Autopilot vs. Horizontal/LNAV Path

[Sabretooth78 56]

One observation I've made with the Pro (P3D4) version since its release is that the autopilot maneuvering of turns, at least in managed mode, seems to be a little off.  I never noticed this in the previous (FSX) version.  Generally, all appears good until the AP begins rolling out of the turn; it's as if it is aiming for a path that is parallel to the outbound path, but some distance before it.  Hopefully I'm describing it well enough.  Anyway, this has only been an observation, no big deal as the AP would eventually smoothly acquire the new path.  For all I know, maybe that's how the actual aircraft operates.

 

Since the most recent update, 1.2.1.0, however, this seems to have become a little more noticeable, in fact on my last flight I had a handful of turns (in the 40-60 degree range) where the turn completed with the aircraft somewhat short of the path and the AP corrected with a noticeable reverse bank to acquire the path - i.e. for a left turn (as most of them happened to be for this particular flight plan), it corrected with a short but sharp right bank to acquire the path.  I don't recall ever noticing this kind of behavior in the Aerosoft Airbus in any version prior, even on very sharp (acute) turns; nor on any other payware aircraft of this caliber.

 

I haven't noticed this in selected mode, but that's not to say it doesn't.  I'm suspecting since a selected turn is not trying to achieve a set path, such an issue won't occur as the AP will simply level out when it achieves the selected heading/track and isn't actually trying to acquire a defined path.

 

Here's a diagram (somewhat exaggerated) to try and explain it visually:

Magenta = Plan

Blue = Flown

Turn begins distance "x" short of waypoint "WAYPT"

At same distance "x" past WAYPT, a/c is offset from path by distance "w"

A/C acquires plan over distance "y" - smoothly (FSX, P3D4 prior to 1.2.1.0) or with right bank (1.2.1.0)

 

Please login to display this image.

 

 

[Aerosoft Team [Inactive Account] 51558]

Okay so taking speed into account the turn starts too early.  Thanks for the great explanation and we'll have a look at it. Determining where these turning points are is tricky and deeply rooted in our prediction code. By far the most complex bit of code we have.

[Sabretooth78 56]

In what seems to be a related issue, I happened into this situation on one of my recent flights:

 

This is on the OOSHN5 arrival into KBOS.  On two of the three times I've flown this (made a save at TOD), the autopilot went berserk on the tight LONEY-RODEE-EURRO turn.  During the initial flight, I quickly took over in selected mode.  The second time it did not happen (I was trying to figure out an ILS issue - which I reported here and resolved), however unlike the first time I was not running in selected speed (I'm not sure if that is important).  The third and final time, I was looking to ensure I had solved the ILS problem (which I had) and it went nuts - and this time only being a test, I decided to let it run its course and see what it would do.

 

Note that the managed speed is way too low (there's no reason based on the STAR or MCDU that it should be under the VNAV descent speed at this stage (say 280 kt), but I wasn't necessarily concerned in flying speeds to the tee.  I'm suspecting that is occurring as part of the acknowledged STAR/descent speed calculation troubles.  Given that getting this to happen seems to be somewhat dependent on managed vs. selected speeds, maybe the two are related and wreaking havoc on the autopilot prediction code.  But I digress - on to the pictures.

 

In the absence of video, I took several screen shots in quick progression:

 

After initially turning left per the LNAV path, things start to go off the rails just before RODEE - could it be because the sweeps from LONEY and EURRO were cutting RODEE "off" the path?

Please login to display this image.

 

Please login to display this image.

 

George finally wakes up and realizes he was heading the wrong way and now overcorrects:

Please login to display this image.

 

Please login to display this image.

 

Sanity returns:

Please login to display this image.

 

Spoke too soon!

Please login to display this image.

 

Please login to display this image.

 

Returning to normal:

Please login to display this image.

 

At this point, the Bus rejoined the path and flew the approach properly.

[Sabretooth78 56]

On 10/23/2018 at 9:13 PM, Sabretooth78 said:

In what seems to be a related issue, I happened into this situation on one of my recent flights:

 

This is on the OOSHN5 arrival into KBOS.  On two of the three times I've flown this (made a save at TOD), the autopilot went berserk on the tight LONEY-RODEE-EURRO turn.

 

I'm reporting the same problem for the FK 1D approach into Moscow, on the GISIN-BESTA-RUGEL-GEKLA turn.  The path ended up a little short on the turn to BESTA and then overcorrected with a sharp left bank just before RUGEL until finally turning back right (once roughly abeam RUGEL) to make GEKLA.  This was in the A319 CFM, 1.2.1.5.

[Sabretooth78 56]

Just to clarify on my last previous post, the particular airport with the problematic approach was UUDD Domodedovo.

[Sabretooth78 56]

Another approach with closely-spaced waypoints which exhibits this problem is the QWENN5 arrival into KSLC Salt Lake City.  Waypoints QWENN and FFU are only about 1 NM apart.  Upon reaching FFU, the a/c began to veer sharply to the right, when in fact it should have gently turned left.  I quickly switched to selected heading mode, so I didn't find out how exactly it would have recovered the situation.