search

septentrio-gnss / septentrio_gnss_driver

ROS 1 & 2 driver for Septentrio GNSS & INS receivers
BSD 3-Clause "New" or "Revised" License
80 stars 38 forks source link

.nan in orientation filed of pose topic #133

Open mirons5678 opened 1 month ago

mirons5678 commented 1 month ago

Hi

I'm using ROS 2 (Humble on Ubuntu 22.04) with the septentrio_gnss_driver with mosaic-H and dual antenna setup. In the web gui I can see that a heading and pitch are being correctly reported. However the ROS "pose" topic contains only ".nan" for the "orientation" field.

Are you able to help please?

thomasemter commented 1 month ago

Hi @mirons5678 ,

thank you for your report. The calculation for the GNSS contained an error, since it can only determine either roll or pitch and the other is .nan. I have a quick fix here and the fix will be part of the next update. Because this is valid just for a heading offset of multiples of 90 °, pitch and roll have to be used with care. Only an INS is able to determine full orientation.

mirons5678 commented 1 month ago

Thanks for the quick reply @thomasemter. I was hoping to fuse heading from the GNSS into a robot localization EKF node, which I think requires a Pose message with orientation - I'm using a heading offset of 90°. Should your fix enable me to do this? Do you have any idea when the next update will be released?

thomasemter commented 1 month ago

Heading can be used for sure. If you are referring to robot_localization it is easily possible to exclude pitch and roll.

I hope the next update will be released very soon.

mirons5678 commented 1 month ago

Thanks for your support @thomasemter . With your updated branch I'm seeing orientation data in the Pose message. I'm fusing this into robot_localization and only including yaw from the GNSS pose topic. This is fused with x and y velocity from a wheel odometry topic. However, when the x and y velocity are 0, the yaw angle is continually rotating through 360° (even when the antenna are static and the web interface is showing an accurate heading). Do you have any idea why this might be?

I've also observed that setting att_offset: heading in the GNSS driver config file does not appear to be honoured (e.g. setting att_offset: heading: 90.0 shows as a heading offset of 0.0 in the web GUI).

thomasemter commented 1 month ago

Great to hear. I am sorry, but I am not very familiar with robot_localization since we use our own fusion framework at our institute.

In my setup, the heading in the WebUI reflects the parameter setting. image If you have set get_spatial_config_from_tf: true, the heading offset is drawn from tf and setting att_offset.heading has no effect. Also, if you have set use_ros_axis_orientation: true, the sign of the heading offset will be reversed since it follows a different convention.