if possible. Recently, Thomas Forster mentioned already having some code to support that. Getting his code integrated with FlightGear CVS is currently one of my immediate term priorities. We'll probably discuss that at LinuxTag next month. > > > Has anyone ever watched an AI aircraft make a > > landing? I wanted to watch > > the AI aircraft on its flight to Sydney. I thought > > this would be easy > > with the FlightGear's "time warp" function. However, > > I discovered that > > AI aircraft are not subject to FlightGear's timewarp > > (effectively, the > > AI aircraft slows down when time speeds up). If time > > is running faster > > shouldn't the AI aircraft also travel faster in > > order to maintain their > > landing schedules? > > As I know, the time warp doesn't affect the AI > Interactive Traffic. Durk knows why. Actually, this is related to several factors. Strangely enough, the FlightGear time warp function only speeds up parts of the normal progression of time, and not others. The sun, moon, and celestial objects will progress faster, as will all internal time calculations. The Flight Dynamics Modules (FDMs) will not update faster, and therefore, one will continue to fly normally in a speeding world. Likewise, the AIModels code is also not sensitive to the time warping functions. The Traffic Manager itself is partly sensitive. > > > This means that I could not accelerate FlightGear to > > watch the AI > > aircraft land in Sydney. Instead, I started > > FlightGear with the command > > fgfs --airport=YSSY --aircraft=ufo > > --start-date-gmt=2008:05:03:03:55 > > > > (That is, at Sydney airport five minute before the > > flight landed.) I > > searched around with the UFO, and found the AI > > aircraft still at 30,000 > > feet at 0400 UTC, when it should have been about to > > land in Sydney. I > > followed it on heading 069 until 0413 hrs, by which > > time it was about > > 100 km out over the Pacific ocean. At that time, it > > made a 180-degree > > turn and went into a 35-degree dive, which would > > have been exciting for > > all concerned. It levelled-out at ~8,500 feet on > > heading 254. It held > > that course while it made a slow descent into > > Sydney. The flight > > touched-down (after flying through some ground > > scenery) at~0430hrs -- 30 > > minutes after its scheduled arrival time. It taxied > > to the end of the > > runway. > > Hmm... I think this is only in 0.9.10. In 1.0.0 and > CVS they landed with some minutes +/- on time. This is > what I noticed after creating some arrival flights for > EDDF. It really can depend a lot on the situation. The big problem here is the initialization. In FlightGear 0.9.10 and 1.0.0. A flight that is due to arrive will be created in cruise, at an approximate position according to schedule. Then, it has to run through the decent, approach and landing phases, each of which can add several minutes to the actual arrival time. The current situation isn't ideal and in need of a better implementation somewhere down the line. > > > I haven't created a ground network for Sydney, and > > because of > > Melbourne/Sydney rivalry I'm not going to :-) . The I have. :-) I have created ground networks for most airports that I have visited in real life. Since I was vacationing in Australia in 2003, I have build a ground network for sydney, about two years ago. It should be in FlightGear CVS / as well as in v1.0.0. Not sure about 0.9.10 though. The ground net I have for Sydney is fairly rudimentary and needs a touchup using the lasted taxidraw editing capabilities. though. > > aircraft taxied > > cross-country on a beeline to Sydney airport's > > origin, where it turned > > around a couple of times like a dog trying to lie > > down! Yes, the turning make one nauseas :-). This was finally fixes a few weeks ago. > > By now the time had passed the departure time for > > the return trip back > > to Melbourne. The AI aeroplane made a couple of > > consecutive > > push-back-and-forward moves (it never actually came > > to a full stop), > > then turned to heading 244 and started to taxi > > towards Melbourne. I > > stopped the simulation when I realised the plane was > > happy to taxi > > across Sydney harbour. > > > > Does that behaviour seem odd to anyone else? Yes, this is more or less the default behavior. When no ground network is present. the Airport's centerpoint is used as an emergency parking. From there, it basically starts taxiing straight to the runway, because it has no routing information. There is also the default push back that is still hard coded in version 0.9.10. > > Yes of course, but I it seems to me that it has > something to do with the groundnetwork. There are > airports without this bugs like turning around on the > taxiways. > > > I started FlightGear again, with the command: > > fgfs --airport=YSSY --aircraft=ufo > > --start-date-gmt=2008:05:03:04:25 > > > > (That is, Sydney airport five minutes before > > departing for the return > > trip.) The aircraft was on the ground at the > > airport's origin, as > > expected. Sure enough, at 04:30 it pushed-back, > > taxied to the runway and > > took off. By the way, while the 737 was at the start > > of the Sydney > > runway two light aircraft, about 100m apart, > > approached and landed > > /through/ it. This would be cause for some > > excitement in real life. Is > > ATC asleep in FlightGear? The two light aircraft are orignating from an older piece of AI code, which is not yet integrated with the traffc system. You can switch this off from the GUI using the AI menu (this has no effect on the Traffic code). Confusing? Yes, I know. :-) We have to clean this up. > > There isn't any ATC for AI InterActiveTraffic yet- so > much as I know, Durk is working on that. But they > react on your own aircraft on the taxiway. They stop > behind if you stop! Yes, that's correct. > > > I started FlightGear with the command > > fgfs --airport=YMML --aircraft=ufo > > --start-date-gmt=2008:05:03:05:55 > > > > (That is, Melbourne airport five minutes before the > > return trip landed.) > > I found the AI aircraft at 31,000 feet on heading > > 110. I followed it for > > another few minutes, until it went into another > > 35-degree dive from > > 31,000 to 8,000 feet. Who says a 737 can't do > > aerobatics? It descended > > very slowly from 8,000 feet, heading 274, until it > > touched-down at YMML > > at 0622 hrs -- 22 minutes after the scheduled > > arrival time. Also this piece of the code is scheduled for refinement. :-) > > > > Do the AI aircraft always arrive late at their > > destinations? Hey, we do aspire to be a *realistic* simulator. :-) Seriously, as mentioned above, the aircraft are placed at an approximate position according to schedule, and then it still has to go through a number of flight phases that causes a delay. > > Yeah, sometimes the movements of the aircrafts seems > very strange. > > > After the AI aircraft landed at YMML, FlightGear > > crashed with a > > segmentation fault and core dump. This may be > > because my very simple > > ground network has no nodes to taxi off the runway > > after landing. I will > > continue to investigate. Yes, that is very well possible. Since there is a groundnetwork, flightgear assumes it is possible to find a route from the runway back to the gate, which it doesn't find. Would have to verify to make sure. > > > > Regards, > > Greg Hawkes > > Yes, it is because there aren't any right nodes in the > network. Have look into other ones and how thy do > that. Notice: with the scenery 1.0.0 the airport > layouts has changed- some groundnetworks aren't right > now. > > And, at last: Durk is still working on AI Interactive > Traffic, so there of course a lot of bugs. But anyway > it is a lot of fun creating traffic files and spotting > aircrafts virtually in FlightGear! ;-) > Yes that is correct. The traffic system is very much still a work in progress. I probably would have progressed further if I hadn't wasted so much time doing virtual plane spotting.... Cheers, Durk ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Flightgear-users mailing list Flightgear-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-users
