Interesting AI landing problem

[clickclickdoh]

I'm working on something for a project and have run into a snag that I can't figure through. I have an airfiled with a 5,000 ft runway with a 1,000 foot pattern altitude. Anything flying VFR to it enters the pattern, turns downwind then stops descending at 2,000 ft and flies away in a straight line never attempting to turn to the runway. Aircraft flying IFR to the runway do missed approaches. AI setup to start on the runway will happily do VFR TNGs.. but pattern at 2,000 ft.

I've tried swapping out differnt types of AI aircraft but get the same behavior.

Tried swapping out tower comms for unicom or CATF. No luck.

[Greg]

Did you try lowering the pattern altitude? Sometimes I needed to go as low as 500' to avoid this.

[clickclickdoh]

Lowering the pattern altitude interestingly also lowered the altitude at which the AI stops descending. Dropping the pattern altitude to 500 cause the descent to stop at 1500 feet. Pushing the pattern altitude up to 1500 feet causes the descent to stop at 2500 feet.

[clickclickdoh]

well, we were on the right track with the pattern altitude, just not extreme enough. When I put it down to 200 the planes started heading for the runway. So, that problem is solved.

Now I've got a new problem... because nothing can ever be easy right? The plane won't stop doing TnGs now.

Here's the segment for the TnG FP as generated by AIFP:

AC#1,9903,1%,4Hr,VFR,00:00:00,TNG00:30:00,050,F,0003,YYYY,00:45:00,01:00:00,035,F,0003,YYYY,01:15:00,01:38:19,050,F,0003,XXXX (airport codes changed to protect the innoccent)

In theory, it should take off from airport XXXX, fly to airport YYYY, do a few TnGs, land at YYYY, wait a few minutes on the ground then go back to XXXX.

What it is doing right now is take off for XXXX, fly to airport YYYY and then do TnGs until I cry myself to sleep. It never stops at YYYY and doesn't return to XXXX even after 2.5 hours of TnGs.

Both runways are inside the each others visual zone so the aircraft is in the sim the whole time. From a cold start on the ground at XXXX to first TnG landing at YYYY is 19.5 minutes, so leg time shouldn't be an issue.

[JohnTenn]

Click

Just theorising here.

Try using the ground loiter time at longer than 30 minutes.
I have heard that this makes them disappear.

This might just be what is precipitating the endless TNG.

John

[Firebird]

We found this quite some years ago whilst doing some C-130s.
There is a missing leg. You have got a TNG leg from XXXX to YYYY, then you need to add a leg from YYYY to YYYY to land. The second leg is best to start 3 mins after the end of the first leg to avoid problems in compiling.

The problem is that you have to have a leg following on from TNG legs start less than 5 mins later or you run the risk of getting what you see, the AI engine ignores the land leg.

The golden rule is that if you want aircraft to TNG at a place and then land that is two legs. If you want an aircraft to TNG at a place then land at a new place then only one leg is needed but you should still have the land leg flight start 3 mins after the end of the TNG leg.

Does this make sense, and does it help?

[JohnTenn]

Not to detract from what Steve stated.

I substituted LTBL and LTFA for XXXX and YYYY respectively.

Started the sim at 07:58 UTC. My aircraft flew from LTBL over LTFA. It then took a wide route back across LTBL and then commenced with the TNG at LTFA. By chance I witched it land, full stop, and taxi to a parking spot. It parked and then almost immediately started again. Taxied out and took off for LTBL.

Very short taxi at LTFA.

I expect that had the time required on the ground been longer then it would have gone into the endless cycle.

I think we now have two alternative solutions to this behaviour.

John