Coordinate Frame about direct_lidar_odometry HOT 3 CLOSED

Hi @danieldive -- thanks for your interest in our work. We use ROS convention (forward left up) as our coordinate system, as defined in REP 103 here: https://www.ros.org/reps/rep-0103.html#coordinate-frame-conventions

The output odometry is relative to the starting location, which is what we define as the origin. I may need more information on your specific setup, but it sounds like a transformation between the ROS convention and your desired coordinate system is needed, as you mentioned.

Let me know if that helps.

from direct_lidar_odometry.

Thank you @kennyjchen Sorry to resuscitate ! In which coordinate frame do you provide your FRD odometry to autopilot. I assume you've used px4 as a autopilot so did you enable external vision rotation to able to take off drone ? It seems that autopilot needs odometry in NED coordinate frame although px4 do FRD to NED transformation by choosing rotation external vision at the parameter section, do you have any other transformation in another module except in slam and autopilot software. I saw no drift your flight video but in reality especially when the slam algorithm exploring new places ( like when you pass an edge ) drift is beginning.

from direct_lidar_odometry.

Hi @danieldive -- we have a custom autopilot solution and use NWU, so we don't apply a transformation between odometry and our autopilot. It sounds like if another coordinate frame is needed, you can apply a transformation upstream of the autopilot. Alternatively, you could also try transforming the raw point cloud data (which is usually in ROS convention) into FRD to get the odometry into NED for your PX4.

from direct_lidar_odometry.