Modified rne algorithm

[Hancheol Choi]

Sometimes, we need to computed Coriolis matrix separated from velocity term.

M(q)*qdd + C(q,qd)*qd + g(q) = tau

But, we can only get vector c(q,qd) ( = C(q,qd)*qd ) in rne algorithm.

Of course, we can use symbolic method (coriolis function) already implemented in Robotics Toolbox.

I found a paper solves this problem effectively.

Alessandro De Luca, 2009, A Modified Newton-Euler Method for Dynamic Computations in Robot Fault Detection and Control.

In this paper, two algorithms want to get Coriolis matrix separately; passivity-control (which motivated me to start this) and momentum observer (used for collision detection)

This algorithm is simple.

Normal rne needs 3 inputs, q, qd, qdd, but modified rne needs 4 inputs q, qd, qda, qdd.

Via this algorithm we compute like this.

M(q)*qdd + C(q,qd)*qda + g(q) = rne_modified(q, qd, qda, qdd)

I attached two files modifying original rne.

I already finished testing this algorithm with passivity-control in my own code.

But, I don't know how to integrate these code in robotics toolbox.

I met error like this.

'rne_modifed_dh' is not defined in 'SerialLink' type.

Should I include class methods somewhere?

If you give some tips about including new methods in SerialLink class, I can proceed to integrate modified rne methtod.

Thank you!

[Erik van Oene]

You are correct. The function should be correctly integrated in the SerialLink class before it is recognized as an method.

I would suggest to use the current RNE function as an example. It should be located in de @SerialLink folder where you should put you function as well.

Erik

[Hancheol Choi]

Thank you. I could finish test putting my code in SerialLink folder.

I made additional three files rne_modified, rne_modified_dh, rne_modified_mdh .

And test code is test_modified_rne_dh, test_modified_rne_mdh.

How can I contribute or pull-request this code to robotics toolbox?

2018년 9월 25일 화요일 오후 10시 11분 58초 UTC+9, Erik van Oene 님의 말:

[Peter Corke]

The coriolis() method gives you the matrix, not a vector.  Why can't you use that?

[최한철]

Of course, I can use coriolis() method, but it is computationally slow as you also mentioned in the comment of coriolis method.

It needs N^2/2 invocations of RNE. But modified one needs N+1 invocations of modified RNE.

And, of course, if we use coriolis() function only for simulation; no real-time consideration, it doesn't matter at all. We can just wait to be computed..

But I think most of control engineers are using this toolbox to apply their algorithm in the real-time application after simulation.

I just point out this coriolis() method could not be a good reference that can be used in the 1kHz real-time application.

2018년 10월 28일 (일) 오전 9:36, Peter Corke <peter....@gmail.com>님이 작성:

--
You received this message because you are subscribed to a topic in the Google Groups "Robotics & Machine Vision Toolboxes" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/robotics-tool-box/7J8aEsmBl3s/unsubscribe.
To unsubscribe from this group and all its topics, send an email to robotics-tool-...@googlegroups.com.
To post to this group, send email to robotics...@googlegroups.com.
Visit this group at https://groups.google.com/group/robotics-tool-box.
For more options, visit https://groups.google.com/d/optout.

[Peter Corke]

Getting MATLAB code to run at 1kHz is going to be challenging anyway, MATLAB Coder won't compile RTB classes.

I'm glad you got it working, but for now I don't think it's a high priority to include into the toolbox.

peter

To unsubscribe from this group and all its topics, send an email to robotics-tool-box+unsubscribe@googlegroups.com.
To post to this group, send email to robotics-tool-box@googlegroups.com.

[최한철]

Ok, i also agree with you. it's not a high priority.

Anyway, thank you for listening to my opinion~

2018년 10월 28일 (일) 오후 2:57, Peter Corke <peter....@gmail.com>님이 작성:

To unsubscribe from this group and all its topics, send an email to robotics-tool-...@googlegroups.com.
To post to this group, send email to robotics...@googlegroups.com.

Visit this group at https://groups.google.com/group/robotics-tool-box.
For more options, visit https://groups.google.com/d/optout.

--

You received this message because you are subscribed to a topic in the Google Groups "Robotics & Machine Vision Toolboxes" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/robotics-tool-box/7J8aEsmBl3s/unsubscribe.

To unsubscribe from this group and all its topics, send an email to robotics-tool-...@googlegroups.com.
To post to this group, send email to robotics...@googlegroups.com.