LuGre friction model implementation

[ale-git]

The LuGre friction model has been implemented in SImulink, ready for automatic code generation.
The output code is included in icub-firmware to add sensorless torque estimation capabilities to the control boards.

[pattacini]

Thanks @ale-git for the contribution 👍🏻
I'll be reviewing the PR in the next few days.

[pattacini]

Hi @ale-git

Before merging this PR, we need to address several points.

Documentation

• The state equations is not correct.
• Missing details on $g\left(v\right)$.
• It'd be appropriate to cite a paper on LuGre, e.g. DOI: 10.1109/mcs.2008.929425.
• Nothing is said about how to tune the LuGre params. This is fundamental to make it usable.

Implementation

• Especially when a dynamic is at stake, it's much better to resort to Simulink rather than MATLAB code. We already have an internal LuGre implementation in Simulink that I can refer you to¹.
• Most of the time $\sigma_1$ is related to $\sigma_0$, as illustrated in the paper I cited above.
• The model interface can be improved.

MATLAB Project

The project has to be fine-tuned, although this is the very last thing.

---

I propose to set up a meeting for this 🗓️

Footnotes

1. It looks like this in Physical signals but can be ported back to original Simulink signals very easily:

   ↩

[ale-git]

I committed the wrong README.md file, whose right version was deleted instead (fixed).
I also added a couple figures.

[pattacini]

Any update on this?

[ale-git]

I've added the Simulink-block made torque estimator to the model.

[pattacini]

Great!
There were other things to address we discussed last week.
I'll give the PR the last touch when the work is complete.

[pattacini]

Hi @ale-git

We still need to address the following important point:

Nothing is said about how to tune the LuGre params. This is fundamental to make it usable.

[pattacini]

Hi @ale-git

Thanks for the notes on how to tune the params.
I've dropped some inline suggestions you may incorporate straight away.

A couple of important points though:

• Nothing is said about the math underlying the brute force estimation. I think that we ought to outline the principle beneath at least, in a brief paragraph.
• I would report the measurement units along with the values.

[ale-git]

I've made the requested changes

[pattacini]

Thanks!