Robot position variables are always required to be a 16-dim vector + robot DoFs

[francesco-romano]

All the blocks which require a position input have dimension 16 + DoFs.
The 16 dim is the Frame object associated with the base.

I think we should move to use quaternion + displacement (then dim = 7) instead of specifying the whole matrix (which in my opinion does not make any sense because it is not a free matrix. There are constraints on the values we can put into it)

cc @DanielePucci @traversaro

[naveenoid]

Popping in because of a prior discussion with @traversaro. I personally
prefer rotation matrix as it leaves a simple typecast to a Frame object
available to the end user. Even a quaternion is not a free R^4 vector, it
is constrained. However quaternion has an issue that its not obvious unless
someone looks up documentation if its a axis-angle or a quaternion. Even
then, there are 2 quaternion conventions : real followed by imaginary
vector and vice versa. So could lead to misunderstandings...On the other
hand there is the advantage of number of variables returned...

---

Naveen Kuppuswamy, PhD
Post-doctoral Fellow,
Cognitive Humanoids Lab,
Department of Robotics, Brain and Cognitive Sciences (RBCS),
Istituto Italiano di Tecnologia, Genova, Italy

On Tue, Mar 3, 2015 at 9:21 AM, Francesco Romano notifications@github.com
wrote:

All the blocks which require a position input have dimension 16 + DoFs.
The 16 dim is the Frame object associated with the base.

I think we should move to use quaternion + displacement (then dim = 7)
instead of specifying the whole matrix (which in my opinion does not make
any sense because it is not a free matrix. There are constraints on the
values we can put into it)

cc @DanielePucci https://github.com/DanielePucci @traversaro
https://github.com/traversaro

—
Reply to this email directly or view it on GitHub
#82.

[DanielePucci]

I prefer the element of SE(3) as input. What do you think of separating the orientation-translation of the base from the joint positions? In this case, we may take as input a 3x4 matrix (i.e. [R,t]) and an nth dimensional vector (i.e. qj).