Quaternion returns NaN value

[kurogane1031]

Hi, firstable, thank you very much for your hardwork.

I am using xsens mti 700 imu. I tried to execute imu filter node from imu madgwick filter, however, everytime it return NaN value. I did tried using the sample bag file and the code returns the quaternions published value just fine. Is there any additional settings that I need to include to my imu?

[mintar]

Can you upload a bagfile somwhere that contains the raw data from the IMU, and the launch file that you're using to start the madgwick filter?

[kurogane1031]

Hi @mintar

The files that you have requested can be downloaded on the following link: the madgwick filter src folder: https://drive.google.com/file/d/1yMOuqZKhyBcqRlxdUPFKxZJ-pouNihv4/view?usp=sharing raw data of imu: https://drive.google.com/open?id=1ga6LNip31eRou205qvXs8fQK7VZx1Euu

I launch the madgwick filter using the following command: rosrun imu_filter_madgwick imu_filter_node _use_mag:=false

Thank you in advance for your help

[mintar]

The problem is the /imu/data_raw topic from your IMU. The linear acceleration is very close to the (0, 0, 0) vector, because your IMU compensates for gravity. However, the imu_filter_madgwick needs the gravity vector to compute the orientation.

You have two options:

1. Don't use imu_filter_madgwick. You don't need it, because your IMU already provides an orientation estimate in /imu/data_raw, and the internal filter is probably much better tuned to the IMU than the default values of imu_filter_madgwick.
2. Figure out how to modify the Xsens ROS driver so that it outputs an acceleration vector that includes gravity.

Also, it looks like you're using the indigo branch. I'd recommend using lunar or melodic instead if you're building from source, or otherwise just use the released binaries.

Anyway, thanks for your bug report! I've added a commit that prints a warning if this happens again: e4166a7fcf47d19c62a42f1ebbfc989cd1f1d8c6.