are there some papers concerning the combination of joint limit avoidance and avoidance of singular configurations available ( of course, during the same trajectory)?
The Orocos Project
Smarter control in robotics & automation!
The Orocos ProjectSmarter control in robotics & automation! |
|
combining joint limit avoidance and singularity avoidance
are there some papers concerning the combination of joint limit avoidance and avoidance of singular configurations available ( of course, during the same trajectory)? |
combining joint limit avoidance and singularity avoidance
On Sun, 12 Oct 2008, manuelbirlo [..] ... wrote:
> are there some papers concerning the combination of joint limit avoidance and
> avoidance of singular configurations available ( of course, during the same
> trajectory)?
>
This is rather trivial: both can be done by applying soft constraints on
the allowable joint values. For example, by implementing virtual repellent
forces on the joints when they come close to their limits and/or their
singular configurations.
Ruben has implemented such things in some visual servoing applications in
Orocos.
Herman
combining joint limit avoidance and singularity avoidance
>This is rather trivial: both can be done by applying soft >constraints on
>the allowable joint values. For example, by implementing >virtual repellent
>forces on the joints when they come close to their limits >and/or their
>singular configurations.
Yes, I know, but my question has the following goal:
Actually, I've written a function which projects joint angel vectors into the null space. And I have 2 gradient vectors:
- a gradient vector which is for joint limit avoidance
- a gradient vector which is for singularity avoidance
If I want to execute the joint limit avoidance, I project this gradient vector into the null space, and if I want to execute the singularity avoidance I project the secont gradient vector int the nullspace. Now, I want to combine both gradient vectors and project the resulting vector into the nullspace, so that both avoidance tasks are executed during the same trajectory. Are there some papers available concerning this combintion of gradient vectors?
I could write an if-constraint:
- if the manipulability measure comes close to zero, then project the singularity avoidance gradient into the nullspace
- else project the joint limit gradient into the nullspace
But I'm not sure if this is the best idea?
RE: combining joint limit avoidance and singularity avoidance
> >This is rather trivial: both can be done by applying soft >constraints on
> >the allowable joint values. For example, by implementing >virtual repellent
> >forces on the joints when they come close to their limits >and/or their
> >singular configurations.
>
> Yes, I know, but my question has the following qoal:
>
> Actually, I've written a function which projects joint angel vectors into the null space. And I have 2 gradient vectors:
>
> - a gradient vector which is for joint limit avoidance
> - a gradient vector which is for singularity avoidance
>
> If I want to execute the joint limit avoidance, I project this gradient
> vector into the null space, and if I want to execute the singularity
> avoidance I project the secont gradient vector int the nullspace. Now, I
> want to combine both gradient vectors and project the resulting vector
> into the nullspace, so that both avoidance tasks are executed during the
> same trajectory. Are there some papers available concerning this
> combintion of gradient vectors?
Probably not, because this is trivial: since both vectors are gradients
from potential fields, and potential fields can be summed, just sum your
two vectors...
Herman
Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
RE: combining joint limit avoidance and singularity avoidance
>Probably not, because this is trivial: since both vectors >are gradients
>from potential fields, and potential fields can be summed, >just sum your
>two vectors...
this was my first idea, but I thought that maybe some "better" ways exist to solve this problem. But if this is the best way.....ok! :)