[ univeral_robot ] UR10: link length differences xacro and 'real' dimensions?

[G.A. vd. Hoorn - 3ME]

Just wondering whether I'm misunderstanding something, or if there is
really something wrong with the current UR10 xacro in the repository.

It would seem the link lengths as specified in the xacro for the UR10
[1] are different from those given by UR in [2]. For instance, the
'upper arm length' is given as 0.60186m in the urdf, while UR seems to
specify 0.6127m in their pdf.

Am I just interpreting these numbers wrong?

thanks,


Gijs

[1]
https://github.com/ros-industrial/universal_robot/blob/hydro-devel/ur_description/urdf/ur10.urdf.xacro
[2]
http://ur-update.dk/URsupport/MechanicalDrawings/2D_drawings/UR10/UR10_working_area/UR10_working_area_and_dimensions.pdf

[Maarten de Vries]

In case Gijs is not interpreting the numbers wrong (and I don't think he is), the shoulder and elbow offset and wrist1 lengths are also wrong. These are not shown in the the drawing [1]. My colleague measured them in solidworks using the step file [2] to be:

shoulder_offset: 176.00 mm
elbow_offset:    127.81 mm
wrist_1_length:  115.70 mm

This can be checked against the drawing by calculating shoulder_offset - elbow_offset + wrist_1 = 163.89 mm which matches the 163.9 mm given in the drawing closely.

Note that that makes wrist_1_length and wrist_2_length the same. This may seem coincidental, but it appears to be correct.

With the unmodified URDF I can see a clear difference between a pose I give to the inverse kinematics from ur_kinematics and the one visualized in rviz. When I use the lengths from the 2D drawing and those measured from the step file, I can't visually see any difference.

Note that the lengths in the ur_kinematics package are also not exactly the same as those from the drawing, but they do match much more closely.

Regards,
Maarten de Vries

[1] http://ur-update.dk/URsupport/MechanicalDrawings/2D_drawings/UR10/UR10_working_area/UR10_working_area_and_dimensions.pdf
[2] http://ur-update.dk/URsupport/MechanicalDrawings/3D_STEP_files/UR10/UR10_step.zip

--
You received this message because you are subscribed to the Google Groups "swri-ros-pkg-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email to swri-ros-pkg-d...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[baron...@gmail.com]

If it can help anyone here are the values i have used in the ros-industrial URDF for the UR10, they are a combination of your suggestions and the drawings from UR.. They seem to give results very close to what the UR controller states (<1 mm error). In comparison the original URDF in ros-industrial resulted in errors above 1cm:

  <property name="shoulder_height" value="0.128" />  

  <property name="shoulder_offset" value="0.176" />  

  <property name="upper_arm_length" value="0.6127" />

  <property name="elbow_offset" value="0.12781" />       

  <property name="forearm_length" value="0.5716" />

  <property name="wrist_1_length" value="0.1157" />     

  <property name="wrist_2_length" value="0.1157" />   

  <property name="wrist_3_length" value="0.0922" />

[GMAIL]

In case this really is a confirmed error, can you please propose the correct URDF in a PullRequest to https://github.com/ros-industrial/universal_robot?

 

Is there a similar issue with the UR5?

[G.A. vd. Hoorn - 3ME]

On 13/06/14 13:52, GMAIL wrote:
> In case this really is a confirmed error, can you please propose the
> correct URDF in a PullRequest to
> https://github.com/ros-industrial/universal_robot?

The issue is that it is currently not entirely clear what would be the
correct values, as far as I understand.

Also personally, I'd like to first understand where the original values
came from, and why they were modelled like this, before starting to
change things. Does anyone know where the UR10 urdf came from?

> Is there a similar issue with the UR5?

Not that I know of, we've only seen it with the UR5.


Gijs


> *From:*swri-ros...@googlegroups.com
> [mailto:swri-ros...@googlegroups.com] *On Behalf Of
> *baron...@gmail.com
> *Sent:* Wednesday, June 11, 2014 1:49 PM
> *To:* swri-ros...@googlegroups.com
> *Subject:* Re: [ROS-Industrial] [ univeral_robot ] UR10: link length

> differences xacro and 'real' dimensions?
>
> If it can help anyone here are the values i have used in the
> ros-industrial URDF for the UR10, they are a combination of your
> suggestions and the drawings from UR.. They seem to give results very
> close to what the UR controller states (<1 mm error). In comparison the
> original URDF in ros-industrial resulted in errors above 1cm:
>
> <property name="shoulder_height" value="0.128" />
>
> <property name="shoulder_offset" value="0.176" />
>
> <property name="upper_arm_length" value="0.6127" />
>
> <property name="elbow_offset" value="0.12781" />
>
> <property name="forearm_length" value="0.5716" />
>
> <property name="wrist_1_length" value="0.1157" />
>
> <property name="wrist_2_length" value="0.1157" />
>
> <property name="wrist_3_length" value="0.0922" />
>
>
> On Tuesday, May 27, 2014 2:59:13 PM UTC+1, maarten wrote:
>
> In case Gijs is not interpreting the numbers wrong (and I don't think he
> is), the shoulder and elbow offset and wrist1 lengths are also wrong.
> These are not shown in the the drawing [1]. My colleague measured them
> in solidworks using the step file [2] to be:
>
> shoulder_offset: 176.00 mm
> elbow_offset: 127.81 mm
> wrist_1_length: 115.70 mm
>
> This can be checked against the drawing by calculating shoulder_offset -

> elbow_offset + wrist_1 = 163.89 mmwhich matches the 163.9 mm given in

> the drawing closely.
>
>
> Note that that makes wrist_1_length and wrist_2_length the same. This
> may seem coincidental, but it appears to be correct.
>
> With the unmodified URDF I can see a clear difference between a pose I
> give to the inverse kinematics from ur_kinematics and the one visualized
> in rviz. When I use the lengths from the 2D drawing and those measured
> from the step file, I can't visually see any difference.
>
> Note that the lengths in the ur_kinematics package are also not exactly
> the same as those from the drawing, but they do match much more closely.
>
> Regards,
> Maarten de Vries
>
> [1]
> http://ur-update.dk/URsupport/MechanicalDrawings/2D_drawings/UR10/UR10_working_area/UR10_working_area_and_dimensions.pdf
> [2]
> http://ur-update.dk/URsupport/MechanicalDrawings/3D_STEP_files/UR10/UR10_step.zip
>
> On 27 May 2014 15:40, G.A. vd. Hoorn - 3ME <g.a.van...@tudelft.nl

[G.A. vd. Hoorn - 3ME]

On 13/06/14 14:16, G.A. vd. Hoorn - 3ME wrote:
> On 13/06/14 13:52, GMAIL wrote:

[..]

>> Is there a similar issue with the UR5?
>
> Not that I know of, we've only seen it with the UR5.

Of course I meant to write: "we've only seen it with the UR10."


Gijs

[Maarten de Vries]

This seems like a good time to look at the lengths used by the kinematics package too. I noticed that the ones used there also do not match the drawings [1][2] provided by Universal Robots. The differences are much less severe than with the UR10 URDF, possibly even negligible, but they're not exact matches either.

For example, for the UR5:

#define d1  0.089159 // 0.0892 according to UR
#define a2 -0.42500  // matches exact
#define a3 -0.39225  // (-)0.392 according to UR
#define d4  0.10915  // 0.1093 according to UR
#define d5  0.09465  // 0.09475 according to UR
#define d6  0.0823   // 0.0825 according to UR

Note that I guessed the matching length in the drawing with each parameter, but given the small differences it seems very likely that the guess is correct.

So for these parameters I have the same question. Where did they come from and is there a good reason that they differ from the drawings?

Kind regards,
Maarten de Vries

[1] http://ur-update.dk/URsupport/MechanicalDrawings/2D_drawings/UR5/UR5_working_area/UR5_working_area_and_dimensions.pdf
[2] http://ur-update.dk/URsupport/MechanicalDrawings/2D_drawings/UR10/UR10_working_area/UR10_working_area_and_dimensions.pdf

[GMAIL]

I opened an issue https://github.com/ros-industrial/universal_robot/issues/74

 

 

--