Hi Tully, The wiki link was very helpful. Here's my two cents: I believe that everything used to define a twist should be explicit in the twist message. As a tf user, I would prefer to provide all the information rather than follow certain conventions. >From the wiki link: "There is a consensus to collapse the reference point and reference frame into a single coordinate frame." Mathematically, this seems ok to me too. However, will this involve a significant overhead in the implementation, such as adding new frames to the tf tree, make the code messy etc, if I want to find the velocity of a large number of different points in the frame of the gripper? Providing a reference point (in reference frame coordinates) does not seem like too much additional information to me. Advait On Tue, Jun 8, 2010 at 1:52 AM, Tully Foote wrote: > Hi Advait, > > There has been a bit of discussion of how to deal with twists well.  However > how to transform them has many different meanings implied by how they are > represented.  In particular we found there were 4 different conventions > without a clear way to distinguish between them.  There is the results of an > API review at http://www.ros.org/wiki/tf/Reviews/2010-03-12_API_Review > > The biggest problem is that there is a lot of semantic information carried > in an instance of a Twist.  Without forcing a lot of conventions on users we > could not find a way to transform Twists without requiring a lot of extra > information. > > There is an API to get the twist between two frames in the cturtle version > of tf. > > And kdl provides may twist manipulation methods which are recommended for > computing things involving twists. > > If anyone else has input we can come back to this.  I would like to add this > capability, but building a consensus on representations will be required. > > Tully > > On Mon, Jun 7, 2010 at 5:00 PM, Advait Jain wrote: >> >> I just realized that the wrench and twist messages currently >> do not explicitly store the point at which the wrench or twist >> is expressed. >> >> Is there is support in TF or somewhere else for getting the >> equivalent twist or wrench at a different point in space? >> >> e.g. to get the equivalent wrench at point B given a wrench >> acting at point A, I would use: >> Force at point A = Force at point B >> Torque at point A  = vector from A to B X Force + Torque at point B >> >> It would be nice to have easy ways to be able to say, for >> example, that given a force that the robot is applying at the >> door handle, what the torque is at the hinge of the door. >> >> Another example could be if the we want a robot to perform >> a bi-manual and need to compute the motion of two points >> on the same rigid body. >> >> I think it would be really cool if TF could also shift twists, >> wrenches correctly in addition to changing coordinate frames. >> >> What are other people's thoughts about this? >> >> >> Advait >> _______________________________________________ >> ros-users mailing list >> ros-users@code.ros.org >> https://code.ros.org/mailman/listinfo/ros-users > > > > -- > Tully Foote > Systems Engineer > Willow Garage, Inc. > tfoote@willowgarage.com > (650) 475-2827 > > _______________________________________________ > ros-users mailing list > ros-users@code.ros.org > https://code.ros.org/mailman/listinfo/ros-users > >