Motion performance is a key differentiator in robotics applications. By introducing GoFa™ Ultra Accuracy, ABB advances path and absolute positioning accuracy beyond existing limits. The newly created potential for robotic automation solutions allows high-precision applications that were previously impossible.

Subscribe to ABB innovation and technology news

_{Arne Wahrburg, arne.wahrburg@de.abb.com

ABB Corporate Research

Mannheim, Germany}

_{Stig Moberg, stig.moberg@se.abb.com

Henrik Nilsson, henrik.x.nilsson@se.abb.com

Tomas Groth, tomas.groth@se.abb.com

Mikael Norrlöf, mikael.norrlof@se.abb.com

Sven Hanssen, sven.hanssen@se.abb.com

ABB Robotics & Discrete Automation

Västerås, Sweden}

_{Richard Roberts, richard.roberts@de.abb.com

Anton Sodja, anton.sodja@de.abb.com

Minda Xia, minda.xia@de.abb.com

ABB Robotics & Discrete Automation

Gilching, Germany}

_{Silke Klose

Former ABB employee}

The trend toward product miniaturization, the ability to program offline yet deploy in real-world scenarios, as well as the need to meet ever tighter application requirements in robot automation is essential for a variety of industrial processes that require accuracy. In electronics manufacturing, high component density, multiple layering and small components require care in assembly. In gluing applications for example, a robot typically handles a work piece as large as a smart phone, a tablet or a laptop. The gluing nozzle usually has an inner diameter of 0.6 mm; the glue profile is shaped like a semicircle (height of 0.5 ~ 1 mm and a width of 0.8 ~ 1.8 mm). Thus, gluing as well as tasks such as cutting, sealing, additive manufacturing, surface inspection, and metrology require robots and robot controllers to achieve very high accuracy, both in terms of absolute positioning and path tracking.

Enter, GoFa Ultra Accuracy which is built on ABB’s TrueMove® concept, providing control for high path accuracy for robots since the 1990s, and on Absolute Accuracy, an ABB calibration option that enables high absolute positioning accuracy. Experimental results demonstrate that GoFa Ultra Accuracy achieves path errors < 0.1 mm, the average width of a human hair. Currently, ABB is integrating this feature into the GoFa™ collaborative robot line to push the limits of robot accuracy further than previously possible.

What is accuracy anyway?

The term accuracy conjures up exactitude and determinism for most people. And yet, ironically, accuracy can often be a somewhat fuzzy term in many technical situations despite the precise definitions found in DIN EN ISO 9283 [1]. In the context of this article, it is critical to distinguish between repeatability, absolute accuracy, and path accuracy, which are often confused. While repeatability can be interpreted as the ability to hit the same spot, or target repeatedly, accuracy refers to how precisely a given target can be hit. This difference is easily understood and visualized →01.

Regarding robotic applications, industrial robots typically excel at very high repeatability, between a range of 0.01 mm and 0.1 mm depending on the model – a key to their success in automation processes. For example, GoFa achieves a repeatability of 0.02 mm – the best among collaborative robots. However, without special capabilities that will be detailed throughout this article, the accuracy of such robotic manipulators is not as high as for repeatability.

Delving further, there is a need to distinguish between absolute- and path accuracy. As visualized in →02, absolute accuracy refers to the capability of reaching programmed targets in space accurately – a key property for robotic tasks such as spot welding, where ever a path must be followed without touching the work piece. In contrast, path accuracy refers to how accurately a prescribed geometry that connects two targets (ie, a path) can be tracked – a key property for continuous robotic tasks such as gluing, sealing, or 3D printing. In reality, industrial robots need to complete automation tasks with high absolute– and path accuracy.

State-of-the-art accuracy and the challenges

Nowadays, thanks to the application of high-resolution position sensing on the gearbox input-side and advanced model-based controls, articulated robots achieve very high repeatability (eg, ± 0.01 mm for an ABB IRB 1100-4/0.58).

When it comes to absolute positioning accuracy however, the difference between a virtual, or ideal, robot and a real robot can be between 8 and 15 mm (depending on the robot model). This discrepancy is due to the resultant errors generated during establishment of the zero positions of position sensors in response to the kinematic tolerances in link lengths and mechanical assembly, as well as to deflections in the robot structure due to load [2,3].

Deflections due to load constitute the structural compliance within the gearboxes, as well as in the bearings and links. Thankfully, algorithms are now available that can identify such modeling errors (referred to as absolute accuracy calibration). With their absolute accuracy option, ABB currently offers a solution that not only identifies the modeling error in the individual robot at hand, but also employs the calibrated model to compensate for the negative effects described above. As a result, maximum absolute positioning errors can typically be reduced to < 1 mm, although exact error values are dependent on the individual robot [4].

Compliance within the mechanical structure poses a major challenge in terms of path accuracy. Not only does compliance induce flexible modes that result in vibrations and oscillations once excited, but the resultant oscillations cannot be traced down to an individual robot joint generally but affect the entire arm due to inertial couplings [5]. Moreover, the reduction gears that are used in each joint to transform the high speed and low torque output of the electrical motors into reasonable ranges for robotic manipulation induce additional challenges. Such specific effects include periodic transmission errors, non-linear stiffness characteristics, hysteresis [6,7], and friction [8,9]. To tackle those challenges, ABB developed the TrueMove concept in the 1990s: This ABB hallmark product ensures that high path accuracy is achieved by means of advanced motion control [10].

Introducing GoFa Ultra Accuracy feature

Thanks to the absolute accuracy option and TrueMove, the starting point for taking both absolute and path accuracy beyond the level currently achieved is simple: increase the breadth of information available and used to eliminate disturbances and diminish error. But how can this be accomplished?

To achieve both accuracy objectives, it is crucial to recognize that the gearbox is one of the major sources of modeling errors and inaccuracies. These modeling uncertainties directly impact accuracy, just as in conventional robot control systems, where only the position of the motor (ie, at gearbox input side) is measured and used for feedback controls [11]. The basic, yet innovative, idea behind GoFa Ultra Accuracy is not only to use such conventional measurements, but to also employ arm-side information in the servo loops →03. This added functionality improves the controller’s ability to reject disturbances and uncertainties introduced by the gearboxes, as well as disturbances that originate externally (such as from process forces). At first glance, use of measurements at the gearbox output (ie, “after” one of the main sources of uncertainty) would seem to be a natural way to improve accuracy compared to use of conventional motor-side instrumentation. Despite this, application of such a concept results in the physical distribution of actuation and sensing with dynamic and, or compliant elements in between. Such non-collocated problems, however pose significant challenges to control systems in general and control of robots specifically →04, which have, until recently, seemed intractable [11,12,13]. New promising progress from academia brings a solution within the realm of possibility [14,15,16]. By developing an improved understanding of the gearboxes and measurement systems, and by integrating such capacity with ABB’s advanced model-based controls, ABB’s GoFa Ultra Accuracy runs stably with high performance within the robot’s complete workspace, covering its full payload- and speed-range despite the challenges arising from non-collocation →04.

Absolute accuracy results

The process of calibrating a robot to achieve high absolute accuracy relies on moving the robot to several programmed calibration targets. The nominal position of the robot tool center point (TCP) is compared to the actual position measured by an appropriate 3D measurement system (typically a laser tracker). The actual calibration step then aims to minimize the difference between the measured TCP positions and the TCP positions of the calibrated model. Once a calibrated model has been obtained, the model is used to calculate compensation parameters that are then applied to each Cartesian position programmed by the user. To validate the calibration and compensation parameters the individual robot is moved to a set of 50 additional targets with compensation activated. While an absolute positioning error of 0.23 mm (average) and 0.47 mm (maximum) is achieved on an ABB GoFa™ 5 robot running motion control with conventional motor-side position sensing, superior absolute accuracy is achieved by performing calibration on the same robot while GoFa Ultra Accuracy is activated →05: In the latter case, absolute positioning errors are reduced by 50 percent, to only 0.10 mm, the width of a human hair, and from a maximum error of only 0.23 mm – phenomenal results →05.

Path accuracy results

To quantify path accuracy, ABB focused on the non-trivial task of separating absolute positioning errors from path errors. To isolate the latter, the motion along small, programmed shapes was performed, where the measurements of the TCP position, obtained by a high-performance laser tracker system (Leica AT 960), are expressed relative to the center of the respective shape. In this way, the impact of absolute positioning errors is minimized and an evaluation of the path errors can be conducted.

The two dimensional (2D) path accuracy evaluation results for a GoFa collaborative robot compare Ultra Accuracy with conventional instrumentation (ie, motor-side position only) →06. Here, it becomes apparent during execution of a circular motion that a very small radius of only 1 mm poses a significant challenge to serial chain manipulators: in other words, the limits of their capabilities, in terms of path accuracy, are reached. The addition of the GoFa Ultra Accuracy feature provides for a notable improvement in path accuracy.

In a second experiment to examine path accuracy in 2D and three dimensions (3D), a small rectangular path (10 mm x 2 mm) was executed. As →07 shows, the path accuracy improvement achieved by implementing GoFa Ultra Accuracy in 3D was significant: Path errors were achieved below 0.1 mm for speeds up to 80 mm/s. The 2D path errors were even smaller.

Applied example results

As mentioned above, absolute positioning errors and path errors are superimposed in actual applications. Ultimately though, what matters is that task completion is ensured at the required quality specifications. In terms of robot motion accuracy, this boils down to moving the TCP along a path that is specified along a physical work piece. This process requires the additional calibration of the tool (to precisely determine the position of the TCP relative to the robot flange) and the work object (to precisely determine the position of the work piece relative to the robot base). Those two application-related calibration steps can be automated, with easy-to-use functionality that are available in ABB’s RobotStudio® Machining PowerPac [17]. By using a robot calibrated for absolute accuracy, running GoFa Ultra Accuracy to push the limits of both absolute positioning and path accuracy, and conducting tool and work object calibration as described above, it is possible to achieve a vastly improved level of offline programming: Complex paths can be programmed in ABB’s RobotStudio® and executed on real robots with very tight tolerances with respect to the physical work pieces.

ABB’s OmniCore Challenge, launched in June 2024, uses the OmniCore™ automation platform to achieve a precision of 0.6 mm with multiple robots (IRB1300) at maximum speeds of 1600 mm/s, thanks to MultiMove®. This new challenge is based on a prominent showcase of ABB’s superior motion control performance enabled by QuickMove® for short cycle times and TrueMove® for path accuracy. Significantly, ABB’s R&D revisited motion control performance at the microscale with GoFa Ultra Accuracy →08. The spacing between the “cans” has been reduced to 2.2 mm, while the diameter of the touch probe moving along the path is 2.0 mm – leaving only a 0.1 mm margin on either side →08. The measurement results obtained by a laser tracker →09 reveal that the path error using conventional motion control exceeds the available tolerance margin. The results show the signed distance of the probe head to any of the cans →09. Ideally (ie, if the TCP perfectly follows the programmed path), the distance should always be 0.1 mm as indicated by the dashed black line in →09. Using conventional control, a signed distance smaller than 0 mm occurs – the probe touches one of the cans. In contrast, the error achieved using GoFa Ultra Accuracy remains well within 0.1 mm on either side of the path. This remarkable result means that the probe can be moved in between the cans without touching them as the signed distance in →09 never becomes negative. Moreover, the same holds true even if the TCP is re-oriented while in motion.

The future of robot accuracy is at hand

Based on the phenomenal real-world test results, ABB’s exciting new feature GoFa Ultra Accuracy is being developed for release in ABB’s GoFa collaborative robot line. By improving robot motion performance, ABB can offer a high accuracy robot to the market, thereby opening the door for higher-precision applications. In this way, ABB helps manufacturers meet superior operational performance, improve productivity and effectively create products to exact specifications.

_{Reference

[1] DIN EN ISO 9283:1999-05 – Manipulating industrial robots – Performance criteria and related test methods; (ISO 9283:1998); German version EN ISO 9283:1998, [Online]. Available: https://www.dinmedia.de/de/norm/din-en-iso-9283/4534078 [Accessed: May 16, 2024.]}

_{[2] A. Nubiola and I. A. Bonev, “Absolute calibration of an ABB IRB 1600 robot using a laser tracker”, In: Robotics and Computer-Integrated Manufacturing, vol. 29, no. 1, 2013, pp. 236 – 245.}

_{[3] Z. Li, et al., “An Overview of Calibration Technology of Industrial Robots”, In: IEEE/CAA Journal of Automatica Sinica, vol. 8, no. 1, 2021, pp. 23 – 36.}

_{[4] ABB Absolute Accuracy datasheet, [Online]. Available: https://library.abb.com/d/PR10072EN_R6, 2011 [Accessed: May 16, 2024.]}

_{[5] S. Moberg, “Modeling and control of flexible manipulators”. Ph.D. dissertation, Dept. Elect. Eng., Linköping Uni., Linköping, SE, 2010.}

_{[6] T. D. Tuttle and W. P. Seering: “Nonlinear Model of a Harmonic Drive Gear Transmission”, In: IEEE Transactions on Robotics and Automation, vol. 12, no. 3, 1996, pp. 368 – 374.}

_{[7] P. Mesmer, et al., “Modeling and Identification of Hysteresis in Robot Joints with Cycloidal Drives”, in Proceedings of the IEEE International Conference on Advanced Motion Control, 2022, pp. 358 – 363.}

_{[8] A. Wahrburg, et al., “Modeling Speed-, Load-, and Position-Dependent Friction Effects in Strain Wave Gears”, in Proceedings of the IEEE International Conference on Robotics and Automation, 2018, pp. 2,095 – 2,102.}

_{[9] M. Iskandar and S. Wolf, “Dynamic friction model with thermal and load dependency: modeling, compensation, and external force estimation”, in Proceedings of the IEEE International Conference on Robotics and Automation, 2019, pp. 7,367 – 7,373.}

_{[10] M. Björkman, et al., “A new concept for motion control of industrial robots”, in Proceedings of the IFAC World Congress, 2008, pp. 15,714 – 15,715.}

_{[11] A. De Luca and W. Book, “Robots with Flexible Elements”, in Springer Handbook of Robotics, B. Siciliano and O. Khathib, Eds. Berlin, Germany: Springer, 2008, ch. 13, pp. 243 – 282.}

_{[12] P. A. Chodavardapu and M. W. Spong, “On noncollocated single control of a flexible link”, in Proceedings of the IEEE Conference on Robotics and Automation, 1996, pp. 1,101 – 1,106.}

_{[13] J.-H. Ryu, et al., “Control of a Flexible Manipulator with Noncollocated Feedback: Time-Domain Passivity Approach”, in IEEE Transactions on Robotics, vol. 20, no. 4, 2004, pp. 776 – 780.}

_{[14] S. Klose and A. Wahrburg: “A feedback control scheme for improving path accuracy of industrial manipulators based on gearbox output sensing”, in Proceedings of the IEEE International Conference on Advanced Motion Control, 2022, pp. 364 – 369.}

_{[15] P. Mesmer, et al., “Robust design of independent joint control of industrial robots with secondary encoders”, in Robotics and Computer-Integrated Manufacturing, vol. 73, 2022.}

_{[16] M. Keppler, et al., “From underactuation to quasi-full actuation: Aiming at a unifying control framework for articulated soft robots”, in International Journal of Robust and Nonlinear Control, vol. 32, no. 9, 2022, pp. 5,453 – 5,484.}

_{[17] ABB RobotStudio Machining PowerPac datasheet, ABB Robotics, 2015, [Online]. Available: https://search.abb.com/library/Download.aspx?DocumentID=9AKK106354A3615&LanguageCode=en&DocumentPartId=&Action=Launch}

Select region & language

Down to a human hair