question: in imu_complementary_filter module, IMU's orientation has some ambiguous.

[asimay]

Hi, dear, I'm using imu_complementary_filtermodule, and combined with robot_localizationpackage. but find something strange of imu orientation output of imu_complementary_filtermodule.

I have an NED global frame imu, and I transform the body-NED data (gyro and accelmeter) to body-ENU, and feed the results into imu_filter module. but the data is not correct after drawing in excel.

my drawing is here: here,

I drive car from north -> west->south->east. the global data and orientation is right, but local odometry orientation data which integrated directly from imu_filter package is not correct, it seems like NWU frame. I wonder to know what is output of imu_complementary_filterorientation data's global frame based on? NWU frame? or do I need to do the frame transform from NWU to ENU? because robot_localizationpackage only accept ENU frame data of imu. Thank you!

[asimay]

or input data is body-ENU W.R.T NED frame, what is the output data frame ? body-ENU W.R.T NWU?

I see in package imu_filter_madgwick ....~world_frame (string, default: "nwu" "ned" "enu") how about the imu_complementary_filter package?

[asimay]

the algorithm in package will only output body-ENU W.R.T NWU data? no matter how i feed body-ENU W.R.T ENU data into package?

[mintar]

The imu_complementary_filter node indeed publishes NWU output. We have recently added the option of publishing ENU (or NED) output to imu_filter_madgwick (in #60), so I would propose that you use imu_filter_madgwick instead of imu_complementary_filter and set the world_frame param to 'enu'.

I have done a new release, but it hasn't arrived in the apt repository yet, so you need to build imu_filter_madgwick from source.

[asimay]

hi, @mintar , thank you very much! sorry for spamming info. and I want to know, if use imu_filter_madgwick, before hitting imu_filter_madgwick, i just only need to transform from body-NED W.R.T NED ->body-ENU WRT NED is ok? or just feed the NED data into it is ok?

[asimay]

and if i directly feed body-ENU WRT ENU data into imu_complementary_filter, could it output correct data which I want? I'm still puzzled... what is the input data frame requirement of imu_complementary_filter ? output format as you said can set world_frameto "enu". thank you!

[mintar]

I don't understand what you mean by body-NED W.R.T NED and so on. Both filters in this repo (imu_filter_madgwick and imu_complementary_filter) read gyro, accelerometer and magnetometer readings and output an orientation. As such, there is no "NED" or "ENU" in the input data, only in the output data.

I think you should do this:

• Orient your robot so that it is upright and faces magnetic east. You can check by doing rostopic echo on your magnetometer topic and checking that the magnetometer x = 0 and y > 0. Your base_link's x axis should now point east and y should point north.
• Start imu_filter_madgwick with the params use_mag = true, world_frame = 'enu' and publish_tf = true. Don't start any other TF publishers. Visualize the result in RViz (using odom as a fixed frame) and note down the orientation of the IMU.
• Kill everything. Load your robot's URDF (which should include the imu frame) and visualize it in RViz (using base_link as a fixed frame). The orientation should be exactly the same as the one you noted down earlier. If it isn't, you should update your URDF until the mounting position of the IMU is correct.

[mintar]

Closing this issue due to inactivity. Feel free to respond if necessary.