Posts by robotrev3

I am using some 3d printed grippers on my fanuc robot. These are softer than metal, and when a collision occurs, they tend to bend. Is there a way to make the collison detection force more sensirive, so that the robot detects lesser force on the grippers and stops the robot? Thanks

Other learnings I had along the way -

- While setting the first plane for 2-plane-calibration, make sure you reposition the robot after calibration-execution and just before setting the first plane

-Make sure you choose the correct user-frame in the irvision

SOLVED

HawkME  Fabian Munoz  jmd2777

The problem was that the camera had shifted slightly after I ran the calibration.

I ran the calibration again today and now everything works perfect.

Calibration is a time intensive process since I had to take out some fixtures and stuff to get the robot to position where we wanted it to, but I should've done the calibration much earlier.

Thanks for everyone who helped

Thanks for your advice HawkME

Fabian Munoz

That was a good thought, that's what I assumed too.

However, The transnational error is the same no matter where I place the part (center, or at the corners of the search window). The transnational error seems to be dependent only on the angle.

And conversely, when the part is parallel to the robot's y axis, the robot can perfectly pick up the part from anywhere in the search window

I tested this out a few minutes ago

HawkME

Thanks for your response!

I am using grippers to pick up the cylindrical part. I tried changing the Z offset such that the gripper center point coincided with the 1) the surface on which the part sits and 2) the center of the part when its lying down (+3.3 mm). In both cases, it did not help the situation, and I got the exact same behavior from the robot.

Despite the parts being cylindrical, I have small bumps on the surface which stop the parts from rolling endlessly. in all the IRvision tests for this part, I also ensured that the part was not moving. Actually, another robot that we have set up also uses the IRvision with similar cylindrical parts and it works flawlessly. I am still going to try running my current system with some type of cuboid parts to see if that has some affect.

Fabian Munoz thanks for your time

The area from which I am picking up is an area of about 15 cm by 10 cm. The search window and snap window are configured to ensure that the parts are recognized within these bounds. I am sorry but I am not sure what you mean by grid? Hopefully you got a little more info of my application

When I put the parts in the search window, the IR vision finds the parts without any problem. I know this since the IRvision program draws a green box around the part. IR vision also shows the origin of each part with a + sign on each part too. Please see an illustration of the same

I tried putting one part in the "horizontal" orientation, and another part at 45 degrees to it, and I got the box and origin at the places where I expected.

The origin is shown with a small green + sign, however the + sign doesn't change its orientation with the angle of the part. I am assuming that is by design, since the robot doesn't make a rotational error while picking up the part, it only makes a transnational error.

HawkME

I think the Red screenshot you mentioned belongs to jmd2777 and they showed the image just for reference. From what I understand, there are no red parts in my IRvision program

Unfortunately due to my policy, I don't have permission to show the silhouette of the part. I can tell you though that the part is a cylinder-like part.

It has varying diameters (in sections) across its length. It has a max diameter of 6.6 mm which is at the ends of the part. The part is 20 mm in length. The parts lies on the surface such that it can roll. In the irvision program, the silhouette is approx 20 by 80pixels.

The irvision can detect the features of the part without much problem, although it does have a little trouble (10% of time) deciding if the part is straight or upside down (e.g. if part is 5 deg or 185 deg) due to the near-symmetry of the part

Fabian Munoz

1. Yes, the Z height is set to zero right now

2. The part is a small cylinder-like part. The application frame was taught on the surface that the part rests on

The cylindrical part has a diameter of 6.6 mm, so its center sits 3.3 mm above the application frame

However, it tried changing the Z height to 3.3 mm (and retouched the picking points) and the problem still persisted for parts lying at an angle. I also ensured that I saved any changes I made to the irvision settings, and I confirmed the fact that I am using the correct IR vision program and TP program

jmd2777 Please see the image below. Those options were hidden under advanced mode

I have to log off now, but I will try out any suggestions on Tuesday. Thanks for your help!

jmd2777 I am a fanuc novice and I don't really know what application frame and grid frame are, or if I have used them in my vision program

In the next few of my comments please find screenshots of the 2d process and the related things

To clarify, the amount of offset I am getting is about 5 mm by 1 mm when the part is at a 45 degree angle (relative to the angle of the part that I use to retouch the picking point).

The direction of the offset depends on if the angle is positive or negative

HawkME, I used the calibration tool to calibrate the camera, please see the screenshot. The focal length of the camera is 12 mm, which is approx 12.14mm given by the calibration tool.

Yes, I believe I did teach a user-frame from the point i am picking up the part from. I did this by copying the values that the camera calibration gave me, and putting those values in a new user-frame.

At the start of the calibration, I used a height Gage to ensure that the calibration grid was at the same height from where I would pickup the parts. I also confirmed that my camera, the picking surface and the calibration grid all were parallel to the ground.

As for the application, I have a camera mounted about 0.8 meters from the picking point, and I have a robot come in to pick up the parts using pneumatic grippers. Precision is necessary since the gripper jaws utilize a small feature on the part to hold them very securely without any slippage. The robot actually has two grippers, such that each gripper is "facing away" from the other. I set my tool frame from the gripper that I am picking up the part from. The other gripper is used to pick up the part from the fixture where the main gripper originally puts in the part.

Thanks for the super quick response

Fabian Munoz , here is a screenshot of 3 parts I set up in FOV. One of the parts was at an extreme angle so was rejected and shown as almost found (the parts are almost symmetrical). All in all the GPM locator tool performs as I would expect. Unfortunately I can't show the silhouette of the part since I don't have the permission to do so. However, the part (a cylinder) measures approx 80 by 20 pixels in the FOV, if that makes sense. I feel that is a little low but the vision is still able to identify the part reliably