EuRoC Forum

Hi,
the coordinate system of IMU measurements seems rotated to us. Does it really follow the conventions described in the instructions document X-front, Y-left, Z-up?
It seems to us that X is up.

Second question - in the callback function do we return the position in the coordinate system of IMU?

Thanks.

Hi,

the alignment of the IMU is a bit different for task 1 and task 2. Only the alignment of the IMU w.r.t. the camera matters, which you can find in the contestant VM in the workspace in euroc_solution_t1/resource/calibration.yaml

maybe this discussion also helps:
http://euroc.iphpbb3.com/forum/20112706nx57588/challenge-3-f5/task-1-reference-frame-definition-t15.html

Hi,

The technical annex might be a bit misleading in this regard. With X-front, Y-left, Z-up, we simply mean the right-hand chirality of the coordinate frame. The initial orientation of the IMU in a world frame is to be estimated by you. As well known from visual-inertial state estimation literature, roll and pitch angle of the IMU is observable. You can set the yaw, which is unobservable, to an arbitrary value. As discussed in the link that Markus provided, we only evaluate the relative position and orientation. Therefore, you can set the initial pose to an arbitrary value - it does not even have to be gravity aligned, as long as your estimated trajectory is consistent with respect to your initial pose.