AC_PrecLand; MAVLink backend supports LOCAL_FRD frame

[rmackay9]

This enhances the AC_PrecLand MAVLink backend so that it can accept LANDING_TARGET messages with the LOCAL_FRD frame (previously it only supported the BODY_FRD frame). This essentially turns off the roll/pitch compensation (but leaves yaw compensation in place) so that a vehicle's main camera gimbal can be used for precision landing.

The PrecisionLoiter autotest has been enhanced slightly to test sending messages using LOCAL_FRD

I've now tested this on a real vehicle running this BlueOS Extension

[Copilot]

Pull Request Overview

This PR updates the AC_PrecLand MAVLink backend to support the LOCAL_FRD frame for LANDING_TARGET messages, replacing previous Companion-related functionality. Key changes include:

• Renaming and macro updates from COMPANION to MAVLINK throughout the Precision Landing code.
• Modifications in los measurement handling to use the unified _los_meas structure with vector and timing updates.
• Updates to the SITL, MAVLink, and IRLock interfaces and corresponding autotest adjustments to allow selection between body-frame and local-frame target messages.

Reviewed Changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
libraries/AC_PrecLand/AC_PrecLand_config.h	Updated macro name from COMPANION to MAVLINK
libraries/AC_PrecLand/*	Replaced _los_meas_body with unified _los_meas usage
libraries/AC_PrecLand/AC_PrecLand_MAVLink.*	Added new MAVLink backend handling LOCAL_FRD frame
libraries/AC_PrecLand/AC_PrecLand_Backend.h	Modified API for los measurement retrieval
libraries/AC_PrecLand/AC_PrecLand.cpp	Updated los_meas retrieval/rotation and estimator logic
Tools/scripts/build_options.py	Adjusted build option feature from Companion to MAVLINK
Tools/autotest/arducopter.py	Extended precision loiter to select frame type for messages

Comments suppressed due to low confidence (3)

libraries/AC_PrecLand/AC_PrecLand_Backend.h:35

• The API has been updated from get_los_body to get_los_meas; please ensure that all backend implementations consistently use the unified _los_meas structure for representing los measurements.

        vec_norm = _los_meas.vec_norm;

libraries/AC_PrecLand/AC_PrecLand.cpp:629

• Verify that _yaw_align is provided in the expected unit (e.g., centi-degrees) since it is converted via cd_to_rad; consistent unit usage is critical for correct vector rotation.

            target_vec_norm.rotate_xy(cd_to_rad(_yaw_align));

Tools/autotest/arducopter.py:5349

• [nitpick] Consider renaming the parameter 'send_bf' to a more descriptive name like 'use_body_frame' to improve clarity on whether the landing target is sent in body-frame or local FRD frame.

    def precision_loiter_to_pos(self, x, y, z, send_bf = True, timeout=40):

[rishabsingh3003]

@rmackay9 thanks! Very requested feature.
Two small optional comments, not a blocker.

[rishabsingh3003]

We should probably have a _los_meas.valid = false here, to prevent stale usage of _los_meas.vec_unit incase this function returns early? We check the timestamp of the last consumed packet to make sure we don't consume duplicates, so this is probably not a problem yet, but my worry is that someone may use this in the future by mistake elsewhere.

[rishabsingh3003]

I think the other backends may need this as well

[rmackay9]

Hi @rishabsingh3003,

Thanks very much for the review, very few others are familiar with this library.

Re adding specifically setting "_los_meas.valid = false", I've just done a light review and I don't think this is necessary because the target's position will only be updated when "_los_meas.time_ms" is changed by a backend and each backend's update() function includes a timeout. Put another way, I don't think we need to invalidate the last known reading before the timeout does.

[rmackay9]

@rishabsingh3003,

Ah, I see your point re _distance_to_target though. You're absolutely right that the driver would consume that field even if the LANDING_TARGET was using an invalid frame. I've added a fix so that the field is only consumed when other fields are also consumed, thanks!