Topic `global_position/local` not consistent linear z velocity in ENU frame

[sam98uele]

When filling mavros topic global_position/local and converting the linear velocity frame from NED to ENU the z coordinate is not consistent with the transformation, in fact, it remains positive toward down.

https://github.com/mavlink/mavros/blob/ros2/mavros/src/plugins/global_position.cpp#L343

    // Linear velocity
    tf2::toMsg(
      Eigen::Vector3d(gpos.vy, gpos.vx, gpos.vz) / 1E2,
      odom.twist.twist.linear);

to be consistent it should be:

    // Linear velocity
    tf2::toMsg(
      Eigen::Vector3d(gpos.vy, gpos.vx, -gpos.vz) / 1E2,
      odom.twist.twist.linear);

You can see it also from this plot:

[elgarbe]

Hi, is global_position/local just the ECEF coordinates reported by GPS? or is it GPS fused with other sensors?

[sam98uele]

global_position/local comes from Mavlink message: GLOBAL_POSITION_INT, which (from the documentation here) it is the GPS fused with the accelerometers.

From the documentation you will also see the input reference frame of the velocity.