Ask your questions - Robotiq Vision System (software)

Olivier_Grenier

Hi everyone,

Here are questions we were asked during the vision system training about the presentation on the software of the Robotiq vision system 

Feel free to ask other ones you may have in mind!

Grady_Turner

Has any testing been done with the Camera Kit used with UR's built in conveyor tracking functionality?

Annick_Mottard

@gradyturner86

As of today, the robot's workplane has to be static in order for the Camera Locate feature to work. This feature allows you to do part detection and go pick the object with the orientation you taught it. This would be difficult to implement with a moving conveyor. The Wrist Camera takes snapshots of the workplane until it detects a part. It can then take actions with positions relative to the object's frame. If the object is placed on a moving conveyor, all the relative movements would not be possible. 

This being said, we have not looked into the application with a moving conveyor yet. We are taking notes and will evaluate the feasibility!

Adam529dp

Are their concerns with exceeding the read/write of the USB?  Also how does an end user handle a corrupt USB?

Ryan_Weaver

Question for Robotiq/Stefan -

What steps is UR and Robotiq taking to ensure that the camera software works well with future versions of Polyscope?  In the past there have been problems where drivers don't work on the latest UR version, until Robotiq is able to release a new driver. 

Are UR+ developers given alpha/beta access to new Polyscope releases to ensure they can develop updates ahead of time?

Grady_Turner

Regarding symmetry or unique features to determine orientation:  Is there an option to ignore orientation?  Some applications will not require a specific orientation of the part in the full +/- 180 degree space.

Simple example is a cylinder.  If all we need to do is pick the object and move it  (all we care about is location not orientation) then the system can choose the best way to pick the object.  For something like a cube, it would just need to rotate +/- 45 degrees maximum to pick up the cube on a given side. 

It might be very useful to have an OPTION in the wizard to choose whether the user cares about model orientation.  Does this make sense?

eric

gradyturner86 said:

Regarding symmetry or unique features to determine orientation:  Is there an option to ignore orientation?  Some applications will not require a specific orientation of the part in the full +/- 180 degree space.

Simple example is a cylinder.  If all we need to do is pick the object and move it  (all we care about is location not orientation) then the system can choose the best way to pick the object.  For something like a cube, it would just need to rotate +/- 45 degrees maximum to pick up the cube on a given side. 

It might be very useful to have an OPTION in the wizard to choose whether the user cares about model orientation.  Does this make sense?

Yeah, this is going to be essential for some customers that are picking up blank pieces to be loaded into a machine for cutting.

Annick_Mottard

@Ryan_Weaver

As discussed by @PierreOlivier_Proulx, our team gets the beta versions for Polyscope before it is released. Thus, we are able to test our new URCaps and legacy driver packages with those beta versions. 

Note that when a Robotiq version is released, it is compatible with the newest available Polyscope version. 

Annick_Mottard

@gradyturner86

Yes, it makes sense! For the first release, this kind of option will not be available. Therefore, for a circular object, or a square, the vision system will pick the part in the orientation it was taught. 

It is definitely something we will want to look into!

Annick_Mottard

@Adam529dp

The read and write process is not a concern for the 16G USB stick we provide, as it is not done often.

 

Kaleb_Rodes

If a part is not detected or recognized by the camera, is there a certain number of attempts/pictures the camera will take until it moves on in the program?

Cody_James

Does the TCP have to be set before going through the calibration process (to avoid running the end effector into the calibration grid)?

Annick_Mottard

@Cody_James

The TCP configuration will not affect the calibration process, nor the scanning process during the object teaching. 

Enric

Is there a difference when calibrating the camera with the force torque sensor and without?

PierreOlivier_Proulx

@Kaleb_Rodes The Camera Locate node takes only one snapshot of the scene. If no object has been found it moves on to the next instruction in the program.

PierreOlivier_Proulx

@Enric With or without sensor, the calibration will complete properly. However, if you calibrated with a sensor, this calibration will only be valid if the sensor is still mounted on the robot. To put it otherwise, don't move the camera's position relative to the flange otherwise you will have to re-calibrate.

PierreOlivier_Proulx

@Adam529dp We write on the flash drive at each calibration and at each taught object. At runtime we do not write the flash drive. For reads, there is not limits on flash memory. In case the drive is corrupted, the customer will have to download the flash drive content and put it on a new usb stick.

UR uses the same memory technology we use but they (my assuption) probably make much more writes than we do on their flash drive. Hence, I expect their drive to fail before ours.