3
$\begingroup$

I am trying to create an accurate IMU model in Gazebo based off of a real IMU that I am using.

The Gyro specs are:

  • Bias instability: 3.5 degrees/hr
  • Angular random walk: 0.1 degrees/sqrt(hr)

The Accel specs are:

  • Bias instability: 0.1 mg
  • Velocity random walk: 0.05 m/s/sqrt(hr)

I understand that I can simply enter in the IMU sensor noise values into the below fields to specify noise characteristics:

  • standard deviation
  • bias standard deviation
  • dynamic bias standard deviation
  • dynamic bias correlation time

Problem is, I have no idea how I can convert the bias instability and random walk values into these statistical parameters. Also, I don't have Noise Density values. Would it be absolutely necessary to generate those or are random walk and bias instability fine?

I've looked at Gazebo's example IMU model with noise parameters but the values obtained from the ADIS16448 didn't seem to match the spec sheet.

default values for use with ADIS16448 IMU Units for values from other picture

Thank you!

$\endgroup$
2
  • $\begingroup$ Perhaps there is a misunderstanding? Would Gyroscope noise density be equivalent to Angular random walk? $\endgroup$
    – zz uu
    Oct 1, 2023 at 1:57
  • $\begingroup$ Any update on this? Facing similar issues here.. $\endgroup$ Nov 6, 2023 at 10:54

1 Answer 1

1
$\begingroup$

Kalibr library has a good tutorial on how to calibrate an IMU with respect to a camera. It also has explanations of parameter data used in IMU calibrations.

From the kalibr wiki:

The parameters for the "white noise" processes are often specified in the datasheet of the sensor manufacturer. A bit misleading, they are commonly denoted as angular random walk in case of the gyro, and velocity random walk for the accel.

In contrast to the "white noise sigmas", and are rarely directly specified in the datasheet. The reason is that in practice, the bias does not truly behave like a "random walk" for longer integration times. Often, the so-called in-run bias (in)stability is specified instead.

Hope this helps, I am not very well versed on IMUs myself, this is just what I found researching.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.