quaternion from IMU interpreted incorrectly by ROS

Question

I have a new CH Robotics UM6 IMU that outputs Quaternion which the datasheet labels as [a,b,c,d]. When I pass the Quaternion to ROS the X and Z axes are swapped, also pointing the IMU UP points the quat DOWN in rviz. I thought Quaternions didn't have "handedness"?

Here is a video of the behavior.

Anyhow, how do I swap axes without converting to Euler angles and introducing singularities?

From the datasheet:

All UM6 angle measurements are made with respect to a North-East-Down (NED) inertial frame. The inertial frame x-axis is aligned with magnetic north, the y-axis is aligned with magnetic east, and the z-axis points down toward the center of the Earth. The "pitch" angle represents positive rotation about the x-axis, "roll" represents positive rotation about the y-axis, and "yaw" represents positive rotation about the z-axis.rotation about the z-axis.

PS I'm using python so numpy code would be great!

---

Originally posted by JeffRousseau on ROS Answers with karma: 1607 on 2011-04-06

Post score: 3

Answer 1

Your sensor is reporting data using the North-East-Down (NED) convention commonly used by aircraft. ROS messages assume the East-North-Up (ENU) coordinates more often used with ground robots. This should really be documented in REP 103, but is not explicitly mentioned. For a (long) discussion of ROS navigation satellite conventions, see the sensor_msgs GPS API review.

Multiplying by a constant quaternion should rotate your data from one frame to the other, but I don't have a good enough intuitive feel for quaternions to give you the exact multiplier. A little experimentation should suffice. Basically, you need to rotate Pi radians about the x-axis followed by -Pi/2 radians about the z-axis.

---

Originally posted by joq with karma: 25443 on 2011-04-06

This answer was ACCEPTED on the original site

Post score: 6

---

Original comments

Comment by JeffRousseau on 2011-04-07:
Also, if needed it turns out transformations.py has some nice functions that rotate quats: q_orig = data_imu; qx = quaternion_about_axis(math.pi, (1,0,0)); qy = quaternion_about_axis(math.pi/2, (0,1,0)); q_tmp = quaternion_multiply(q_orig, qy); q_rot = quaternion_multiply(q_tmp, qx)

Comment by JeffRousseau on 2011-04-07:
so it turns out it's a data misinterpretation problem! The UM6 outputs [abcd] which is actually [wxyz], so I had w and z swapped when passing the quat to ROS (I had assumed the scalar was at the end). So my solution is just w=a,x=b,y=c,z=-d (up/down swap)

Answer 2

Not sure if this is absolutely correct, but a quaternion just encodes rotation around an axis.

So you could just correcly swap axes, when filling the quaternion, e.g. when you have xyzw from your IMU fill the ros quaternion with: y, x, -z, w (might be wrong like this)

---

Originally posted by dornhege with karma: 31395 on 2011-04-06

This answer was NOT ACCEPTED on the original site

Post score: 1

---

Original comments

Comment by JeffRousseau on 2011-04-11:
yes, this does work. thanks