SDF <noise> model details?
I haven't been able to find much documentation on how the parameters specified in sensor <noise>
tags are used. Is there any documentation on how noise is implemented?
In particular, I'm curious what underlying process is used to model time-correlated noise?
Additionally, I was wondering if using quantized noise results in a quantized output measurement or if it is just the noise added that is quantized?
Asked by shonigmann on 2021-06-16 14:09:51 UTC
Answers
I am assuming you have seen this:
Sensor Noise Tutorial ? If not I would start there. Here is a brief explanation of the tags:
You could also look here: Noise Class Reference if you were planning on getting very detailed with it.
All that said, I haven't personally used this myself. I just figured I could get the obvious stuff out of the way if you hadn't discovered them yet. Based on your last two questions I am guessing you will need to make your own noise type.
Asked by Adam Gronewold on 2021-06-17 18:29:38 UTC
Comments
Thanks for the answer. I'm familiar with the stock Gaussian noise model in Gazebo, although it falls a bit flat when simulating sensors with time correlated noise (drift) like gyros. The SDF specification for IMU's offer a number of additional parameters to approach more realistic noise representations. But I haven't found any indication of their use in Gazbo; but I suppose not everything in SDF is necessarily in Gazebo. In any case, SetCustomNoiseCallback() seems like a potential workaround
Asked by shonigmann on 2021-06-22 13:32:53 UTC
If you are building your model via URDF, many of the SDF tags will never be parsed by Gazebo when you launch your simulation. If you are having trouble setting those tags, I would recommend you make a very basic plugin to fill the vital parameters. It is a pain, but it seems that is the way to go right now for many things in Gazebo.
Best of luck.
Asked by Adam Gronewold on 2021-07-02 18:37:09 UTC
Comments