[prev in list] [next in list] [prev in thread] [next in thread] 

List:       orocos-dev
Subject:    [Orocos-Dev] KDL: proposal for medium-term design. Please discuss!
From:       Herman.Bruyninckx () mech ! kuleuven ! be (Herman Bruyninckx)
Date:       2008-07-31 16:56:57
Message-ID: alpine.DEB.1.00.0807311653350.6556 () roble
[Download RAW message or body]

On Fri, 25 Jul 2008, Advait Jain wrote:

> On Wed, Jul 23, 2008 at 11:52 PM, Herman Bruyninckx
> <Herman.Bruyninckx@mech.kuleuven.be> wrote:
>>
>> Maybe I looked at the wrong place, but in your code I don't see where you
>> adapt the inertia during an iteration over the joints... The inertia than
>> joint i feels must be a function of the inertias of all more distal
>> joints...
>>
>
> In this method we solve for the constraint forces and the joint
> toques. The effect
> of the inertias of the distal joints is reflected in the constraint
> forces. The backward
> recursion (solving for the forces) starts from the last link and goes
> up the chain.

And where are these inertial "constraint" forces calculated...?

>> A side note: "Newton_Euler" or "Lagrange" is not useful information, since
>> the recursive techniques that are derived in both frameworks are the same
>> :-)
>> See @Article{           Silver82,
>>  author          = {Silver, D. B.},
>>  title           = {On the equivalence of {L}agrangian and {N}ewton-{E}uler
>>                    dynamics for manipulators},
>>  journal         = {The International Journal of Robotics Research},
>>  year            = {1982},
>>  volume          = {1},
>>  number          = {2},
>>  pages           = {60--70}
>> }
>> <http://ijr.sagepub.com/cgi/content/abstract/1/2/60>
>> (Free download till the end of the month...)
>>
>
> I agree, I will change the name of the class to something more
> appropriate.
>
>> As a summary: the classes for Motion, Solver and State are still
>> practically nowhere; Chain is already okay to some extent.
>>
>> Ruben (and some others, including myself) would give priority to a damped
>> least-squares inverse kinematics (or dynamics, if you include the inertia
>> as weighting matrix), for a generic serial and tree structure, including a
>> "CLIK" solver (Closed-Loop Inverse Kinematics). See e.g.
>> <http://ijr.sagepub.com/cgi/content/abstract/10/4/410>.
>>
>
> In addition to this, I feel that handling joint limits while computing inverse
> kinematics would also be very useful.
>
Absolutely! Personally, I currently prefer a "potential field" approach, in
which a virtual repelling force is added to joint that come close to their
limits. (Or, an attractive force that tries to keep them close to their
optimal configuration.) But I would welcome very much a more in-depth
discussion about all sorts of alternatives!

Herman

[prev in list] [next in list] [prev in thread] [next in thread]