[ros-users] multi_dof_joints parameter in arm_navigation stack

Gil Jones gjones at willowgarage.com
Fri Feb 4 18:15:32 UTC 2011

Hi Felix,

As Sachin said, the bug you were seeing should be fixed.  By way of
explanation, we were trying to accomplish a few things with the multi-dof
transform.  The primary purpose is to supply a mechanism for a world->robot
transform that supplements the urdf, which doesn't and shouldn't really
supply information about how to transform between a fixed world frame and a
frame for the movable base.  The world transform should be either be a 3-dof
Planar Joint with (x,y,th) or a 6-dof Floating Joint with a full orientation
matrix.  We found on the PR2 that we had to use a floating joint as there
can occasionally be enough of a pitch and roll in our IMU such that they
need to be accounted for.  It's entirely fine to do what you've done below,
which is to say that you essentially don't care about the fixed transform
and will always be planning based on the current base position of the
robot.  So that should be all you need to do.

Having a fixed reference transform should allow two important pieces of
functionality - as you say, it should allow you to plan to move the base,
and it should also allow you to pose the robot some where else in the world
and plan based on that new pose - for instance, you might be interested to
know whether an object across the room from the robot's current location can
be picked up given a particular approach pose.  By specifying a new
multi-dof transform in the start_state message our system should support
such things, though it is not the case that our current planners fully
support the former use-case.

The other purpose of the multi-dof joints is to equip the system to use tf
transforms instead of just joint_state messages to populate the current
position of the robot.  The system is set up to first try to set multi-dof
joints from individually labeled entries in the joint state and then to use
tf to get a pose transform for the entire joint if it needs to.  The message
you are seeing refers to one of the labeled single dofs - 'floating_trans_x'
- of the multi-dof joint 'base_joint'.  The bug was that for the purposes of
checking joint limits the individually-labeled dofs were not being mapped to
the multi-dof joint.

As arm_navigation matures we will try to do a better job of making migration
guides that explain new features and inform users what they need to do to
adopt a new version.  At this point we're focused on making relatively large
changes that we hope will make things work better and make it easier to get
our stacks working on other arms, but we still have a ways to go.  We're
targeting e-turtle for a 0.9 or 1.0, at which point things should stabilize
a good deal.

Please let us know if the bug fix doesn't work or if you have other


E. Gil Jones (gjones at willowgarage.com)
Research Engineer
Willow Garage, Inc.
68 Willow Road
Menlo Park, CA 94025
650.475.9772 <tel:+16504759772>

On Fri, Feb 4, 2011 at 1:23 AM, Sachin Chitta <sachinc at willowgarage.com>wrote:

> Hi Felix,
> This should be fixed in motion_planning_common 0.3.5 (thanks to Gil).
> Can you try it out and let us know.
> Thanks,
> Sachin
> On Wed, Feb 2, 2011 at 4:09 AM, "Felix Meßmer" <Felix_Messmer at web.de>
> wrote:
> > Dear community,
> >
> > there is another feature in the unstable/diamondback version of the
> arm_navigation stack that I have problems with:
> > the multi_dof_joints parameter-option!
> >
> > When I use our old launch file
> (cob_arm_navigation/cob3_planning_environment.launch) that loads the
> collision_checks and planning_groups paramters to the parameter server, I
> get the following error message from both the trajectory_filter and the
> ompl_planning node:
> >
> > [ WARN] [1296646033.585385464]: No multi dof joints specified, including
> root to world conversion
> > [ WARN] [1296646033.585623806]: Can't do anything without a root
> transform
> >
> > I've adapted the planning_environment.launch file to also load the
> multi_dof_joints parameter (looking at the according pr2-config files).
> > But now planning reqest as in move_arm_simple_pose goal fail due to the
> following error message from the environment_server node and the move_arm
> action server node:
> >
> > [ WARN] [1296647088.916018038, 830.563000000]: No joint with name
> floating_trans_x
> > [ WARN] [1296647088.916446633, 830.563000000]: Joint limits violated.
> >
> > [ERROR] [1296647088.925356070, 830.563000000]: Will not plan to requested
> joint goal since it violates joint limits constraints
> >
> >
> > The multi_dof_joints.yaml file I use looks like this:
> >
> > multi_dof_joints:
> >    - name: base_joint
> >      type: Floating
> >      parent_frame_id: base_footprint #odom_combined
> >       child_frame_id: base_footprint
> >
> >
> >
> >
> > I guess the multi_dof_joint parameter is meant to introduce a joint that
> stands for the moving platform with 3 DoFs (trans_x, trans_y, rot_z)
> > Does that mean that if the pose goal is specified in a world/global
> coordinate system, the planner also considers and calculates a possibly
> necessary platform motion to reach the specied goal?
> >
> > Is there a parameter (use_multidof?) for the environment_server to
> deactivate multi_dof_joints?
> > OR:
> > What else do I have to add to use that feature?
> >
> >
> > I hope you can give me a short explanation on that. Thanks a lot!
> >
> >
> >
> > Regards,
> > Felix
> > ___________________________________________________________
> > WEB.DE DSL Doppel-Flat ab 19,99 €/mtl.! Jetzt mit
> > gratis Handy-Flat! http://produkte.web.de/go/DSL_Doppel_Flatrate/2
> > _______________________________________________
> > ros-users mailing list
> > ros-users at code.ros.org
> > https://code.ros.org/mailman/listinfo/ros-users
> >
> --
> Sachin Chitta
> Research Scientist
> Willow Garage
> _______________________________________________
> ros-users mailing list
> ros-users at code.ros.org
> https://code.ros.org/mailman/listinfo/ros-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ros.org/pipermail/ros-users/attachments/20110204/e3f1192e/attachment-0003.html>

More information about the ros-users mailing list