Precision, Accuracy, and Repeatability Defined
How do Standards Measure Robot Performance?
Why Repeatability Outpaces Accuracy?
Heat, Load, and Real-World Drift
Closing the Gap Through Calibration
Conclusion
References and Further Reading
Industrial robots perform tasks like welding car bodies, drilling aircraft panels, and placing tiny electronic parts. Each task relies on how accurately the robot's tool can reach the intended spot. Engineers describe this behavior with three related terms, precision, accuracy, and repeatability, which people often mix up. As manufacturers increasingly rely on digital models to program robots, these concepts become more crucial.
Image Credit: Gorodenkoff/Shutterstock
The programs direct the robots to coordinates they have never physically visited, so their ability to reach a computed point is very important. This article explains how these three terms differ, how standards measure them, why robots are better at repeating motions than at hitting exact targets, and how calibration improves this accuracy.1
Precision, Accuracy, and Repeatability Defined
Precision describes how tightly a set of results clusters together, regardless of where that cluster happens to sit. Accuracy means how close a result is to the true target. In robotics, repeatability is used to describe precision. It measures how consistently a tool returns to a position it was taught. Absolute accuracy measures how close the tool gets to the commanded position in a coordinate system.1
A dartboard offers a helpful picture. A player who lands every dart in a tight group near the outer ring shows high repeatability and poor accuracy. A player whose darts scatter loosely around the bullseye shows the reverse. Most industrial robots behave like the first player. Their tool returns to the same spot again and again, and that spot may sit several millimeters from the intended location.1
Which property matters most depends on how the robot is programmed. When a technician uses a handheld teach pendant to move the robot to each point, it only needs to return to those taught spots. In this case, repeatability is what matters. For tasks programmed offline, like drilling and laser cutting, the robot must reach specific coordinates. Here, accuracy is the key factor.2
The international standard ISO 9283 sets out test methods for industrial robot performance. It covers accuracy, pose repeatability, path accuracy, and related measures. For repeatability, the robot visits the test position 30 times and approaches it from the same direction each time. Engineers then calculate the radius of a sphere that contains all the measured points. A smaller sphere signals a more repeatable robot.2
In terms of accuracy, testers direct the robot to specific positions and then check how closely it follows those commands. The distance between where the robot ends up and where it was told to go is its accuracy score. Manufacturers usually do not share this accuracy figure, which can be a few millimeters for an uncalibrated arm. Repeatability for the same machines typically measures only a few hundredths of a millimeter, a difference of roughly two orders of magnitude.1
Why Repeatability Outpaces Accuracy?
The gap occurs because a robot may not know its exact location. Its control system uses a formula to map joint angles to the tool's position. If the actual arm differs slightly from this formula, it will always make the same calculation error. This means the robot can repeat its movements accurately but may not hit the correct target.1
Researchers categorize these errors into two types. Geometric errors include link lengths that differ from their intended values, joint axes that are not perfectly aligned, and joint positions that are misaligned. Non-geometric errors include bending due to weight, the effect of gravity, gear looseness, friction, wear, and sensor limitations. Studies show that about 90% of positioning inaccuracies stem from geometric errors.1
Heat, Load, and Real-World Drift
Temperature can cause problems in robotic arms. Motors and gear friction warm the arm during operation, and the links expand as they heat. A study published in the International Journal of Automation Technology found that the robot's parts experience more thermal changes than machine tools, making heat a major source of error throughout its workspace. Since heat builds up slowly, the resulting error can increase over the course of a shift.3
Welding illustrates how these temperature changes affect accuracy. A recent report in Advances in Mechanical Engineering and Transport shows that a temperature rise of 10-30 °C in the welding area can shift the tool path by 0.1-0.5 mm, depending on the arm length. The study estimated that under normal conditions, the total positioning error could be around 0.24 mm, while during intense welding, it could reach about 0.50 mm. Heat is one of the main causes of these errors.4
Closing the Gap Through Calibration
Calibration corrects the controller's model to match the physical robot's calculations. Engineers first build a model containing the parameters they wish to identify. They then measure the actual tool position across multiple poses using laser trackers, vision systems, ballbars, or coordinate measuring machines. Software estimates parameter errors from these measurements, and the corrected values are finally fed into the controller.1
Saving this for later? Download a PDF here.
Published results show major improvements in robot positioning. A Springer Nature study used a low-cost draw-wire encoder on a six-axis robot. This reduced the largest position error between different arm setups from 3.3 mm to 0.8 mm. Another work published in MDPI Sensors refined an already accurate robot using a laser tracker, reducing its mean error from 75.4 to 60.1 μm and demonstrating that gains persist at finer scales.1,5
A corrected model effectively manages geometric errors. During operation, heat-driven drift and bending from payloads change and require separate solutions. For demanding tasks, it is important to combine calibration with temperature compensation, machine vision, and adaptive seam tracking.4
Conclusion
Robot accuracy, precision, and repeatability describe different aspects of a robot's performance. Repeatability measures how consistently a robot returns to a taught point, and accuracy reflects how faithfully its internal model matches the physical arm. Errors in the model and heat-shape accuracy significantly impact performance, making repeatability figures less informative on their own.1
Calibration and compensation allow engineers to bring accuracy close to repeatability, and that shift opens the door to precise, offline-programmed work. Robots that once relied entirely on taught points can now drill, cut, and assemble parts from digital instructions with confidence. As factories lean further on digital models, the ability to hit a computed target will define how far robotic automation can reach.4
References and Further Reading
- Boschetti, G. et al. (2024). A novel step-by-step procedure for the kinematic calibration of robots using a single draw-wire encoder. Int J Adv Manuf Technol, 131, 4129–4147 (2024). DOI:10.1007/s00170-024-13219-1. https://link.springer.com/article/10.1007/s00170-024-13219-1
- Robot Accuracy vs Repeatability. Association for Advancing Automation. https://www.automate.org/glossary/robot-accuracy-vs-repeatability
- S. Ibaraki & K. Kawano. (2023). On Thermal Positioning Error of a Planar Robot Arm over Entire Workspace. Int. J. Automation Technol., 17(5), pp. 504-511. DOI:10.20965/ijat.2023.p0504. https://www.fujipress.jp/main/wp-content/themes/Fujipress/pdf_subscribed.php
- Samchuk, L. et al. (2026). Analytical review of approaches to ensuring positioning accuracy of industrial robots in welding operations. Advances In Mechanical Engineering and Transport, 1(26). DOI:10.36910/2zmncs77. https://eforum.lntu.edu.ua/index.php/jurnal-mbf/en/article/view/2234
- Khanesar, M. A. et al. (2023). Precision Denavit–Hartenberg Parameter Calibration for Industrial Robots Using a Laser Tracker System and Intelligent Optimization Approaches. Sensors, 23(12), 5368. DOI:10.3390/s23125368. https://www.mdpi.com/1424-8220/23/12/5368
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.