navigation user guide

[Frank Dellaert]

All,

We have embarked on the GTSAM documentation project this semester. I am getting a lot of help from 2 smart undergrads, Porter Zach and Katie Kim, to do so.

The first more or less complete section is the Navigation user guide. Many of the source notebooks were generated with the help of Gemini 2.5 - which is *amazing* - after which I reviewed and edited them, and made sure things worked in colab.

Please consider helping by flagging any problems/hallucinations that remain, and also let me know if there are ways that this could be even more helpful. Note this is not the API docs, we are working on revamping those in parallel.

Finally, last but not least, Yashas Aambati has been working on Python packaging. You can now do “pip install gtsam-develop” on Linux, Mac version coming soon.

Best

Frank

[Valerio Spagnoli]

Hi! I was looking at the Navigation User Guide, which is actually very useful!

However, I have a question regarding the accelerometerCovariance parameter. It's described as a "3x3 matrix representing the continuous-time noise covariance of the accelerometer," with units specified as (m/s²)²/Hz. This implies that the standard deviation units are (m/s²)/sqrt(Hz), aligning with the units for accelerometer noise density.

According to VectorNav's documentation, there's a distinction between standard deviation and noise density: the standard deviation σ is calculated from the noise density (ND) and the sample rate (SR) using the formula σ = ND * sqrt(SR). This results in units of (m/s²) for the standard deviation.

My question is: which value should be used for implementation? Should I directly use the noise density, or is it more appropriate to calculate the standard deviation following the VectorNav documentation? This question comes out also from the fact that in this page it is specified that "the units for stddev are σ = m/s²/√Hz.".

Could you please clarify which approach is correct for setting the accelerometer covariance in the implementation? The same holds also for the gyroscope.

Thank you for your time.

Then a feedback about the user experience in navigating the documentation. It would be useful to include direct links to the PreintegrationParams and PreintegrationCombinedParams pages within the IMUFactor page or in the main navigation page. Currently, these pages can only be accessed via the sidebar, which may be somewhat confusing for users. Direct links would streamline the navigation process and enhance the user experience comprehension, but this is just my personal observation :)

[Dellaert, Frank]

Properly documenting the noise models for the IMU is actually a bit of a project in and of itself, but I agree it should be done. Fan is defending soon and he might have some time to convert some of his latex documents into notebooks that we can link to. It would also be good to have some example pipelines that illustrate the difference between IMUfactor and combined IMUfactor.

Frank

---

From: gtsam...@googlegroups.com <gtsam...@googlegroups.com> on behalf of Valerio Spagnoli <spag...@arpsoft.it>
Sent: Thursday, April 10, 2025 6:29:10 AM
To: gtsam users <gtsam...@googlegroups.com>
Subject: [GTSAM] Re: navigation user guide

 

You don't often get email from spag...@arpsoft.it. Learn why this is important

--
You received this message because you are subscribed to the Google Groups "gtsam users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gtsam-users...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/gtsam-users/7601efa6-41ba-4aca-9f29-dd168c62f3d9n%40googlegroups.com.

[Jiang, Fan]

I do have some documents with detailed documentation of IMU noise models. But I am too busy now for my defense. Will have more time for organizing and publishing them soon.

Best,

Fan

Fan Jiang

5th year Robotics PhD
Georgia Institute of Technology

E-mail: fan....@gatech.edu

Try TeXmacs today!

To view this discussion visit https://groups.google.com/d/msgid/gtsam-users/SJ0PR07MB90744D383DE71D30CA795FB2CFB72%40SJ0PR07MB9074.namprd07.prod.outlook.com.

[Valerio Spagnoli]

Thank you! Good luck for your defense! ;-)