[Bug] The body angular velocity (body rate) values published on the ROS 2 topic /fmu/out/vehicle_angular_velocity differ significantly from the values shown in QGroundControl (QGC) via MAVLink messages.

[tuananh0606-tech]

Describe the bug

The body angular velocity (body rate) values published on the ROS 2 topic /fmu/out/vehicle_angular_velocity differ significantly from the values shown in QGroundControl (QGC) via MAVLink messages. This discrepancy causes confusion during controller debugging and system analysis.

To Reproduce

1. Run make px4_sitl gz_x500 in terminal PX4-Autopilot.
2. Run MicroXRCEAgent udp4 -p 8888 in other terminal.
3. Run ros2 topic echo /fmu/out/vehicle_angular_velocity in a terminal within the ros_ws that contains px4_msgs.
4. At the same time, monitor body angular velocity (e.g., gyro rates) in QGroundControl through the MAVLink Inspector.
5. Compare values between the ROS 2 topic and QGC – they show noticeable differences.

Expected behavior

The angular velocity (body rate) values from /fmu/out/vehicle_angular_velocity should closely match those shown in QGroundControl, as both should represent the same onboard state estimation.

Screenshot / Media

Flight Log

https://docs.px4.io/main/en/msg_docs/VehicleAngularVelocity.html

Software Version

PX4 version: v1.15.2

ROS 2: Humble

PX4-ROS2 Bridge: uXRCE v2.4.2

Simulation: Gazebo-harmonic

Flight controller

No response

Vehicle type

None

How are the different components wired up (including port information)

No response

Additional context

I am currently developing a custom controller that relies on accurate body rate feedback via ROS 2. The mismatch with QGC values raises concerns about which source is accurate, or whether a transformation (e.g., NED to FLU or timestamp alignment) is needed.

[beniaminopozzan]

Hi @tuananh0606-tech , The ATTITUDE MAVLink message uses exactly the vehicle_angular_velocity xyz filed to get rollspeed, pitchspeed and yawspeed.
The data does not match because you are seing messages captured at different timestamps.

Try matching the timestamps (you can turn off the time synchronized of uxrcedds so that the ROS 2 messages will keep the PX4 timestamp and make the matching agints the MAVLink timestamp easier), you will see the value to to match too (well, only if you get an exact time match).

[tuananh0606-tech]

@beniaminopozzan but i still get correct information about position, velocity and angle, only angular velocity is not correct

[beniaminopozzan]

@tuananh0606-tech correct as the data matches perfectly, not a single digit is off?
I might expect that from pos and vel as the publication rate is low enough that it is likely that you see the same messge but I won't expect this from attitude. Maybe it looks correct just because the data is close enough.
I would argue that even the angular velocity is correct as the difference between ROS2 and MAVLink is in the order of 1e-3 rad/s

[tuananh0606-tech]

@beniaminopozzan Do I need to convert to another coordinate frame?

[beniaminopozzan]

no conversion is needed if you compare raw MAVLink and urceDDS

Conversion is needed if you use MAVROS instead as it perform NED->ENU and FRD->FLU