Force/Torque feedback of UR robots

[Athanasios Polydoros]

Hi all,

I have a simple question about the feedback that is provided by  the joint_states topic on UR robots.

The topic publishes info about the force/torque on each joint, but the UR robots do not have embedded torque sensors.

Thus, those values are the current on each joint , or something else ?

Best

Athanasios

 

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

On 8-10-2014 19:15, Athanasios Polydoros wrote:
> Hi all,

Athaniasos,

The values you are referring to are those returned by the
get_joint_torques() URScript function. According to the scripting manual
(v1.8, Aug 1, 2013):

"Returns the torques of all joints

The torque on the joints, corrected by the torques needed to move the
robot itself (gravity, friction, etc), returned as a vector of length 6."

You are righ that the robot does not have sensors to measure that
directly, so your suggestion that it uses joint currents could very well
be true.


Gijs

[Leopold Palomo-Avellaneda]

A Dimecres, 8 d'octubre de 2014, G.A. vd. Hoorn - 3ME va escriure:

AFAIK, the UR5 have a innovate method (protected by a patent) to calculate the
torque from the current of the motors. It's not a torque sensor, but gives a
good approximation to the real value.

Regards,

Leopold




--
--
Leopold Palomo-Avellaneda <leopold...@upc.edu>
Institut d'Organització i Control de Sistemes Industrials -IOC-
Universitat Politècnica de Catalunya -UPC-

Institute of Industrial and Control Engineering
Technical University of Catalonia
Avda. Diagonal 647, pl. 11
08028 BARCELONA (Spain)

Tel. +34-934016655 (office)
Tel. +34-934017163 (lab)
Fax. +34-934016605

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

On 9-10-2014 9:09, Leopold Palomo-Avellaneda wrote:
> A Dimecres, 8 d'octubre de 2014, G.A. vd. Hoorn - 3ME va escriure:
>> On 8-10-2014 19:15, Athanasios Polydoros wrote:
>>> Hi all,
>> Athaniasos,
>>
>>
>>> I have a simple question about the feedback that is provided by the
>>> joint_states topic on UR robots.
>>> The topic publishes info about the force/torque on each joint, but the UR
>>> robots do not have embedded torque sensors.
>>> Thus, those values are the current on each joint , or something else ?
>> The values you are referring to are those returned by the
>> get_joint_torques() URScript function. According to the scripting manual
>> (v1.8, Aug 1, 2013):
>>
>> "Returns the torques of all joints
>>
>> The torque on the joints, corrected by the torques needed to move the
>> robot itself (gravity, friction, etc), returned as a vector of length 6."
>>
>> You are righ that the robot does not have sensors to measure that
>> directly, so your suggestion that it uses joint currents could very well
>> be true.
>>
> AFAIK, the UR5 have a innovate method (protected by a patent) to calculate the
> torque from the current of the motors. It's not a torque sensor, but gives a
> good approximation to the real value.
>
> Regards,
>
> Leopold

Is that a patent filed by UR themselves, or are they licensing someone
else's? I could only find two patents, but no mention in either of these
of their torque control / derivation.


Gijs

[Leopold Palomo-Avellaneda]

A Dijous, 9 d'octubre de 2014, G.A. vd. Hoorn - 3ME va escriure:

Well, I wouldn't like to spread false information or similar. But, one year
ago, we studied and finally bought two UR5. I remembered that I searched
information about the security and how the control worked, because we have the
LBR4+ and we wanted to compare.

I remember that I read in someplace that they (UR, not another) had developed
and patented some algorithm to obtain the torque from the current. But, don't
ask me where I found it because I have reviewed my notes, sources and I don't
find it. Maybe I got confused, I don't know or maybe my brain make some
assumption.

Anyway, the UR has some mechanism to obtain the torque values without any
torque sensor, and this value can be obtained measuring the current (I don't
know if _only_). From my poor knowledge, the conversion from current to
torque, using an harmonic drive isn't obvious and has a lot of noise. So, if
they develop a fine current sensor, or whatever to obtain that value is
something that I ignore. The question is that you can obtain a good value of
the torque without any torque sensor.

That's my 2ct,

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

Right. Thanks for the clarification.


Gijs