robot calibration method?

What method did you use to calibrate an arm robot?

I know we offten use home switch to calibration CNC machines, but for robots i dont know what method to use for arm robot to set zero point for joints.If any expert can recommend me the method that is used for arm robot of ABB,KUKA,Yaskawa,..

KUKA has a physical switch that you have to mount on the joint you are calibrating, wired to the base of the robot. It has to be bought separately and the price is......high.
On robots you have to also do the tool calibration after joint calibration that can be done by entering the exact values and offsets to the tool tip or KUKA has a nice helper wizard that you point the tip of the tool to a tip mounted firmly on the ground or table, from 3 different angles and it calculates the tool tip from those readings.
Having limit switches is nice but for the tip you really need accurate info.

Many use some form of absolute encoder in the joints.

If a kuka robot has had a big collision, you can screw up a zero point sensor device, this device costs about 1800 euro. It's a sensor clock. You have to repeat this proces for all individual axis. Your robot will be back in position within 1mm.

But there is a trick. First they home in for example 5% speed. After the sensor has triggered, they do it again with very slow speed.
This searching is not to sensor, but also over sensor. So the calibration is to the sensor has triggered and after the sensor has triggered off. And also very very slow for final adjustment.

Conclusion is, keep on the encoder power with battery.

I think when you modify robot with linuxcnc. You can also do second homing with less speed, but second less speed home procedure is not possible at this moment. Only a different linuxcnc config startup procedure will do what you want. :woohoo:

Good luck.

What precision do you need?

I work with Fanuc and ABB robots, they have absolute encoders. If the home position is lost due to a failed backup battery or disconnected cable, I can restore the zero point for each axis by lining up indicators on each axis. The control then goes through a procedure moving each axis a small amount, presumably searching for the closest major division on the encoder. Once the zero point is set no further calibration is needed as long as there wasn’t any mechanical damage that caused the robot to loose its zero position.

No limit switches are used, the same procedure is used to set the zero position on every other machine I have worked on that uses absolute encoders, you line up an indicator and perform a zero return. I don’t have experience with absolute encoders in Linux cnc, but I imagine setting the zero position is very similar to any other machine controller, it should be a function that is baked in, but I don’t have any experience setting it up so I can’t help much.

With absolute encoders the machine remembers its position when turned off, and knows if it has moved without power. There aren’t a lot of reasons to re-zero a machine with absolute encoders, so home switches are an unnecessary cost and use up IO unnecessarily. In the rare instance an axis needs to be zeroed, the indicators are more or less a manual home switch.