Hi,

 

I've been testing out the software today with unfortunately not with a great deal of success. I really like the idea so hopefully any bugs etc can be ironed out eventually.

 

I've been basing my tests around EGLL and only EGLL. I did an aiconv from the latest PMDG sids and stars for EGLL and made no other changes to the files. After starting up FSX it took something like 40mins before I had an arrival with the aircraft seemingly being vectored to STARs on the other side of Heathrow. For example aircraft close to LOGON where being told to direct to OCK2C which made little sense when a LAM3A would have been far better. Correct me if I am wrong but I thing pretty much all STARs are available to use no matter what runway is in use at EGLL. I changed the option AIautostar to AIautostar=0 in the hope this may help but the program does not honor that change, it stays at 1 as shown when the program loads regardless of what you set that to in the ini file. Another issue I came across is when I did actually have some arrivals the first one would land and then any other aircraft afterwards would get to say 50ft above the ground before climbing away and being vectored to a SID. The runway was clear all times.

 

I was trying to figure out why it too so long for the arrivals to start coming in but there didn't really appear to be any pattern. The aircraft were just going all over the place at odd heights must of which were far too higher for an approach to work.

 

The most I had was two aircraft on approach at the same time with the next aircraft coming in probably about 15 mins later.

 

I'm happy to produce a logfile for you should this help. In fact I'll kick it all off again now and start logging for the next hour.

 

Thanks for having a go and trying to crack this.

 

Best

 

Andy

Andy,

 

The latest beta version (which hopefully I can get out soon) addresses many of the issues you described.  Arrival flow should be a lot better (including when holding is enabled).  I've gotten the most feedback on this issue.  Another problem I received a lot of feedback on was the program crashing after about an hour.  This was caused by AI turning around at airports (arriving, sitting at the gate, and departing) and by AI leaving the FSX reality bubble and then coming back, which has all been fixed.

 

The STAR selection takes into account the arc length of standard turn (going from memory the radius is about 15 nm), so an entry waypoint that appears close (e.g., 5 nm away), but behind, is actually farther away than an entry waypoint 10 nm directly ahead (in the view of the program).  Having said that, there was a bug in the arc length calculation that was imposing way too much of a penalty for turns, so I'm not surprised unfortunately to hear about the AI being vectored from one side of Heathrow to the other.  This has been fixed.  I have more than bug fixes though.  One important new feature is SIDs are selected based on the AI's destination airport.

 

Regarding the AI go-around for no apparent reason, I had the "no-vector" zone around the airport too low.  FSX wants about 7 nm (to be safe), I was at 5 nm.   When this zone is violated, FSX sometimes gets confused and starts clearing AI that are second in line to land, but of course that AI might go around, which then causes every single AI behind that AI to go around as well regardless of runway obstacles.  This has been fixed - it is now 7 nm.

 

Regarding odd heights and approaches too high, I've redone the holding altitudes to reflect the AI's closeness to the airport.  Even when holding is disabled, the new version will detect if an AI is too far above its descent profile and order a circling descent.  The main reason the AI comes in too high sometimes is because it is spawned too high in relation to the closest STAR entry point.  There was also a bug in the AIConv.exe utility that was not deleting all waypoints too close to the airport, which can cause problems when AIDSIDSTAR hands off to FSX for the final approach (STAR) (partially for the reasons discussed above wrt to the go-around issue) or when FSX hands off to AISIDTAR for the departure (SID).  This has been fixed as well, but the files will have to be run through AIConv again.

 

I did not know about AIautostor=0 being ignored.  I will look into this and get it fixed for the new version (probably an easy fix once I find it).

 

Thanks for taking the time to test out the beta and then write up your observations.  Much appreciated.  Please give the new version a try and let me know what differences you observed.

 

-Roland

 

 

 

 

 

 

 

Hey Roland,

Got FSX back in but now trying to reconfigure and rebuild my AI. Doesn't help that there was a new Airbus model fleet released. Keeps me busy.

 

Hope to test again soon.

Hi Roland,

 

Thanks for the reply and for continuing to develop the software. It sounds like you have seen and fixed all the probs I was experiencing so as and when you get the new release out I'll certainly give it another go and let you know the outcome. I like the idea of the assigned SID taking into consideration the destination airport, I had noticed that, nice fix!

 

Any ideas when you can put the new release out? No pressure just want to know when to come back here and test that's all.

 

All the best

 

Andy

There is no FSX data converter to implement because FSX doesn't come with any built in SID or STAR data only approach data and this program has no use for that data as it doesn't control the planes on the approach stage only on the SID or STAR stage. if you want to take full advantage of the software you will need to buy one cycle of AIRAC data.

BTW Clipper Ocean Spray, why are you using the PMDG data and not for example the Captain Sim data? In my experience having written an AIRAC approach data to FSX XML approach data converter the Captain Sim data is easier to parse since it is similar in format to CSV files while the PMDG format is actually written in a pseudo human language (probably because when it was invented PMDG wanted it to be easy for people to update and write new SID's and STAR's) which is harder to parse correctly.

Rev 1.2J (beta) can be found here:

http://www.mediafire.com/download/r9dajzm2ceq5es6/AISIDSTAR12J.zip

NEW: automatic SID selection now takes into account where the AI is going (in addition to the distance/direction between the AI at takeoff and the SID entry point). In most cases, this will result in the automatic selection of a SID appropriate for the AI’s departing runway and destination airport. You can see a good example of the effect at KLAX, where I added the ZILLI and FIXIT SIDs. AI destined to cross the Pacific are assigned ZILLI or FIXIT while AI destined points east are assigned LOOP6.

I think this new feature (along with automatic STAR selection based on runways in use) provides versatile SID/STAR selection.

NEW: SIDs/STARs only for propeller-equipped aircraft or jets. Just change the pattern name field inside each SID and STAR file to begin with the four letters "prop" (case not important) and that SID/STAR will only get assigned to propreller equipped AI (e.g., Prop KBOS STAR ORW.ORW6.TXT). Similarly, begin the pattern name field with "jet" for turbo-jet only SID/STARs.

NEW: Circling descents (related to holding, but different). If the AI is judged to be too high above the descent profile, it will be commanded to descend while circling the exit waypoint until it acheives a better altitude. This situation often happens when a FSX situations is first started and quite a few AI are spawned close to the airport at too high of an altitude for the actual STARs. This feature should also reduce instances of divebombing on final approach, which happens I might add under the control of FSX (not AISIDSTAR).  This feature will always be on.

IMPROVED: Holding Flow (if enabled). Received a lot of feedback about how few AI were allowed to land when holding is enabled! My prior focus re: holding was to prevent go-arounds, but I've decided to loosen the reigns. If too many AI are placed on hold, the program will start forcing their release at diminishing time intervals (sliding scale). If there are more than 25 AI holding at anytime, the AI will be released immediately from holds (provided they have obtained the commanded holding altitude first) until the queues are cleared. This will compromise spacing and increase go-arounds, but the number of go-arounds should still be less than if no spacing was enforced.

IMPROVED: Better holding altitudes. If the AI is assigned to hold at an entry or exit waypoint and a waypoint altitude restriction is not available, the AI will be assigned a holding altitude calculated to be reasonable in view of its "closeness" to the destination airport. Moroever, the AI will not be released from hold until it acheives this atltiude during. This should reduce instances of divebombing on final approach.

IMPROVED: CPU Occupancy - should now average about 0% after initial startup (code optimizations). Still, it is best to run AISIDSTAR on a networked, SimConnect client.

IMPROVED: Memory management. The program gradually increases memory size over time, but this is not a leak. Rather, the data to track (AI tables and related information) tends to increase over time. I've added more aggressive pruning of these tables, so that a memory steady-state value should be reached about an hour after the program is started.

IMPROVED: Regular status messages (non-verbose) in console window are now much more informative and should give you a better picture of what is happening in the monitored area while still being fairly succinct.

FIXED: Program crashes due to AI "turn-around" scenarios. I recently revised the code to handle "turn-around" situations, but specific situations really test the code! One interesting bug was caused by my assumption that the same AI aircraft would only get "turned around" at an airport in the reality bubble, thus if I detected it flying again, I should assign a SID. I discovered however that every once in awhile an AI object that leaves the FSX reality bubble will COME BACK some hours later destined for a monitored airport! Obviously assigning a SID in that situation would not be appropriate and indeed caused a vector out-of-range error that crashed the program.

FIXED: The conversion utility was not deleting all waypoints determined to be too close to monitored airport (e.g., default 2 nm for SIDs, 5 nm for STARs). The samples included with version 1.2J have been re-converted and should be OK, but if you have your own files that you converted using an earlier version, you may want to consider converting them again using the latest AIConv.exe.

FIXED: Program was attempting to vector AI within 7 nm of the destination airport, which sometimes causes chronic go-arounds with AI.

FIXED: STAR/SID selection was assigning too much of a penality for the turning distance required to hit the entry waypoint, causing inappropriate STAR and SID selection. Please bear in mind however that even with this fix, it is still extremely unlikely the AI cruising at flight level will choose an entry point that is behind it. However, waypoints at bearings of +/- 90 deg. are more likely to be chosen after the fix.

-Roland

Hello Roland,

 

I'm looking forward to test you software, but I have an error when I try to download the file.

It says that the files has been removed , renamed o temporary unavailable.

 

Thanks!!

Hi,

 

Download working ok for me.

 

Andy

mmm I think the problem it's on my side!!

 

Thanks

BTW Clipper Ocean Spray, why are you using the PMDG data and not for example the Captain Sim data? In my experience having written an AIRAC approach data to FSX XML approach data converter the Captain Sim data is easier to parse since it is similar in format to CSV files while the PMDG format is actually written in a pseudo human language (probably because when it was invented PMDG wanted it to be easy for people to update and write new SID's and STAR's) which is harder to parse correctly.

 

Seemed like a format the most users subscribed to, but I agree, the format is easy to read, but hard to parse!  BTW I forgot to mention in the write-up the issue you identified has also been fixed ("escape" key should only be recognized when console window is in focus).

can someone post a tutorial on how to create SID/STARS for the airports from the Navdata pro db ?

Any chance Roland you could maybe put together a video displaying the program working at an airport like LHR,LAX,JFK?

Any chance Roland you could maybe put together a video displaying the program working at an airport like LHR,LAX,JFK?

 

+1

 

Dirk.

Hi Roland,

 

I've been testing the new release. It now picks up the correct STAR and doesn't vector the aircraft over to the other side of Heathrow which is great. I'm still having probs though with only the occasionally arrival, I've had 1 in the last 30 mins although there is a build up of aircraft now trying to get to land they just never manage to make it. Here are my findings.

 

Holding set to 1

Minimum sep ranges from 0,1-4.

 

- I've noticed aircraft that are miles apart (I'm talking at least 30nm here) are being turned around and put back into the hold due to minimum separation being reached. See screenshot 1 for an example. This happens constantly.

- Your program takes control again of the aircraft released to FSX ATC for final approach. They get directed back to the STAR entry point and the whole process starts again. The reason they are taken control of again is due to minimum separation. I didn't think once they are released that your program would take control again. Even with this set to off i.e. 0.0 it still happens when holding is enabled. See screenshot 2.

 

Holding set to 0

Minimum sep set to 0

 

- Some aircraft after being released for approach close to the airfield turn around and head off in the wrong direction for miles. They don't appear to be controlled by your program going by the output but this does seem very odd. Could be FSX ATC of course. See my comment about the aircraft height below, this may explain it.

- Certainly more approaches but many go arounds due to no separation.

 

Holding set to 0

Minimum sep ranges from 0,1-4.

 

- Seems better, still very few approaches but the best I have had it so far.

 

More often than not with any settings aircraft are being released to FSX ATC far too high. I think this might be why they get vectored away from the airfield again before finaly trying to land. I just noticed and aircraft being released at FL100, after release it climbed back to FL180 and headed off away from the airfield. Is there a minimum height that FSX ATC wants the aircraft at before it's ready to vector them onto the approach?

 

Some ideas for enhancement. Whilst it's nice to see the text scrolling through the window it's not always easy to follow what's going on. One thing that would really enhance your program is to change your scrolling list into an auto updating table. The table could show for example...

 

flight number

aircraft type

departure airfield

arrival airfield

assigned sid/star

status i.e. holding, on approach, going around, next waypoint etc.

 

This table would be great asset to the program. Just a thought anyway.

 

Hope the above helps.

 

All the best

 

Andy

 

Any chance Roland you could maybe put together a video displaying the program working at an airport like LHR,LAX,JFK?

 

I'll look into it.  Maybe JFK and its interesting climb procedures.  With the improved SID selection, I could probably get the AI to do all the different climbs (e.g., Bridge, Canarsie, Gateway, etc.) although it may require some custom waypoints here and there.