Computed Torque Control

[Muhammad Tufail]

Hi Peter,

I have some questions regarding computed torque control of your text (see page 213). I have tried to summarize my understanding of it in the attached document. My questions are: 

1) In the computed torque control law why are you multiplying the velocity gain with the desired acceleration? 

2) From your block diagram it seems OK then that Kv should not be multiplied with the qdd desired (starred) and just summed with the proportional and differential terms. 

3) But then again your block diagram doesn't seem to match with its Simulink implementation. There desired qdd is not coming from the desired trajectory but rather is calculated as a sum of proportional and differential error terms. 

Would you please clear the confusion?

My goal actually is to implement computed torque control in Cartesian space and then proceed to impedance control of PUMA 560.

thanks,

Tufail

[Peter]

Muhammad,

> 1) In the computed torque control law why are you multiplying the velocity
> gain with the desired acceleration?
> 2) From your block diagram it seems OK then that Kv should not
> be multiplied with the qdd desired (starred) and just summed with the
> proportional and differential terms.

this is indeed an error and has been previously pointed out, but I
neglected to post it to the errata page. I apologize for the
inconvenience, and the errata is now posted.

> 3) But then again your block diagram doesn't seem to match with its
> Simulink implementation. There desired qdd is not coming from the desired
> trajectory but rather is calculated as a sum of proportional and
> differential error terms.

The Simulink diagram should add qdd from the trajectory generator,
\ddot{q}^*, to the signal going into the qdd input of the RNE block.
That is, the required acceleration is due to the position error,
velocity error and desired (feedforward) acceleration, which is the
same as the third argument to the function D() in the equation at top
of p 213. This error is now posted on the errata page.

peter.

[Federica]

Hi Tufail,
Did you implement impedance control of PUMA 560 with Robotics Toolbox?
Federica

[Peter]

No, I've not done impedance control, but I believe all the tools are
there to do so.

peter

[Tufail]

Hi Federica,

First of all sorry for the late reply. 

I have not implemented impedance control of PUMA 560 arm. I had started working on it using the Robotics Toolbox but i didn't pursue it till completion. After having suggested Peter with modification in his computer torque control (as can be seen in previous post), i simulated them and surprisingly ended up with large error in joint variables. I really don't know why.

Thanks,

Tufail

[Alex Smith]

Hi, 

Has there been any more development with this? I am currently trying the same thing, so I will post results when I have them. 

Regards,

Alex

[Peter Corke]

With respect to computed torque or impedance control?

peter

--
You received this message because you are subscribed to the Google Groups "Robotics & Machine Vision Toolboxes" group.
To view this discussion on the web, visit https://groups.google.com/d/msg/robotics-tool-box/-/c9S7wnB5DQYJ.
To post to this group, send an email to robotics...@googlegroups.com.
To unsubscribe from this group, send email to robotics-tool-...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/robotics-tool-box?hl=en-GB.

[Alex Smith]

Both. I have it working now anyway... small position errors but that is expected. 

[Alex Smith]

Hi again,

I've been working on the computed torque control method from the book for a while now, but I have noticed a couple of things which seem strange... My model is based on fig. 9.20 on page 214 of the book.

1. Is there a particular reason for having a zero-order hold between the RNE block and the Robot block? My model seems to work fine without it. 

2. Why has the demanded trajectory acceleration been left out of the computed acceleration, whereas it is included in fig. 9.18? 

Regards,

Alex

[Peter Corke]

Alex,

thanks for your comments and careful reading.

1. Is there a particular reason for having a zero-order hold between the RNE block and the Robot block? My model seems to work fine without it. 

It will work fine without it.  What I didn't explain clearly enough is that the robot controller is typically a discrete time system with a specific sample rate, say 500Hz.  Since all the other blocks in this model are continuous the ZOH was introduced to make the system discrete time.

The same goes for the feedforward controller of Fig 9.19 except there we can have the feedback and feedforward loops running at different rates.

2. Why has the demanded trajectory acceleration been left out of the computed acceleration, whereas it is included in fig. 9.18? 

This is an error, it'll be fixed in the second printing.

peter

[Peter Corke]

2. Why has the demanded trajectory acceleration been left out of the computed acceleration, whereas it is included in fig. 9.18? 

This is an error, it'll be fixed in the second printing.

this has been mentioned in the errata for a while now.

[Alex Smith]

I'm afraid I can't share my model, as it will be used in a published paper for a conference. 

Regards,

Alex

On Thursday, 22 November 2012 00:52:21 UTC, angper...@gmail.com wrote:

I'm trying to do the same as u, impedance control with inner motion to minimize the error. But I'm a bit clumsy with matlab. Could u share your model? . I started with the dynamical model in joint angles which Peter Corke provides.

[hushein bai]

hi sir...

now i have a problem with "Quaternion". cant execute properly...
--
With Warm Reg,
Hushein.R

Mobile No:9944528538

[Peter Corke]

 I don't have a problem with Quaternion.