AI Aircraft SID and STAR Controller

[alpha117 73]

I'm sure there is, but will have to ask Roland. Not sure what the AI does if there are no STARs as in this example.

 

Never been to TUNIS, will have a more detailed look at the charts and let you know

 

 

Clive

Currently monitor AI traffic at EDDM, using EFB, it looks good. Get more delays and cancellations on departures and also diversions on arrivals. Looking good this end

 

Being monitoring for over 3hrs, how sad is that!!

 

BTW running AISISSTAR on a netwrok PC and using UT2. Having converted ALL German SIDSTARs and made changes to the .ini file.

[alpha117 73]

Just converted all DT*.* SIDSTAR files. Now sat at Tunis Carthage(DTTA). Nor much going on down here at the moment. Only one inbound, TUNISAIR TU375 from Algiers Houari Boumediene(DAAG)

 

Inbounds are 'vectored flights' by ATC onto the ILS approach. Looks like therefore these are not controlled by AISIDSTAR?? I'm sure Roland will confirm this.

 

Anyway , nice weather down here (32C), should have brought my shorts!!

[Clipper Ocean Spray 189]

For example, when AI aircraft arrive close to eachother at the edge of the "bubble", one seems to be put into a hold immediately, at the start of a very long star (plenty of time for speed differentiation). Instead of speeding the lead up and slowing the other down, the program seems to put one of them into a hold immediately, and this is compounded when there's lots of traffic.

 

 

Yes, that's correct.  For STAR entry point holding, the AI will get put on hold immediately if there is a conflict even if the STAR is lengthy.  Good point though, I calculate the STAR path distances, so perhaps I could factor that in before putting the AI on hold (e.g., let the first few through if the STAR is lengthy).  

 

 

Sometimes you see freshly spawned traffic that gets put into a hold due to another aircraft that is on the same STAR while 15+ miles away, so something doesn't seem right.

 

 

Several people have noticed this, but that is how it works by design.  The problem is I'm not dealing with AI arriving on one STAR, the AI are arriving on multiple STARs.  An AI approaching from the north could indeed be projected to conflict with an AI approaching from the south by the time they both line up for final.

 

I explain in more detail here at the start of the video:

 

http://www.mediafire.com/watch/t8nvlywnceoh67a/AISIDSTAR_Video_3_Operating_Parameters_Part_2.mp4

 

That being said, enough people have asked me about this that I've decided to add it as an option in the next version (even though in some cases it may compromise spacing on final).

 

 

The holding patterns flown by the AI seem really really large and spread out

 

 

Yes, they're doing standard turns with a 15 nm turn radius at typical airliner speed.  Increase the AIturnscalar and that should tighten up the turns across the board.  The AI are also changing speeds while holding to change minimum separation.  That decreases the coherency of the hold pattern, but it lets the AI take advantage of the circling distance to obtain minimum separation.  I keep going back and forth on that.  I'll probably add an undocumented option to turn off speed changes while holding.

 

 

Is there a way to increase the "Hold release" frequency. Ideally, lets say if 10 aircraft suddenly spawn at close to the same track distance on their respective STARs, the first few would continue at varying speeds to increase separation, while the remainder tightly orbit the entry waypoint

 

 

I think this relates back to your first two questions.  I'll think about letting the first few pass through the entry point if the STAR path distance is determined to be long enough to afford separation just by speed changes.  Re: turns, try bumping up the turn scalar and I might add an option to turn off speed changes while holdings.

 

 

Then, at intervals of 45 seconds (enough for lead aircraft at ~355 knots to go 4.5 nm for separation), you would release the aircraft in the race track from their holds, leading to a nice "string of pearls".

 

 

The program already does this part, but based on a decreasing interval when an excessive number of AI start holding until AIMaxHolding, at which point AI start getting released from holding without regard to separation.  This "clear out too many holding AI" phase will often produce decent separations.

 

A note of caution however regarding an apparent assumption you're making.  A strict 45 interval release would work if all of the AI were holding at one waypoint, but not (as would be typical at busy FSX airport) the AI are holding at different waypoints and on different STARs.  Each unique combination of holding waypoint and STAR has a unique distance to the final approach fix, so releasing the AI from hold using only a strict timed release would likely result in minimum separation violations as the AI approaching different directions reached the fix.

 

 

 

It seem to fix the bunch of planes landing behind one another at UK2000 EGLL, but now it seems that the AI planes are keeping on doing a flyby around the runway even though the runway is clear.

 

 

Chronic go-arounds is a sign the NoVectorAltitude and NoVectorDistance are too low in regard to a specific airport.  Try bumping one or both up a little.  I would start by bumping NoVectorDistance up to 8.

 

I explain it in more detail in the following video at the 5:20 mark:

 

http://www.mediafire.com/watch/t7bhbs0x2gnrekn/AISIDSTAR_Video_3_Operating_Parameters_Part_1.mp4

 

 

yes, that's the file. I had a look too and I noticed it had no STARs, but I thought that maybe there is a way to make changes in order to make AI fly the waypoints in there.

 

 

If there are no STARs specified for the airport, then AISIDSTAR will not provide vectoring services and the AI will fly approaches under the control of FSX per normal.

 

You can add custom procedures though, see the following video:

 

http://www.mediafire.com/watch/t9irz80c497beo7/AISIDSTAR_Video_2_Custom_Edit_SID_STAR_Files.mp4

 

I think you can also use the ADE editor to program nice approach (not STAR) patterns for any airport.

 

Regards,

-Roland

[semtex 0]

Hi Roland.

 

I have a question. I noticed that when AI is released from slewing and into the hands of FSX the aircraft almost always speeds up for a few seconds before returning to whatever speed it had. In a lot of cases this is extreme. I watched an A380 going from 220 knots to 530 knots and then back down to 220 again. Is this a FSX problem or is it something that can be fixed in some way?

 

The reason why I ask is I have added the transition fixes to the STARS at EGLL to try and funnel the traffic better onto the approach. This has actually worked well except the problem I described, it causes aircraft to sometimes catch up to the aircraft ahead of them that is on approach or in some cases actually miss the approach.

 

Regards

Pierre

[MeneMene 1]

Several people have noticed this, but that is how it works by design.  The problem is I'm not dealing with AI arriving on one STAR, the AI are arriving on multiple STARs.  An AI approaching from the north could indeed be projected to conflict with an AI approaching from the south by the time they both line up for final.

 

If you're already sorting by track distance, then at any one given time there is always a "lead" aircraft and a "trailing" aircraft, no matter where those aircraft are in physical space, correct? So lets say we have 10 aircraft spawned, at various different places and on different stars. The lead aircraft gets priority and is sped up to the maximum acceptable speed (within reason of course) until it is out of the STAR. Similarly, the second aircraft goes at slower or normal speed until it has the desired clearance with the lead aircraft, no matter where it may be, and then is sped up to match the lead's speed, etc etc for the third and fourth. For the last aircraft in the star, slow them down to maximize the rate of separation. You can figure out the distance needed for separation given the rate of separation in knots and the speed of the aircraft in front, and only then if you don't have enough space put aircraft into holds. Lets say 5 aircraft or so are in various holding patterns at star entry points. You know at any given time the track distance of the "last" aircraft on a star; pick the aircraft next in line to be released from a hold (either random or a queue of some sort), and just wait until the track distance of the last aircraft is the desired separation less than the track distance from the hold, and then release the aircraft from the hold (or maybe a bit earlier to account for the time required for the aircraft to leave the hold and line up with the star again). Then repeat for the next aircraft due to be released from a hold, using the recently released aircraft as the new "last aircraft". And lastly, after all this, you still have the star exit hold available to wait for space to land in the case that something gets messed up along the way.

 

Anyways, I'm sure you've already put a great deal of thought into this, those are just my two cents, I'm tremendously grateful for the work so far. This is just an issue for me because when an aircraft gets put into a hold, you in see in the console the traffic that it has a conflict with, and sometimes it seems that this aircraft is one that's on the same star and many miles ahead. Sorry that I don't have a log available, i'll try to find the time to sit and let it run for a bit.

[Bjoern 1,166]

Suggestion: Have aiconv write a log regarding which files need manual intervention and implement a "silent" mode. And extend the wildcard support.

 

 

I'm trying to convert the entire PMDG database (as I can't convert only airports starting with "C" and "K") and I'm basically sitting in front of an ever scrolling window (quite cool if you've got some techno running in the background) holding down a key because of the breaks whenever SIDs can't be deleted.

[masneoquil 66]

Hello there,

 

I have a little question about the lat version with the option for UT2.

-I have UT2. What I'm missing if I don't use that option?

- And the only way to use that option is to have AISIDSTAR working on a network configuration? Is there a technical reason for that?

 

Thanks again.

[Clipper Ocean Spray 189]

I noticed that when AI is released from slewing and into the hands of FSX the aircraft almost always speeds up for a few seconds before returning to whatever speed it had. In a lot of cases this is extreme. I watched an A380 going from 220 knots to 530 knots and then back down to 220 again. Is this a FSX problem or is it something that can be fixed in some way?

 

 

It is an FSX problem calibrating the AI's speed.  I calibrate too, but I have a pretty good idea of what the conversion values should be going in, so I don't have any dramatic speed changes when I take over from FSX.  Still, I'm looking at a new way to try to fix this problem.  If it works out, I'll mention it when the new version is released.

 

 

 

You know at any given time the track distance of the "last" aircraft on a star; pick the aircraft next in line to be released from a hold (either random or a queue of some sort), and just wait until the track distance of the last aircraft is the desired separation less than the track distance from the hold, and then release the aircraft from the hold (or maybe a bit earlier to account for the time required for the aircraft to leave the hold and line up with the star again).

 

 

I'm revising the code to release from hold directly using the sorted table along the lines you mentioned, just to see if it results in better performance.  If it works out well, it will be in the next version.  Thanks for the detailed feedback.

 

 

Have aiconv write a log regarding which files need manual intervention and implement a "silent" mode. And extend the wildcard support.

 

 

Good suggestion, I have received feedback from other users regarding this issue.  I will definitely implement mode that requires no manual intervention.  I'll probably interpret the use of a wildcard to be an implied "yes" to the various questions about airports not being in the database.  I'm not sure about extended wildcard support though, the operating system's file management capabilities allow you to quickly put a subset of desired files into a folder for conversion.

 

 

-I have UT2. What I'm missing if I don't use that option?

 

 

 

The main difference is departures. Using UT2 and AISIDSTAR without the UT2 compatibility option enabled, some AI will have a tendency to complete their SID, then do a 180 deg. and head back toward the airport before turning around once again toward their destination.  It happens at flight levels so you don't really notice it if you remain in the user aircraft.  However, if you're looking at a traffic map, it is more noticeable.  Similar deal with arrivals, but to a much lesser extent if you turn down the AImaxspawndistancefromstar to 60 or 50.  So, nothing really that earth shattering I guess.  With UT2 compatibility on, none of this should happen.

 

 

- And the only way to use that option is to have AISIDSTAR working on a network configuration? Is there a technical reason for that?

 

 

For some reason, I've found that if you run AISIDSTAR (with the UT2 option enabled) on the same FSX computer as UT2, then FSX will crash.  When I run AISIDSTAR (with the UT2 option enabled) on a networked computer, FSX doesn't crash.  It probably has something to do with Simconnect more fully segregating programs that come in via a network interface.  Frankly though, I'm not 100% sure it isn't something particular to my computer, so if you're up for it, please try to AISIDSTAR (with UT2 enabled) and UT2 on the same FSX computer and let me know what happens.

 

If you run AISIDSTAR with UT2 compatibility disabled (default), then it doesn't make a difference whether you run AIDISTAR on the FSX computer or a network client, FSX should operate normally. 

 

Thanks,

-Roland

[masneoquil 66]

 

 

For some reason, I've found that if you run AISIDSTAR (with the UT2 option enabled) on the same FSX computer as UT2, then FSX will crash.  When I run AISIDSTAR (with the UT2 option enabled) on a networked computer, FSX doesn't crash.  It probably has something to do with Simconnect more fully segregating programs that come in via a network interface.  Frankly though, I'm not 100% sure it isn't something particular to my computer, so if you're up for it, please try to AISIDSTAR (with UT2 enabled) and UT2 on the same FSX computer and let me know what happens.

 

Thanks Roland, I will try it and let you know!

[Bjoern 1,166]

I'm not sure about extended wildcard support though, the operating system's file management capabilities allow you to quickly put a subset of desired files into a folder for conversion.

 

Didn't think of that, haha!

 

I did the conversion and *then* weeded out the regions I don't want to use. The other way around would have saved me from 3249032493294 additional keystrokes. :lol:

[masneoquil 66]

Hi Roland,

 

Well, I tried with the option on, and I did have a crash. But is happened almost 30 minutes after I start AISIDSTAR, so I don't know if it was in fact your program.

The error was on util.dll, don't know if that means something to you.

I will try again anyway.

 

Thanks again.

[masneoquil 66]

I tried again but this time it CTD on approach. Again on util.dll.

Will try the same flight with out AISIDSTAR

[alpha117 73]

Phew..I thought it was just me, but I also get the crash with 'util.dll'

[jambo50 15]

Yup, also getting a 'util.dll' crash. I'm running networked with UT2 compatibility on and also the telemetry option on. Haven't had a chance to try and troubleshoot the error but if I do get a chance I will report back.

[MeneMene 1]

 

 

I'm revising the code to release from hold directly using the sorted table along the lines you mentioned, just to see if it results in better performance.  If it works out well, it will be in the next version.  Thanks for the detailed feedback.

 

THanks, I hope it works out. This also ties in with the holding pattern the traffic fly- I don't know how you've implemented the turning, but it would be nice if the AI could keep the current default turning rate when actually flying the star, while keeping a tighter, slower holding pattern so they are in the "right" physical space with regards to the separation calculations. Thanks for the response.