RTL_ALT_FINAL is relative to home, not terrain, causing hard landing at rally point

[pa-hagen]

Bug report

Issue details
RTL_ALT_FINAL is calculated relative to HOME, even if RTL_TYPE is terrain and TERRAIN is enabled.
This causes the drone to descend all the way to ground at LAND_SPEED_HIGH if RTL target (RALLY point) is located higher in the terrain relative to HOME and RTL_ALT_FINAL > 0

Version
Arducopter 4.5.2 in IRL. Verifired in Arducopter 4.7 in SITL.

Platform
[ ] All
[ ] AntennaTracker
[x] Copter
[ ] Plane
[ ] Rover
[ ] Submarine

Airframe type
Hex

Hardware type
Cube orange

Logs
https://www.pahagen.com/ardupilot/00199-2026-01-15_10-34-13.tlog
(RTL start at approx 95% to end. Mode changed to break approx 5 meters above ground to prevent hard landing).

[lthall]

Please provide a dataflash log.

[peterbarker]

Ping @pa-hagen

[pa-hagen]

@peterbarker @lthall

2026-01-15 10-34-13.zip

[lthall]

Thanks @pa-hagen !

Looks like we set a decent target here:

ardupilot/ArduCopter/mode_rtl.cpp

Line 403 in b720755

rtl_path.descent_target = Location(rtl_path.return_target.lat, rtl_path.return_target.lng, g.rtl_alt_final, Location::AltFrame::ABOVE_HOME);

We also appear to be going to a rally point rather than home in this log. However, I am not sure why we do the decent phase at all.

[pa-hagen]

We also appear to be going to a rally point rather than home in this log. However, I am not sure why we do the decent phase at all.

@lthall Yes, we are going to a rally point, and the rally point is located higher (16 meters i believe) relative to home. Going to a rally point with a different AMSL than home seems to be the only situation where this is a problem.

"LAND" seems to work fine and will slow down at expected elevation above terrain (even in higher terrain) based on terrain database.

[lthall]

Ok, I think I have worked it out. You are also using RTL_ALT_FINAL = 10 m. However this is referenced to home. So because you are going to a rally point that is 16 m above home the target altitude is below the terrain.

I will discuss this with @peterbarker and @rmackay9 .