Efficiently Improving and Quantifying Robot Accuracy In Situ

Geraldine Cheok; Joe Falco; Karl Van Wyk

REVIEW 3 major objections 3 minor 26 references

Efficiently Improving and Quantifying Robot Accuracy In Situ

T0 review · 3 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A low-cost motion capture system plus Bayesian inference can calibrate a robot's joint offsets in place, matching laser-tracker results and cutting average positioning error roughly four-fold.

desk verdict A useful, honest robot-calibration paper whose central practical claim holds in the dexterous workspace but whose 'throughout its entire workspace' headline outruns its evidence. read the letter →

arxiv 1908.07273 v1 pith:L3UPXWX2 submitted 2019-08-20 cs.RO

Karl Van Wyk , Joe Falco , Geraldine Cheok This is my paper · ORCID

classification cs.RO

keywords robotcalibrationzero-offsetremasteringBayesianinferenceMetropolissamplingmotioncapturelasertrackerpositioningaccuracyinsitu

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper aims to show that the expensive laser tracker traditionally used for robot calibration can be replaced by a low-cost, portable motion capture system without losing calibration quality. The target is the set of joint zero-offsets: small angular corrections added to the robot's measured joint angles, which the paper says cause over 90 percent of typical Cartesian positioning error. Bayesian inference on measured marker positions yields both optimal offset values and their uncertainties, and the offsets agree with those computed from laser tracker data. The sympathetic reading is that in situ, software-only remastering with commodity hardware gives roughly a four-fold improvement in average positioning accuracy across the robot's workspace, at a fraction of the cost and a fraction of the data-collection time.

What carries the argument

The load-bearing object is the zero-offset vector $\Theta_2 = \{\delta\theta_1,\ldots,\delta\theta_7\}$: seven small angular constants added to the commanded joint angles inside the forward-kinematics model (Eq. 9). Around this vector the paper builds a joint posterior $p(\Theta,\sigma^2 \mid \text{data}) \propto \sigma^{-3n-2}\exp(-E/(2\sigma^2))$ (Eq. 15), where $E$ is the sum of squared residuals between measured marker positions and the robot's predicted marker position under an isotropic Gaussian error model (Eq. 10). This posterior is sampled with the Metropolis algorithm, producing most-likely offsets together with their standard deviations. A six-parameter correctional transform $T^*(\Theta_1)$ is optimized simultaneously to absorb errors from the initial three-point registration between the reference sensor and the robot base, which is what lets the low-cost sensor's unknown low-frequency distortion be compensated rather than leak into the offsets.

What would settle it

Compute the residuals $N p_{i,\text{ref}} - N p_{i,\text{robot}}$ after fitting the fourteen parameters and check whether their per-axis means are zero and whether their variance changes with joint configuration, arm reach, temperature, or marker position in the motion-capture volume. If the residuals show a nonzero mean in any axis for some region of the workspace, or if their spread grows systematically with reach, the isotropic Gaussian model in Eq. (10) is false and the reported offset uncertainties are overconfident.

Watch

Extended reading notes

Core claim

The paper's central claim is that a robot's joint zero-offsets can be estimated in situ from a low-cost motion capture system using Bayesian inference, and that the resulting offsets and their uncertainties are essentially the same as those obtained from a benchmark laser tracker. On a held-out validation set, applying the motion-capture-derived offsets to a seven-degree-of-freedom arm reduced mean post-registration absolute error from 4.715 mm to 1.838 mm, while the laser-tracker-derived offsets reduced it from 4.397 mm to 1.509 mm. The paper also proposes three accuracy metrics: relative Cartesian accuracy across joint configurations, post-registration absolute position error, and a theoretical uncertainty-based error computed by sampling the validated robot model. The last metric, evaluated over 1000 configurations spanning the full workspace, supports the claim of a four-fold improvement in absolute positioning accuracy after zero-offset remastering.

Load-bearing premise

All uncertainty numbers and the four-fold improvement claim depend on the assumption that, after fitting six registration corrections and seven zero-offsets, every leftover difference between measured and modeled marker positions is independent, symmetric bell-shaped noise with the same spread in every direction.

Editorial extensions

If this is right

Zero-offset remastering can be done in situ with a pre-calibrated motion capture unit, about 20 minutes of data per pose set, and no high-accuracy marker plate fixed to the robot base.
Because the recovered offsets from the motion capture system and the laser tracker agree within their reported uncertainties, the low-cost sensor is sufficient for this calibration task.
After remastering, the remaining positioning error is dominated by non-zero-offset sources such as unmodeled link parameters and structural deflection, since zero-offset uncertainty explains only about 15 percent of the residual error.
The same posterior machinery transfers to any robot with known forward and analytical inverse kinematics, provided a marker can be mounted on the tool.
Across the full workspace sampled with 1000 configurations, the theoretical average absolute error falls from above 5 mm with visual Vernier-scale remastering to below 1.5 mm with Bayesian remastering.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

An extension the authors leave implicit is using the posterior to choose the next measurement poses adaptively; a testable variant is selecting poses that most reduce posterior variance in the offset estimates.
The equivalence claim rests on a single arm and a single motion capture unit; a stronger test would place the same low-cost unit in several positions and check whether the recovered offsets stay within the reported standard deviations.
Because the motion capture system's absolute accuracy is unknown, the six-parameter registration correction may be absorbing smooth optical distortion; if so, the same pipeline with a different sensor placement could shift the offsets by more than the reported uncertainty.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Worth a look if you care about practical robot calibration. The genuinely new thing here is the combination: low-cost motion capture as the reference sensor, Bayesian MCMC over the zero-offsets plus a six-parameter registration correction, and analytical inverse kinematics that command every joint solution for each Cartesian pose. The held-out validation is the backbone of the paper and it mostly holds up. Applying the motion-capture-derived offsets to a fresh set of 156 configurations gets mean post-registration error down from 4.7 mm to 1.8 mm, with the laser tracker at 1.5 mm. That is real, direct evidence that the cheap sensor route works at the level of the dexterous workspace where data were collected. The agreement between the two sensors' offset estimates, with the caveat that J1 and J7 differ, is also genuinely encouraging for the method.

The soft spot is the headline, which outruns the evidence. The 'four-fold throughout its workspace' claim comes from Table VI, a model extrapolation to 1000 random joint configurations, many explicitly outside the dexterous volume where all training and validation data live. The model is the isotropic-Gaussian lumped error plus seven zero-offsets and six registration terms. That is fine for interpolation inside the data support, but using it to predict accuracy in uncharted workspace regions assumes systematic effects like link deflection and uncorrected D-H parameters stay Gaussian and zero-mean far away. The reader's stress-test note is on target here. The paper would be more accurate claiming three-fold improvement in the dexterous workspace, with the full-workspace numbers as a model-based projection, not an experimental result.

The in-sample theoretical tables (V and VI) are also partially circular—they use the fitted posterior to generate the after-errors—but the authors themselves seem aware, and the validation section is the honest check. That check covers only the dexterous volume, so the circularity concern for the full-workspace claim stands.

Minor things: no code or data released, which slows independent reproduction; the Gaussian error model gets no residual diagnostics; and the sample sizes (173 and 156 configurations) are modest but acceptable for the claims made within the sampled volume. The citation pattern looks fine—self-citations are to NIST registration work that is directly relevant.

Who is this for? Practitioners in SME settings who want an inexpensive remastering procedure, and calibration researchers looking for a low-cost reference sensor comparison. It deserves a real peer review; the core idea is sound and the validation is honest, but the authors should be pushed to either soften the full-workspace claim or validate it with data.

Referee Report

3 major / 3 minor

Summary. The paper proposes a robot zero-offset calibration method that combines a low-cost OptiTrack motion capture system with Bayesian inference to estimate seven joint zero-offsets and six registration-correction parameters. Data are collected from a 7-DOF KUKA arm within a restricted dexterous workspace (Section III-A), and the fitted model is used to report three accuracy metrics: relative Cartesian accuracy, post-registration absolute error, and a theoretical uncertainty-based accuracy. The authors report a roughly four-fold improvement in average positioning accuracy 'throughout its workspace' based on a theoretical extrapolation to 1000 random joint configurations (Table VI), and a held-out validation dataset (Section VI, Table VIII) that shows post-registration absolute error of 1.838 mm (motion capture) versus 4.715 mm before remastering.

Significance. The core experimental result is valuable: the held-out validation in Section VI demonstrates that the low-cost motion capture system, which is far cheaper and faster than a laser tracker, produces zero-offset estimates that improve dexterous-workspace absolute accuracy from 4.715 mm to 1.838 mm, compared with 1.509 mm for the laser tracker. The paper also provides a clear, reproducible pipeline (Latin Hypercube sampling, analytical inverse kinematics, Metropolis sampling) and compares two reference sensors on the same pose set. These contributions are useful for in-situ robot remastering and for quantifying positioning accuracy. However, the headline claim of 'four-fold improvement throughout its workspace' goes beyond the experimental evidence and depends on an unvalidated model extrapolation, so the paper's significance is strongest when the claim is restricted to the tested dexterous workspace.

major comments (3)

[Section V-C / Table VI] The abstract's 'four-fold throughout its workspace' rests on Table VI, which extrapolates the model in Eq. (10) to 1000 randomly generated joint configurations that the authors themselves state lie 'well outside the dexterous workspace' (Section V-C). All training and validation data were collected in the restricted volume described in Section III-A (r between 200 mm and 500 mm, z between 400 mm and 800 mm), and the validation dataset in Section VI was generated by the same restricted sampler. Systematic errors such as link deflection, uncorrected D-H parameters, and thermal drift are configuration-dependent and need not remain zero-mean, isotropic, or Gaussian hundreds of millimeters outside the sampled region. Without measured residuals at full-workspace configurations, the 'throughout its workspace' claim is unsupported; the direct experimental evidence supports only dexterous-workspace improvement.
[Section V-B / Tables IV-VII] Tables IV, V, and VII are not independent verifications of the method: the 'after' errors are computed by applying the fitted MLE zero-offsets (Table I) and the fitted noise model (Eq. 10) to the same configurations used for fitting (Tables IV and V) or to random configurations under the same model (Tables VI and VII). The only genuinely held-out evidence is the validation dataset in Section VI, Table VIII, which indeed shows post-registration absolute error of 1.838 mm (motion capture) versus 4.715 mm before remastering. The paper should present Table VIII as the primary evidence for improvement and explicitly label Tables IV-VII as model predictions, not measurements, so that the strength of the empirical claim is not overstated.
[Eq. (10) / Section IV] The posterior in Eq. (15) and the uncertainty intervals in Tables V-VII all assume that the residual in Eq. (10) is zero-mean, isotropic, independent Gaussian noise with variance sigma^2. The paper provides no residual diagnostics to justify this distribution, such as plots of residuals versus joint configuration, marker location, or time. If systematic errors remain after fitting the zero-offsets and registration corrections, the MLE offsets will be biased and the reported standard deviations will be overconfident. The authors should add residual analysis or explicitly temper the uncertainty claims.

minor comments (3)

[Table VI] The header row spells 'Vernier' as 'Verier'; please correct this typo.
[Section II-B] The claim that the motion capture system's distance perception does not significantly distort as a function of marker location is important for the method's validity, but no quantitative check of this distortion is provided. Consider including a short experiment or reference to a prior calibration study.
[Section VI] The validation uses only the motion capture-derived zero-offsets because the robot interface could not roll back to a previous remastering state. This is a reasonable practical limitation, but it should be stated in the abstract or conclusions so that readers do not infer a direct experimental comparison of both sensors' offsets on the validation set.

Circularity Check

0 steps flagged · score 2.0 of 10

No load-bearing circularity: the held-out validation dataset provides independent support for the central viability claim; the full-workspace 'four-fold' claim is a model extrapolation, not a circular reduction.

full rationale

The paper's central viability claim is not circular. The zero-offset posterior in Eq. (15) is fitted to the optimization dataset, and the theoretical accuracy metrics in Tables V-VII are computed from that same fitted posterior, so those tables are in-sample model re-statements rather than independent predictions. However, Section VI provides a genuine held-out check: the MLE zero-offsets from motion capture data were physically applied to the robot, and post-registration absolute errors were measured with both the motion capture system and the laser tracker on a separate validation dataset (Table VIII, means 1.838 mm and 1.509 mm). This independent evidence supports the method's practical viability and the general magnitude of the claimed improvement. The 'four-fold throughout its entire workspace' statement in the abstract and Section VII relies on Table VI, which extrapolates the fitted isotropic-Gaussian model to 1000 joint configurations outside the dexterous workspace where all data were collected. This is a validity or robustness limitation, not a circularity: the model output is a genuine extrapolation rather than a restatement of the fitted parameters by definition. No load-bearing self-citation chain or imported uniqueness theorem is used; self-citations such as [20] and [23] are supporting references for standard registration techniques and are not the basis of the central result. Overall, the derivation is self-contained and the main claim has independent validation, so no significant circularity is present.

Assumptions & free parameters 3 free parameters · 6 assumptions · 0 invented entities

The method fits seven zero-offsets, six registration corrections, and one noise sigma against experimental data. No fundamentally new physical entities are introduced. The main unproven inputs are the Gaussian error assumption, the completeness of the registration correction, and the extrapolation from the dexterous workspace to the full workspace.

free parameters (3)

Zero-offsets Θ2 (δθ1..δθ7) = Mocap MLE: 0.477, -0.192, 0.139, 0.099, 0.392, -0.114, 0.936 deg
The seven joint zero-offsets are the target parameters of the calibration, jointly fitted with registration corrections and noise sigma.
Registration correction Θ1 (x,y,z,α,β,γ) = Mocap MLE: -2.096, 0.502, 3.311 mm and -0.252, -0.334, -0.960 deg
Six parameters correcting the initial three-point registration between reference sensor and robot base; fitted simultaneously with zero-offsets.
Isotropic Gaussian noise sigma σ = Mocap 0.890 mm, laser tracker 0.734 mm
Standard deviation of lumped unmodeled effects and sensor noise in Eq (10), fitted within the Bayesian model.

assumptions (6)

standard math Bayes theorem and Metropolis-Hastings sampling produce the posterior in Eq (15).
The paper invokes standard Bayesian inference and MCMC, which are accepted mathematical tools.
domain assumption The robot kinematic model in Eq (9) is exact except for zero-offsets; all other D-H parameter errors are negligible (<10%).
Invoked in Section IV and Introduction, citing literature that more than 90% of positional error is caused by zero-offsets.
domain assumption Residual errors after fitting Θ1 and Θ2 are zero-mean, isotropic, independent Gaussian (Eq 10-11).
This assumption underlies the likelihood in Eq (12) and the posterior in Eq (15); no residual diagnostics are provided.
domain assumption The six-parameter registration correction T*(Θ1) fully absorbs the erroneous three-point initial registration.
The paper fits six correctional offsets to compensate for the poor initial registration, assuming this is sufficient for unbiased zero-offset estimation.
domain assumption The Metropolis sampler converged after 1.75e5 samples and the sample mean equals the MLE.
The paper relies on trace plots and histograms in Figs 3-4 to assert convergence, without formal convergence diagnostics.
domain assumption The dexterous workspace (r=200-500 mm, z=400-800 mm) is representative enough to validate a model used to predict full-workspace accuracy.
The full-workspace four-fold improvement in Table VI is a model extrapolation beyond the validated dexterous workspace; no physical measurements outside this volume were taken.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Efficiently Improving and Quantifying Robot Accuracy In Situ." pith.science (2026). https://pith.science/paper/L3UPXWX2

@misc{pith2026190807273,
  author       = {Pith},
  title        = {Pith review of: Efficiently Improving and Quantifying Robot Accuracy In Situ},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L3UPXWX2}},
  note         = {Machine review of arXiv:1908.07273}
}

read the original abstract

The advancement of simulation-assisted robot programming, automation of high-tolerance assembly operations, and improvement of real-world performance engender a need for positionally accurate robots. Despite tight machining tolerances, good mechanical design, and careful assembly, robotic arms typically exhibit average Cartesian positioning errors of several millimeters. Fortunately, the vast majority of this error can be removed in software by proper calibration of the so-called "zero-offsets" of a robot's joints. This research developed an automated, inexpensive, highly portable, in situ calibration method that fine tunes these kinematic parameters, thereby, improving a robot's average positioning accuracy four-fold throughout its workspace. In particular, a prospective low-cost motion capture system and a benchmark laser tracker were used as reference sensors for robot calibration. Bayesian inference produced optimized zero-offset parameters alongside their uncertainty for data from both reference sensors. Relative and absolute accuracy metrics were proposed and applied for quantifying robot positioning accuracy. Uncertainty analysis of a validated, probabilistic robot model quantified the absolute positioning accuracy throughout its entire workspace. Altogether, three measures of accuracy conclusively revealed multi-fold improvement in the positioning accuracy of the robotic arm. Bayesian inference on motion capture data yielded zero-offsets and accuracy calculations comparable to those derived from laser tracker data, ultimately proving this method's viability towards robot calibration.

Figures

Figures reproduced from arXiv: 1908.07273 by the authors.

**Figure 2.** Hardware setup and placement of (a) coordinate systems for robot, [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Trace plot of zero-offsets of all robot joints during sampling process [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Converged zero-offset distributions of all robot joints during the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Robot TCPs calculated for a single joint configuration using Vernier [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 25 canonical work pages

[1]

Robot metrology and calibration systems - a market review,

F. Hidalgo and P. Brunn, “Robot metrology and calibration systems - a market review,” Industrial Robot: An International Journal , vol. 25, no. 1, 1998

work page 1998
[2]

2015: By far the highest volume ever recorded,

I. F. of Robotics, “2015: By far the highest volume ever recorded,”

work page 2015
[3]

Recent progress on programming methods for industrial robots,

Z. Pan, J. Polden, N. Larkin, S. Van Duin, and J. Norrish, “Recent progress on programming methods for industrial robots,” Robotics and Computer-Integrated Manufacturing, vol. 28, pp. 87–94, 2012

work page 2012
[4]

Comparative peg- in-hole testing of a force-based manipulation controlled robotic hand,

K. Van Wyk, M. Culleton, J. Falco, and K. Kelly, “Comparative peg- in-hole testing of a force-based manipulation controlled robotic hand,” IEEE Transactions on Robotics , in press

work page
[5]

Addressing perception uncertainty induced failure modes in robotic bin-picking,

K. N. Kaipa, A. S. Kankanhalli-Nagendra, N. B. Kumbla, S. Shriyam, S. S. Thevendria-Karthic, J. A. Marvel, and S. K. Gupta, “Addressing perception uncertainty induced failure modes in robotic bin-picking,” Robotics and Computer-Integrated Manufacturing , vol. 42, pp. 17–38, 2016

work page 2016
[6]

Dex-net 1.0: A cloud-based network of 3d objects for robust grasp planning using a multi-armed bandit model with correlated rewards,

J. Mahler, F. T. Pokorny, B. Hou, M. Roderick, M. Laskey, M. Aubry, K. Kohlhoff, T. Kr ¨oger, J. Kuffner, and K. Goldberg, “Dex-net 1.0: A cloud-based network of 3d objects for robust grasp planning using a multi-armed bandit model with correlated rewards,” in Robotics and Automation (ICRA), 2016 IEEE International Conference on . IEEE, 2016, pp. 1957–1964

work page 2016
[7]

Kinematic noise propagation and grasp quality evaluation,

S. Liu and S. Carpin, “Kinematic noise propagation and grasp quality evaluation,” inAutomation Science and Engineering (CASE), 2016 IEEE International Conference on . IEEE, 2016, pp. 1177–1183

work page 2016
[8]

A technique to calibrate industrial robots with experimental veriﬁcation,

R. Judd and A. Knasinski, “A technique to calibrate industrial robots with experimental veriﬁcation,” IEEE Trans. Robotics and Automation , vol. 6, no. 1, pp. 590–596, 1990

work page 1990

Show all 26 references

[9]

Robot calibration issues: accuracy, repeatability and calibration,

K. Conrad, P. Shiakolas, and T. Yih, “Robot calibration issues: accuracy, repeatability and calibration,” Proc. 8th Medit. Conf. Control and Autom., 2000

2000
[10]

Developing an efﬁcient calibration system for joint offset of industrial robots,

B. Gao, Y . Liu, and Y . Shen, “Developing an efﬁcient calibration system for joint offset of industrial robots,” Journal of Applied Mathematics , pp. 1–9, 2014. [Online]. Available: http://dx.doi.org/10.1155/2014/769343

2014 doi
[11]

A geometrical approach for online error compensation of industrial manipulators,

S. Mustafa, P. Tao, G. Yang, and I. Chen, “A geometrical approach for online error compensation of industrial manipulators,” Int. Conf. Adv. Intel. Mech. , pp. 738–743, 2010

2010
[12]

Calibration of industrial robots with product- of-exponential (poe) model and adaptive neural networks,

P. Tao and G. Yang, “Calibration of industrial robots with product- of-exponential (poe) model and adaptive neural networks,” Int. Conf. Robotics and Autom. , pp. 1448–1454, 2015

2015
[13]

Zero-offset calibration using a screw theory,

Y . Choi, J. Cheong, J. Kyung, and H. Min Do, “Zero-offset calibration using a screw theory,” Int. Conf. Ubiquitous Robots and Ambient Intel. , pp. 526–528, 2016

2016
[14]

A new kind of accurate calibration method for robotic kinematic parameters based on extended kalman and particle ﬁlter algorithm,

Z. Jiang, W. Zhou, H. Li, Y . Mo, W. Ni, and Q. Huang, “A new kind of accurate calibration method for robotic kinematic parameters based on extended kalman and particle ﬁlter algorithm,” IEEE Transactions on Industrial Electronics , 2017

2017
[15]

Practical industrial robot zero offset calibration,

H. Chen, T. Fuhlbrigge, S. Choi, J. Wang, and X. Li, “Practical industrial robot zero offset calibration,” Int. Conf. Autom. Science and Eng. , pp. 516–521, 2008

2008
[16]

Development and sensitivity analysis of a portable calibration system for joint offset of industrial robot,

Y . Liu, N. Xi, J. Zhao, E. Nieves-Rivera, y. Jia, B. Gao, and J. Lu, “Development and sensitivity analysis of a portable calibration system for joint offset of industrial robot,” IEEE Int. Conf. Int. Robots and Sys. , pp. 3838–3843, 2009

2009
[17]

An automated method to calibrate industrial robot joint offset using virtual line-based single-point constraint approach,

Y . Liu, N. Xi, G. Zhang, X. Li, H. Chen, C. Zhang, M. Jeffery, and T. Fuhlbrigge, “An automated method to calibrate industrial robot joint offset using virtual line-based single-point constraint approach,” IEEE Int. Conf. Int. Robots and Sys. , pp. 715–720, 2009

2009
[18]

An automated method to calibrate industrial robot kinematic parameters using spherical surface constraint approach,

Y . Liu, D. Shi, and J. Ding, “An automated method to calibrate industrial robot kinematic parameters using spherical surface constraint approach,” IEEE Int. Conf Cyber Techn. Autom. Control and Intel. Sys. , pp. 365– 370, 2014

2014
[19]

Nistir 8093: Tools for collaborative robots within sme workcells,

J. Marvel, E. Messina, B. Antonishek, L. Fronczek, and K. Van Wyk, “Nistir 8093: Tools for collaborative robots within sme workcells,” National Institute of Standards and Technology, Tech. Rep. , 2015

2015
[20]

Simpliﬁed framework for robot coordinate registration for manufacturing applications,

J. A. Marvel and K. Van Wyk, “Simpliﬁed framework for robot coordinate registration for manufacturing applications,” Proce. IEEE Int. Symp. Assem. Manufact. , pp. 56–63, 2016

2016
[21]

M. W. Spong, S. Hutchinson, and M. Vidyasagar, Robot modeling and control. Wiley New York, 2006, vol. 3

2006
[22]

Monte carlo sampling methods using markov chains and their applications,

W. K. Hastings, “Monte carlo sampling methods using markov chains and their applications,” Biometrika, vol. 57, no. 1, pp. 97–109, 1970

1970
[23]

Strategies for improving and evaluat- ing robot registration performance,

K. Van Wyk and J. A. Marvel, “Strategies for improving and evaluat- ing robot registration performance,” IEEE Transactions on Automation Science and Engineering , 2017

2017
[24]

Prior distributions for variance parameters in hierar- chical models,

A. Gelman et al. , “Prior distributions for variance parameters in hierar- chical models,” Bayesian analysis , vol. 1, no. 3, pp. 515–534, 2006

2006
[25]

The selection of prior distributions by formal rules,

R. E. Kass and L. Wasserman, “The selection of prior distributions by formal rules,” Journal of the American Statistical Association , vol. 91, no. 435, pp. 1343–1370, 1996

1996
[2016]

Available: http://www.ifr.org/industrial-robots/statistics/

[Online]. Available: http://www.ifr.org/industrial-robots/statistics/

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Robot metrology and calibration systems - a market review,

F. Hidalgo and P. Brunn, “Robot metrology and calibration systems - a market review,” Industrial Robot: An International Journal , vol. 25, no. 1, 1998

work page 1998

[2] [2]

2015: By far the highest volume ever recorded,

I. F. of Robotics, “2015: By far the highest volume ever recorded,”

work page 2015

[3] [3]

Recent progress on programming methods for industrial robots,

Z. Pan, J. Polden, N. Larkin, S. Van Duin, and J. Norrish, “Recent progress on programming methods for industrial robots,” Robotics and Computer-Integrated Manufacturing, vol. 28, pp. 87–94, 2012

work page 2012

[4] [4]

Comparative peg- in-hole testing of a force-based manipulation controlled robotic hand,

K. Van Wyk, M. Culleton, J. Falco, and K. Kelly, “Comparative peg- in-hole testing of a force-based manipulation controlled robotic hand,” IEEE Transactions on Robotics , in press

work page

[5] [5]

Addressing perception uncertainty induced failure modes in robotic bin-picking,

K. N. Kaipa, A. S. Kankanhalli-Nagendra, N. B. Kumbla, S. Shriyam, S. S. Thevendria-Karthic, J. A. Marvel, and S. K. Gupta, “Addressing perception uncertainty induced failure modes in robotic bin-picking,” Robotics and Computer-Integrated Manufacturing , vol. 42, pp. 17–38, 2016

work page 2016

[6] [6]

Dex-net 1.0: A cloud-based network of 3d objects for robust grasp planning using a multi-armed bandit model with correlated rewards,

J. Mahler, F. T. Pokorny, B. Hou, M. Roderick, M. Laskey, M. Aubry, K. Kohlhoff, T. Kr ¨oger, J. Kuffner, and K. Goldberg, “Dex-net 1.0: A cloud-based network of 3d objects for robust grasp planning using a multi-armed bandit model with correlated rewards,” in Robotics and Automation (ICRA), 2016 IEEE International Conference on . IEEE, 2016, pp. 1957–1964

work page 2016

[7] [7]

Kinematic noise propagation and grasp quality evaluation,

S. Liu and S. Carpin, “Kinematic noise propagation and grasp quality evaluation,” inAutomation Science and Engineering (CASE), 2016 IEEE International Conference on . IEEE, 2016, pp. 1177–1183

work page 2016

[8] [8]

A technique to calibrate industrial robots with experimental veriﬁcation,

R. Judd and A. Knasinski, “A technique to calibrate industrial robots with experimental veriﬁcation,” IEEE Trans. Robotics and Automation , vol. 6, no. 1, pp. 590–596, 1990

work page 1990