Robot performance
A robot's repeatability is not its accuracy: the measured gap
A robot arm rated for tight repeatability was off by 1.7730 mm in absolute accuracy before calibration, many times its repeatability spec. Calibration then cut that error to 0.1041 mm.
Repeatability and accuracy are different things, and buyers routinely confuse them. Repeatability is how closely a robot returns to the same taught point; accuracy is how close it gets to a commanded point it was never taught. The datasheet quotes the first, which is far better than the second.
This page traces the measured before-and-after figures to their study and draws the distinction that a repeatability spec hides.
Data covers Measured positioning-accuracy study of an industrial arm, before and after calibration. Last reviewed by a human editor before publication.
The figures and where they come from
Each figure is rated for how safely you can cite it today. Ratings judge current usability, not whether a number was ever correct.
| Figure | What it is | Source | Citation Confidence | Notes |
|---|---|---|---|---|
| 1.7730 mm | Mean positioning error, before calibration | [A] | Medium | Absolute positioning accuracy of the arm before calibration, far worse than a repeatability spec would suggest. Accuracy and repeatability are not the same number. |
| 0.1041 mm | Mean positioning error, after calibration | [A] | Medium | After kinematic calibration and neural-network compensation, the mean error fell to 0.1041 mm, about a 94% reduction. |
| 0.8497 mm | Average distance accuracy, before | [A] | Medium | A second accuracy measure, distance accuracy, before calibration. |
| 0.0493 mm | Average distance accuracy, after | [A] | Medium | The same measure after calibration, showing the large improvement calibration delivers. |
| 2.0140 to 0.2392 mm | Maximum distance error, before to after | [A] | Medium | The worst-case distance error fell from 2.0140 mm to 0.2392 mm with calibration. |
Why the numbers disagree
The confusion is baked into the datasheet. A robot's repeatability, how closely it returns to a taught point, is often a few hundredths of a millimeter and is the number vendors advertise. Its absolute accuracy, how close it gets to a commanded point in space it was never taught, is a different and much worse number, because it depends on the robot's model of its own geometry. The measured arm here was off by 1.7730 mm on accuracy while its repeatability was far tighter.
That gap matters whenever a task needs the robot to go to a computed point rather than a taught one, as in offline programming, vision-guided work, or CAD-driven paths. In those applications repeatability is not enough; absolute accuracy is what the robot actually delivers, and the datasheet does not quote it.
Calibration closes much of the gap but does not erase the distinction. The study cut the mean error from 1.7730 mm to 0.1041 mm with kinematic calibration and neural-network compensation, a large improvement. It is one arm in one study, so the exact numbers are specific, but the lesson, accuracy starts far from repeatability and calibration is how you recover it, is general.
How to cite these figures
Never quote a repeatability spec as if it were accuracy. They are different measures, and accuracy is much worse out of the box.
If a task uses computed points, offline programming, or vision, plan for absolute accuracy and calibration, not the datasheet repeatability.
Cite the measured improvement as an example of what calibration recovers, from about 1.77 mm to 0.10 mm here, while noting it is one arm in one study.
Where people go wrong
Reading the repeatability spec as the accuracy the robot delivers. Absolute accuracy is a different, worse number.
Assuming offline-programmed or vision-guided points land as tightly as taught points. They depend on accuracy, not repeatability.
Treating the exact millimeter figures as universal. They are one arm in one study; the accuracy-versus-repeatability gap is the general lesson.
How we checked
The figures come from a peer-reviewed positioning-accuracy study accessed through PubMed Central. We retrieved it and confirmed the before-and-after positioning and distance accuracy figures, including the 1.7730 mm and 0.1041 mm mean errors, in its text.
We use the study to illustrate a general distinction, repeatability versus absolute accuracy, while stating that the exact values belong to the tested arm. The distinction holds broadly; the numbers are one data point.
We did not find a broad, independent distribution of out-of-box absolute accuracy across industrial arms. A single well-documented study is the anchor, and the page is explicit about that scope.
Full source list
Primary sources, with live links. Every figure above traces to one of these.
- [A]PMC (PubMed Central), Sensors (MDPI)2020
Nguyen et al., "Absolute Positioning Accuracy Improvement in an Industrial Robot" (Sensors), PMC
https://pmc.ncbi.nlm.nih.gov/articles/PMC7471972/
Common questions
- Is a robot's repeatability the same as its accuracy?
- No. Repeatability is how closely a robot returns to a taught point, often a few hundredths of a millimeter. Absolute accuracy, how close it gets to a commanded point it was never taught, is much worse. One measured arm was off 1.7730 mm on accuracy before calibration.
- Why does the difference matter?
- Because tasks that use computed points, offline programming, or vision guidance depend on absolute accuracy, not repeatability. The datasheet quotes repeatability, so it overstates how precisely the robot hits untaught points.
- Does calibration fix it?
- Largely. In the study, kinematic calibration and neural-network compensation cut the mean error from 1.7730 mm to 0.1041 mm, about 94%. It closes most of the gap but does not make accuracy equal to repeatability.
- Do these numbers apply to every robot?
- No. They are one arm in one study. The general lesson, that accuracy starts far worse than repeatability and calibration recovers much of it, transfers; the exact millimeters do not.
More data, traced to source
- Robot repeatability under load: the spec versus the measurement
A robot arm rated at plus or minus 0.1 mm repeatability measured worse under load, with the spread reaching about 0.2 mm at 16 kg. The datasheet is one number; the measured accuracy depends on what the robot is carrying.
- Robot reliability numbers: the vendor claims and the one independent study
Manufacturers advertise robot uptime in the high nineties and mean time between failures in the tens of thousands of hours. The one independent study of more than 400 factories found a robot cell is reliable 88 percent of the time, with 87 minutes between failures.