Accelerometer / Gyroscope Biases and IMU Intrinsic Models

[sgraybi]

I'm curious as to why the accelerometer and gyroscope biases are not saved in the yaml file. The biases are computed as can be seen in the "estimated accelerometer bias (imu frame)" and "estimated gyroscope bias (imu frame)" plot show in the output pdf. To obtain a calibrated accelerometer measurement I would think you would need to compute acc_calibrated = M * acc_raw + b, where M is the 3x3 scale and orthogonality matrix in the yaml file and b is a 3x1 bias vector.

Also, maybe related, what is the A (3x3) matrix, under Gyroscope in the yaml file. The pdf suggests that it has units of (rad/s)/(m/s^2), angular velocity over linear acceleration.

Is there a document that goes into these details then please let me know. Thanks in advance!

stereo_imu-report-imucam.pdf
stereo_imu-imu.yaml.txt
stereo_imu-camchain-imucam.yaml.txt

[goldbattle]

The biases are time varying so each time the sensor turns one should expect it to be different.

For the IMU intrinsics, you could take a look at the original publication or a work from our group:

• https://timohinzmann.com/publications/icra_2016_rehder.pdf
• https://pgeneva.com/downloads/preprints/Yang2022ARXIV.pdf

A should be the gravity sensitivity parameter.

There isn't much documentation but it is on the todo list. See #520 and #189 (comment)