Applying Vision Offsets with a KUKA Robot

The first questions are:

Is this camera fixed, or mounted to the robot?

Are these offsets in a fixed reference frame (Base), or a Tool offset?

What Base is your path taught in?

We need to see what your path program looks like, and your I/O setup between the robot and Keyence.

Typically, you teach the path and the vision to a part that becomes the 0 reference -- that is, the vision system produces offsets of 0 for a part exactly there, and the path (dispense in your case?) is taught to that part.

Then, ideally, when the part moves, the vision system measures the difference between the new position and the 0-reference position, and sends offsets to the robot. The robot applies those offsets, and runs the same path.

If we assume a fixed camera, and that the vision system and robot have already been calibrated to each other (sometimes called hand-eye calibration), then you're probably using Base offsets. Something like:

DECL FRAME _fShift ; create temporary Frame for offset
_fShift = $NULLFRAME ; init to all 0s
_fShift.X = KeyenceX ; get the X offset from the vision
_fShift.Y = KeyenceY ; get the Y offset from the vision
_fShift.Z = KeyenceZ ; get the Z offset from the vision

BASE_DATA[2] = BASE_DATA[1] ; assuming the "nominal" frame is Base 1
BASE_DATA[2] = BASE_DATA[2] : _fShift ; apply the shifts to Base 2
$BASE = BASE_DATA[2] ; activate Base 2
PTP P1
LIN P2 C_DIS
LIN P3 C_DIS
LIN P4 C_DIS

Applying offsets to Frames in KRL is done using the Geometric Operator, which has a lot of discussions in the forum archives. But this is the basic idea. If you ran this code and KeyenceX was +10, the path would be run 10mm from where it was originally taught, on the X+ axis of Base 1.

You don't want to change your nominal Base, which is why in this example I copy it to Base 2, then shift Base 2, then run the program with Base 2 active. That (hopefully) prevents Base 1 from being corrupted by a programming error.

Quote from TheCarebear

Probably base 0 or 1, we are matching the camera base to the robot base so they will be the same.

Mmm... Base 0 doesn't always work. Depends on what vision system you're using. It also means that your offsets are being calculated from a reference point fairly far away from the actual working position, which means small rotations (like minor noise or rounding errors in the offset measurement) incur larger motion errors. Ideally, the origin of the reference frame should be as close to the work as is practical.

Typical approach is to put the calibration grid on the work table (so, directly under the camera), then perform the robot and camera calibration to that grid without moving it. That generally ties the reference frames of both robot and camera together nicely. This often puts the origin of the shared reference frame in the middle of the work area, so any rounding errors tend to get averaged out evenly.

Also, I noticed you said XYZ offsets. Are you using 2D or 3D? Typically from a 2D overhead camera you get X, Y, and Rz -- Z offsets from a 2D camera are relatively rare, though not unheard of, and generally rely on measuring the apparent "size" of the target to get a rough value for how much closer or further it is from the camera than the 0-reference part.

Why do you want to use PLC in the middle between Keyence (which one?) and robot? Ask your Keyence support for all manuals for your vision system and read them. Keyence vision systems CV-X and XG-X both can control robots as slaves as long as you run provided solution.

I hope your camera is not IV series, as these are not vision systems but vision sensors and can not give you coordinates.