Vision Pick

[Kinne]

Hey guys,

I'm need to create a Vision Pick but don't know how, unfortunately I don't have access to the training portal.

It's supposed to be a very simple pick and place of an static object that's somewhere on a table. I know how to create a locator in the snapshot galery and also how to pick and place. What's missing is the connection inbetween. How do I tell Sawyer to pick where the object is located?

 

Best regards and thank you,

Katharina

 

[lsobczuk]

Hello Katharina ,

If you know how to create a locator and how to do pick and place it will be simple for you to create Vision Pick.

Below you can find a short instruction on how to create it :

1. Create Locator in Snapshot gallery. (Just in case instruction here)- DO NOT MOVE YOUR OBJECT   

2. Close all windows/nodes in Intera.

3. Open Vision Locate node and choose your object.

4. Click on the button TEST (in Vision locator node - at the top left corner)

5. If your object is found, it should be highlighted in green and then click RELOCATE. (there should be confirmation - RELOCATED) - DO NOT MOVE YOUR OBJECT

6. Create Pick & Place (most important are : Pick should be children of the Vision Locator node - toggle on the vision locator node, click on this and it should be highlighted in orange, then create Pick & Place)  

Remember, your object has to be in one place all time after you execute (Point 5 - RELOCATE).

Good luck.

Please let me know if it is helpful for you. 

Cheers

Lukasz

Edited February 4, 2019 by lsobczuk

[Steffen Scherf]

Thank you Lukasz!! 

 

[Kinne]

Hi Lukasz,

Thanks a lot for the nice description!

But we have a problem as you can see in the attached picture. We are not able to choose "relocate" if we test the snapshot.

Do you have any idea?

 

Thanks in advance!

Katharina & Max

[lsobczuk]

Hello Katharina ,

This button is available only when your object was recognized by vision system.

First press the button GEHE ZU  - the robot goes to the position where you were doing the snapshot.

If it does not help so.. I suppose that you have to change settings e.g. 
1. similarity threshold
2. start angle
3. end angle

or try to create snapshot and vision locate from the beginning and remember that you can not move your object during programming. 

Cheers

Lukasz

Edited February 6, 2019 by lsobczuk

[Steffen Scherf]

Hello Katharina, Max and Lukasz,

I´m Steffen Scherf the Head of Application Engineering  from Rethink Robotics GmbH

Very nice description from Lukasz!

 

If you have any question you can also contact me directly.

s.scherf@rethinkrobotics.com 

[Kinne]

Hey,

thanks again for your response. Unfortunately it's still not working. 

We pressed "Gehe zu" and set the similarity to 20% and the angles to -180 and 180, but still couldn't press "Verschieben". We tried over and over again, but without success.

The problem might be that it says on the screen "Armcamera" instead of "Armcamera is proofing" (or similar). The camera does find the object though and there is the green rectangle around it.

 

Best, 

Katharina and Max

 

[lsobczuk]

Hello,

Your object is recognized (highlighted in green) but look that not all.

 

Two edges are outside of green rectangle.

Try to broaden object field, because all edges should be inside of field,  green rectangle (during snapshot) cannot cut of any object edge.

Also i have few question : 

1. If you create surface in snapshot editor you put Landmark on the table or on your object ? 

2. Does your object lie on the hard or soft base ?

Cheers 

Lukas

Edited February 8, 2019 by lsobczuk

[mwatt]

Hoping you'll be able to help me with a couple questions I have about SAWYER and Intera.

 

First of all, I was wondering what the difference between Intera 5.3 and SDK is. I understand SDK uses ubuntu and ROS (I've used a bit of this before) but do the capabilities of the arm change drastically?

 

Secondly, I'm trying a vision pick and place where the arm needs to scan a surface for a filter (about the size of a 5p coin) and then drop it into a tray. If I put the filter in the same location every time, the task suceeds, but if the filter moves, the arm missed the filter. It does track there's a change in location and the arm moves close to the new location, but the further away from the original position the filter is, the more error there is in picking it up.  

 

I've ensured the move to nodes are children of the vision locator and have used a landmark to mark the surface.