When a KRC4 pose value changes after the first pass through a loop, the fastest explanation is often “the tool offset was applied twice.” That conclusion is tempting—and premature. The result can be influenced by active tool and base data, approximation, advance run and the exact point at which a system variable is evaluated.

Separating commanded and measured paths helps distinguish execution timing from persistent tool or base errors.

The robot should not be recalibrated until those influences have been separated. A compensating frame can hide the symptom while damaging every program that uses a previously correct tool.

Reproduce the symptom without changing calibration

Save the controller project and record KSS version, active tool, active base and program mode. Run the loop at reduced speed with approximation disabled around the suspect capture point if the process allows. Log the relevant values on the first and second iterations at the same program line.

Use KUKA Xpert for the exact system-variable definitions of the installed release. Similar names can refer to commanded, measured or transformed positions, and release details matter.

What advance run changes

KRL prepares upcoming motion ahead of physical execution so paths can blend. Non-motion statements may be evaluated while the robot is completing an earlier move. A deliberate synchronization point can test this hypothesis. If the mismatch disappears, the issue is execution timing or approximation—not automatically bad mastering.

Check the frame chain before the math

Confirm tool and base selections are not changed by an inline form, subroutine, interrupt or submit interpreter between iterations. Search for assignments to tool, base and frame variables, including indirect references.

Compare values in one declared frame. Mixing flange-referenced and tool-referenced data can resemble a repeated angular offset. Keep Euler conventions explicit; near orientation singularities, large displayed angle changes can represent a smaller physical change.

Turn the finding into a robust interface

If a PLC consumes robot coordinates, specify whether values are commanded or measured, which frames are active and what event makes the sample valid. Capture only after the robot and PLC agree on state, then copy the complete set atomically before acknowledging it.

Broader robotics investment is accelerating, as discussed in the related report on Delta’s robotics and semiconductor automation push. New platforms do not remove the need for clear frame and sampling contracts.

Our view: a pose mismatch is an evidence problem before it is a calibration problem. Freeze execution, identify the frame and compare commanded with measured data before changing persistent values.

Questions engineers ask before making the change

Can advance run change when an assignment is evaluated?

Yes. Upcoming instructions can be processed ahead of physical motion. Use the documented synchronization behavior for the installed KSS release.

Does a different Euler display prove wrong motion?

No. Equivalent orientations and singularities can produce unintuitive angles. Compare the physical pose and reference frame.

Should the tool be recalibrated first?

No. Verify timing, approximation, active frames and variable semantics first, or the compensation can spread to healthy programs.