On Wednesday 21 December 2011 12:34:04 Azamat Shakhimardanov wrote:
> Dear Ruben,
>
> after long testing, corrections and discussions with Herman Vereshchagin's
> hybrid dynamics algorithm is ready to be used with serial chains.
> My first question is how I could contribute it back? I could just send the
> current version I have to you. Some parts (many not necessary comments,
> printfs) of it still need to be cleaned up. I could write a short README
> explaining what changed to keep you updated.

The easiest way is to contribute is to provide a patch against the latest git
HEAD which is available on
http://git.mech.kuleuven.be/robotics/orocos_kinematics_dynamics.git

> The other question is related to this algorithm's extension to trees. Herman
> already confirmed theory behind it, so I am working my way through KDL
> tree primitives to see how it could be done. Regarding this I had one
> point, so far.
>
> - What is the semantics of q_nr member in TreeElement. Does it describe the
> joint's number which is part of this element? Since it gets assigned joint
> index number in Tree::addSegment, I was wondering about it.

Since we use a plain JntArray for the joints in a tree each Tree element gets
an index number that will use to find its jointvalue in this array

> - Another point of clarification, Tree primitive assumes that the joints at
> the root of branches are co-located. Example: If tree branches of at
> segment L1 then the joints of branches at L1 tip are co-located. Is it so?

Yes, but you could make the first joint of type None, which means you can put
the next branched joints at any place you would like.

> Also, some docu is missing, but I would be glad to contribute. I could send
> you my notes.

Patches (against the latest git) for documentations are very welcome.

> Sincerely
> Azamat

Ruben