Pith. sign in

REVIEW 3 major objections 5 minor 36 references

Towards a Modular Bin-picking Framework for Handling Object Pose Uncertainties

T0 review · 3 major / 5 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read This paper presents a modular bin-picking framework that jointly handles object pose uncertainty and grasping errors, and claims to be the first of its kind.

arxiv 2607.13698 v1 pith:YUM6OG75 submitted 2026-07-15 cs.RO cs.CV

Frederik Hagelskj{ae}r This is my paper
classification cs.ROcs.CV
keywords binpickingobjectposedistributionuncertaintygraspingerrorrecoverymodularroboticssecondviewpointfusionproductoflikelihoodsSO(2)estimation
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's central claim is that bin-picking can be made robust and efficient by treating object pose uncertainty and grasping errors as separate, modular problems rather than one monolithic pipeline. It introduces a framework built on an existing pose-distribution-based pipeline, adding three modules: an in-hand pose check, a re-orientation tray, and a second camera view whose pose distributions are fused. In real-world tests, the full configuration achieved 100% insertion success, and the re-orientation tray alone or in-hand verification alone also reached 100%, while the second view improved efficiency. The paper argues this is the first modular framework for bin-picking that jointly handles both error types, and that the SO(2) design can be extended to full 6D pose space.

What carries the argument

The load-bearing mechanism is the product-of-likelihoods fusion of pose distributions from two views, p(θ|I1,I2) = p(θ|I1)p(θ|I2), justified by the conditional-independence assumption I1⊥I2|θ. Around this sit three interchangeable modules: in-hand pose verification, a re-orientation tray, and a second-view distribution estimator. The modularity lets each component be used alone or combined, and the decision flow routes uncertain objects to the second view or to the tray.

What would settle it

Collect a dataset of the same object in a bin under two fixed views with ground-truth poses, and compare the fused distribution p(θ|I1)p(θ|I2) against the empirical posterior from many trials. If the fused distribution is systematically overconfident — low entropy and wrong mode — when lighting or occlusion changes between views, the independence or calibration assumptions fail.

Watch

Extended reading notes

Core claim

The author proposes a modular bin-picking framework where each module addresses a specific source of failure. Object pose uncertainty is handled by estimating a full distribution over in-plane rotation and reflection rather than a single pose; when the distribution is too wide, a second fixed-view camera provides another distribution, and the two are multiplied under the assumption that the images are conditionally independent given the pose. Grasp errors are handled by verifying the object's pose in the gripper before insertion and by moving ambiguous objects to a re-orientation tray where they are re-estimated from a stable surface. Experiments on a real workcell with three objects show th

Load-bearing premise

The second-view fusion only works if the two camera views are conditionally independent given the true object pose, and if the per-view pose distributions are calibrated likelihoods; neither is verified in the paper.

Editorial extensions

If this is right

  • If the framework's claims hold, bin-picking systems can be assembled from independent modules rather than custom one-off pipelines.
  • The second-view module demonstrates that a fixed additional viewpoint can reduce pose uncertainty enough to avoid costly reorientation cycles.
  • The 100% insertion success with either in-hand pose verification or re-orientation tray suggests that grasp-error recovery alone suffices for reliable insertion for these objects.
  • The modular structure allows deployment even when some modules are physically impossible in a given workcell.
  • Extending to SE(3) would let the framework handle arbitrary object geometries and explicitly model grasping uncertainty.

Reading between the lines

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

  • The product-of-likelihoods fusion assumes calibrated pose distributions; if the network outputs are overconfident, the fused distribution may be too peaked, so a calibration check would be a natural extension.
  • The conditional-independence assumption can be tested by comparing fusion against a jointly estimated distribution under varying lighting or occlusion; violations would show up as overconfident wrong peaks.
  • The framework's success suggests that treating uncertainty as a first-class object, rather than a single pose, could be applied to other manipulation tasks like assembly or kitting, not just bin feeding.
  • The in-hand pose verification currently uses a single-point estimate; converting it to a distribution, as the paper suggests, could let the system decide whether to re-grasp based on confidence rather than a binary match.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper presents a modular bin-picking framework that combines object pose distribution estimation with modules for handling grasping errors. The core modules are in-hand pose verification, a re-orientation tray, and a second-view pose-distribution fusion step. The framework builds on an earlier SO(2) pose-distribution strategy [17] and is evaluated in a real workcell with three objects. The main empirical claim is that using all modules yields the best performance: 100% insertion success at 1.91 grasps per insertion, compared with 89% success and 2.5 grasps per insertion for the original method. Additional experiments on two World Robot Summit Assembly Challenge objects report 100% success in reflection-discriminative kitting. The paper claims to be the first modular framework for bin-picking that jointly handles object pose uncertainty and grasping errors.

Significance. If the empirical results are reliable, the paper makes a useful systems-level contribution: it demonstrates a reusable modular recipe for combining pose-distribution-driven decision making with grasp-error recovery, and it introduces a second-view fusion module that is not common in bin-picking practice. The ablation-style comparison across module configurations is a strength, as is the real-world validation on multiple objects, including difficult WRS objects. However, the current evidence is not yet strong enough to support the paper's central efficiency claims: the fusion rule is not justified probabilistically, one module combination contradicts the abstract's blanket efficiency claim, and the experiments lack statistical grounding. The paper is a good starting point, but the load-bearing points need further work.

major comments (3)
  1. [§III-E, Eqs. (3)–(4)] The second-view module is a central contribution, but the fusion rule is not justified. Under the stated conditional independence I1⊥I2|θ, Bayes' rule gives p(θ|I1,I2) ∝ p(I1|θ)p(I2|θ)p(θ) = p(θ|I1)p(θ|I2)/p(θ) when each p(θ|Ii) is computed with the same prior. Equation (4) omits the division by p(θ), so it is valid only for a uniform prior over θ (and after renormalization). No uniformity argument is given for the discrete revolution/reflection states in §III-A. Moreover Eq. (3) is asserted rather than tested; shared lighting, sensor noise, occlusions, and object surface properties across the two fixed views can induce dependence given θ. If the network posteriors are not calibrated likelihoods, the product can be overconfident and the cutoff decisions in §III-H can be wrong. Please either provide calibration/independence validation, derive the correct Bayesian combination, or explicitl
  2. [Abstract and Table I] The abstract's statement that 'all modules are shown to improve efficiency' is not supported by Table I. Adding IHP to the original system changes success rate from 89% to 100% but worsens grasps per insertion from 2.5 to 2.8 (rows 1–2). If 'efficiency' means success rate, then IHP improves it, but the paper's primary efficiency metric is grasps per insertion; by that metric IHP alone reduces efficiency. The later comparison O+RT+NV (2.05) vs O+RT+NV+IHP (1.91) does support IHP in combination, but not the blanket statement. Please define 'efficiency' and specify which metric each module improves.
  3. [§IV, Table I and §IV-A] The experimental comparisons are based on a single run per configuration stopping after 40 successful insertions. No repeated trials, confidence intervals, or significance tests are reported. The relevant differences are small: 2.25 vs 2.05 grasps per insertion for adding NV, and 2.05 vs 1.91 for adding IHP on top; with bin-picking variability these differences may be within noise. Also, because each run stops at 40 successes, the number of observed failures differs across runs but is not reported, so success-rate comparisons lack denominators. Please report total attempts, run-level results, repetitions, and an appropriate uncertainty estimate before claiming that the new-view module and the full configuration are the most efficient.
minor comments (5)
  1. [Abstract] The phrase 'evaluated in a real-world setup with three different objects, with no errors' is grammatically and semantically unclear. Presumably the intended meaning is 'achieving no insertion errors' or 'with 100% success.' Please rephrase.
  2. [§IV, first paragraph] The text says 'The results are shown in Table IV' but the table is labeled Table I. Please correct the cross-reference.
  3. [§III-A, §III-E] The pose-uncertainty cutoff is a central decision threshold, but its value and origin are never specified. State whether it is inherited from [17], and give its numerical value or how it is selected.
  4. [§III-E] The second-view angle is manually defined as 30 degrees from the front. No justification or sensitivity analysis is given. At minimum, state that this is a fixed parameter chosen from the workcell geometry and note that the results may depend on it.
  5. [§I, last paragraph of Introduction] The sentence beginning 'While our framework currently operates in SO(2), leveraging only in-plane rotations...' is incomplete. Revise for grammatical clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's contribution is system integration validated by real-world experiments, and its fusion rule is an explicit modeling assumption rather than a fitted result.

full rationale

I walked the claimed derivation chain. The paper is a system-integration contribution: the pose-distribution estimator, uncertainty cutoff, and workcell are inherited from prior work ([17], [31], [4], [9]), several by the same author. However, the paper does not derive its headline numbers from these citations; it measures them on a real workcell. Table I compares module combinations and reports empirical success rates and grasps-per-insertion (89%→100%, 2.5→1.91), and no parameter is fitted to force these outcomes. The second-view fusion rule (Eq. 4) is stated as an assumption citing Product of Likelihoods [32]; it is not derived from data, nor is it a renamed version of the experimental result. Whether Eq. (4) is statistically sound—because p(θ|I) are posteriors rather than likelihoods and because conditional independence (Eq. 3) is untested—is a correctness/calibration concern, not circularity. The 'first modular framework' claim is a novelty assertion, not a computation. Heavy self-citation is present, but it is cumulative research use of the author's own components, not a load-bearing self-referential proof. Thus no step reduces by construction to its own inputs.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The modular framework is a system composition of existing techniques; the main assumptions are calibrated pose distributions from prior same-author work and a hand-set cutoff.

free parameters (2)
  • pose uncertainty cutoff = not reported (inherited from [17])
    The threshold on the pose distribution decides whether to grasp, reorient, or acquire a second view; all efficiency numbers depend on its value.
  • second-view angle = 30 degrees from front
    Chosen manually from bin geometry; affects how much the second view disambiguates poses.
assumptions (4)
  • domain assumption Conditional independence of the two camera views: I1 ⊥ I2 | θ (Eq. 3).
    Underlies the Product-of-Likelihoods fusion in Eq. (4); if false, the fused distribution is overconfident and the second-view module's reported improvement may not generalize.
  • domain assumption Pose distributions p(θ|I) produced by [17] are calibrated likelihoods.
    Needed for cutoff decisions and for multiplying the distributions as likelihoods; no calibration check is reported.
  • domain assumption Objects lie flat and stable in the re-orientation tray, so grasps from the tray are error-free.
    Basis for the RT module's 100% success rate and reduced grasps; assumed rather than measured separately.
  • domain assumption SO(2) representation captures the relevant pose ambiguity for the test objects.
    The paper explicitly limits to cylindrical-symmetric objects; extension to SE(3) is not demonstrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards a Modular Bin-picking Framework for Handling Object Pose Uncertainties." pith.science (2026). https://pith.science/paper/YUM6OG75

@misc{pith2026260713698,
  author       = {Pith},
  title        = {Pith review of: Towards a Modular Bin-picking Framework for Handling Object Pose Uncertainties},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YUM6OG75}},
  note         = {Machine review of arXiv:2607.13698}
}
read the original abstract

In recent years, there has been growing interest in robust robotic systems for precise bin-picking applications. To achieve reliable performance, such systems must address errors arising from both the object pose estimation and the grasping process. Although various approaches have been proposed, they typically target specific challenges and do not offer general solutions. In this paper, we present a modular framework that jointly handles both error types. The framework incorporates object pose distribution estimation to account for pose uncertainty, which frequently arises in situations with ambiguous observations where a single correct pose cannot be determined. To further reduce uncertainty, we introduce a second-viewpoint module that computes complementary pose distributions, which are subsequently fused. This fusion decreases overall uncertainty and improves system efficiency. Additionally, two independent modules are included to compensate for grasping errors. The modular design allows the components to be combined for optimal performance or used individually, depending on the physical setup. The proposed method is evaluated in a real-world setup with three different objects, with no errors, and all modules are shown to improve efficiency. These results suggest that incorporating pose distributions with grasping pose errors is a promising direction for developing more flexible and reliable robotic production systems. To the best of our knowledge, this is the first framework that jointly addresses both grasping and object pose uncertainties using interchangeable modules. We believe there is ample opportunity to integrate additional modules, resulting in improved performance and flexibility. The current framework is limited to pose uncertainties in SO(2), but it could be extended to SE(3), enabling additional modules to improve the system.

Figures

Figures reproduced from arXiv: 2607.13698 by the authors.

Figure 1
Figure 1. Example of the framework when inserting an object. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Activity diagram of the bin-picking process with an added in-hand [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The framework for the proposed bin-picking strategy. White: Original [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The workcell used for experiments. The overhead camera for in-hand [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The objects used in the experiments shown from both sides. For [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

36 extracted references · 6 linked inside Pith

  1. [17]

    Object pose distribution estimation for determining revolution and reflection uncertainty in point clouds,

    F. Hagelskjær, D. Arapis, S. Madsen, and T. M. Iversen, “Object pose distribution estimation for determining revolution and reflection uncertainty in point clouds,” in2025 7th International Conference on Control and Robotics (ICCR), 2025, pp. 201–208

    2025

  2. [1]

    Bin-picking—5 decades of research,

    D. Buchholz, “Bin-picking—5 decades of research,”Bin-Picking: New Approaches for a Classical Problem, pp. 3–12, 2015

    2015

  3. [2]

    Current research trends in robot grasping and bin picking,

    M. Alonso, A. Izaguirre, and M. Gra ˜na, “Current research trends in robot grasping and bin picking,” inThe 13th International Conference on Soft Computing Models in Industrial and Environmental Applications. Springer, 2018, pp. 367–376

    2018

  4. [3]

    Bin picking approaches based on deep learning techniques: A state-of-the-art survey,

    A. Cordeiro, L. F. Rocha, C. Costa, P. Costa, and M. F. Silva, “Bin picking approaches based on deep learning techniques: A state-of-the-art survey,” in2022 IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC). IEEE, 2022, pp. 110–117

    2022

  5. [4]

    Good grasps only: A data engine for self-supervised fine-tuning of pose estimation using grasp poses for verification,

    F. Hagelskjær, “Good grasps only: A data engine for self-supervised fine-tuning of pose estimation using grasp poses for verification,” in 2025 IEEE/SICE International Symposium on System Integration (SII). IEEE, 2025, pp. 957–964

    2025

  6. [5]

    Bop: Benchmark for 6d object pose estimation,

    T. Hodan, F. Michel, E. Brachmann, W. Kehl, A. GlentBuch, D. Kraft, B. Drost, J. Vidal, S. Ihrke, X. Zabuliset al., “Bop: Benchmark for 6d object pose estimation,” inProceedings of the European conference on computer vision (ECCV), 2018, pp. 19–34

    2018

  7. [6]

    Bop challenge 2020 on 6d object localization,

    T. Hoda ˇn, M. Sundermeyer, B. Drost, Y . Labb ´e, E. Brachmann, F. Michel, C. Rother, and J. Matas, “Bop challenge 2020 on 6d object localization,” inComputer Vision–ECCV 2020 Workshops: Glasgow, UK, August 23–28, 2020, Proceedings, Part II 16. Springer, 2020, pp. 577–594

    2020

  8. [7]

    Bop challenge 2022 on detection, segmen- tation and pose estimation of specific rigid objects,

    M. Sundermeyer, T. Hodan, Y . Labbe, G. Wang, E. Brachmann, B. Drost, C. Rother, and J. Matas, “Bop challenge 2022 on detection, segmen- tation and pose estimation of specific rigid objects,”arXiv preprint arXiv:2302.13075, 2023

    arXiv 2022

Show all 36 references
  1. [8]

    Bop challenge 2023 on detection segmentation and pose estimation of seen and unseen rigid objects,

    T. Hodan, M. Sundermeyer, Y . Labbe, V . N. Nguyen, G. Wang, E. Brach- mann, B. Drost, V . Lepetit, C. Rother, and J. Matas, “Bop challenge 2023 on detection segmentation and pose estimation of seen and unseen rigid objects,” inProceedings of the IEEE/CVF Conference on Compute...

    2023

  2. [9]

    Off-the-shelf bin picking workcell with visual pose estimation: A case study on the world robot summit 2018 kitting task,

    F. Hagelskjær, D. Kraftet al., “Off-the-shelf bin picking workcell with visual pose estimation: A case study on the world robot summit 2018 kitting task,” in2024 21st International Conference on Ubiquitous Robots (UR). IEEE, 2024, pp. 145–152

    2018

  3. [10]

    Towards robot cell matrices for agile production–sdu robotics’ assembly cell at the wrc 2018,

    C. Schlette, A. G. Buch, F. Hagelskjær, I. Iturrate, D. Kraft, A. Kram- berger, A. P. Lindvig, S. Mathiesen, H. G. Petersen, M. H. Rasmussen et al., “Towards robot cell matrices for agile production–sdu robotics’ assembly cell at the wrc 2018,”Advanced Robotics, vol. 34, no. 7...

    2018

  4. [11]

    A smooth representation of belief over so (3) for deep rotation learning with uncertainty,

    V . Peretroukhin, M. Giamou, D. M. Rosen, W. N. Greene, N. Roy, and J. Kelly, “A smooth representation of belief over so (3) for deep rotation learning with uncertainty,”arXiv preprint arXiv:2006.01031, 2020

    2006 arXiv

  5. [12]

    Implicit-pdf: Non-parametric representation of probability distributions on the rotation manifold,

    K. Murphy, C. Esteves, V . Jampani, S. Ramalingam, and A. Makadia, “Implicit-pdf: Non-parametric representation of probability distributions on the rotation manifold,”arXiv preprint arXiv:2106.05965, 2021

    2021 arXiv

  6. [13]

    Ki-pode: Keypoint- based implicit pose distribution estimation of rigid objects,

    T. M. Iversen, R. L. Haugaard, and A. G. Buch, “Ki-pode: Keypoint- based implicit pose distribution estimation of rigid objects,”arXiv preprint arXiv:2209.09659, 2022

    2022 arXiv

  7. [14]

    Spyropose: Se (3) pyramids for object pose distribution estimation,

    R. L. Haugaard, F. Hagelskjær, and T. M. Iversen, “Spyropose: Se (3) pyramids for object pose distribution estimation,” inProceedings of the IEEE/CVF International Conference on Computer Vision, 2023, pp. 2082–2091

    2023

  8. [15]

    Confronting ambiguity in 6d object pose estimation via score-based diffusion on se (3),

    T.-C. Hsiao, H.-W. Chen, H.-K. Yang, and C.-Y . Lee, “Confronting ambiguity in 6d object pose estimation via score-based diffusion on se (3),” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 352–362

    2024

  9. [16]

    Corr2distrib: Making ambiguous correspondences an ally to predict reliable 6d pose distributions,

    A. Brazi, B. Meden, F. M. de Chamisso, S. Bourgeois, and V . Lepetit, “Corr2distrib: Making ambiguous correspondences an ally to predict reliable 6d pose distributions,”IEEE Robotics and Automation Letters, 2025

    2025

  10. [18]

    Assembly challenge: a robot competition of the industrial robotics category, world robot summit–summary of the pre-competition in 2018,

    Y . Yokokohji, Y . Kawai, M. Shibata, Y . Aiyama, S. Kotosaka, W. Ue- mura, A. Noda, H. Dobashi, T. Sakaguchi, and K. Yokoi, “Assembly challenge: a robot competition of the industrial robotics category, world robot summit–summary of the pre-competition in 2018,”Advanced Roboti...

    2018

  11. [19]

    Large-scale 6d object pose estimation dataset for industrial bin-picking. in 2019 ieee,

    K. Kleeberger, C. Landgraf, and M. F. Huber, “Large-scale 6d object pose estimation dataset for industrial bin-picking. in 2019 ieee,” in RSJ International Conference on Intelligent Robots and Systems (IROS), 2019, pp. 2573–2578

    2019

  12. [20]

    Scara+ system: Bin picking sys- tem of revolution-symmetry objects,

    G. Jin, X. Yu, Y . Chen, and J. Li, “Scara+ system: Bin picking sys- tem of revolution-symmetry objects,”IEEE Transactions on Industrial Electronics, 2024

    2024

  13. [21]

    Two-stage grasping: A new bin picking framework for small objects,

    H. Cao, J. Zhou, J. Huang, Y . Li, N. C. Meng, R. Cao, Q. Dou, and Y . Liu, “Two-stage grasping: A new bin picking framework for small objects,”arXiv preprint arXiv:2303.02604, 2023

    2023 arXiv

  14. [22]

    Robotic general parts feeder: Bin-picking, regrasping, and kitting,

    Y . Domae, A. Noda, T. Nagatani, and W. Wan, “Robotic general parts feeder: Bin-picking, regrasping, and kitting,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 5004–5010

    2020

  15. [23]

    Robust bin-picking system using tactile sensor,

    S. Tajima, S. Wakamatsu, T. Abe, M. Tennomi, K. Morita, H. Ubata, A. Okamura, Y . Hirai, K. Morino, Y . Suzukiet al., “Robust bin-picking system using tactile sensor,”Advanced Robotics, vol. 34, no. 7-8, pp. 439–453, 2020

    2020

  16. [24]

    Towards high precision: An adaptive self-supervised learning framework for force-based verification,

    Z. Duan, F. Hagelskjær, A. Kramberger, J. Heredia, , and N. Kr ¨uger, “Towards high precision: An adaptive self-supervised learning framework for force-based verification,” 2025. [Online]. Available: https://arxiv.org/abs/2508.02153

    2025 arXiv

  17. [25]

    Using vision for pre-and post- grasping object localization for soft hands,

    C. Choi, J. D. Preto, and D. Rus, “Using vision for pre-and post- grasping object localization for soft hands,” inInternational Symposium on Experimental Robotics. Springer, 2016, pp. 601–612

    2016

  18. [26]

    Robust, occlusion-aware pose estimation for objects grasped by adaptive hands,

    B. Wen, C. Mitash, S. Soorian, A. Kimmel, A. Sintov, and K. E. Bekris, “Robust, occlusion-aware pose estimation for objects grasped by adaptive hands,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 6210–6217

    2020

  19. [27]

    Reorientating objects with a gripping hand and a table surface,

    W. Wan and K. Harada, “Reorientating objects with a gripping hand and a table surface,” in2015 IEEE-RAS 15th International Conference on Humanoid Robots (Humanoids). IEEE, 2015, pp. 101–106

    2015

  20. [28]

    Robotic task success evaluation under multi-modal non-parametric object pose uncertainty,

    L. Naik, T. M. Iversen, A. Kramberger, and N. Kr ¨uger, “Robotic task success evaluation under multi-modal non-parametric object pose uncertainty,”Industrial Robot: the international journal of robotics research and application, 2025

    2025

  21. [29]

    Deep learning a grasp function for grasping under gripper pose uncertainty,

    E. Johns, S. Leutenegger, and A. J. Davison, “Deep learning a grasp function for grasping under gripper pose uncertainty,” in2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2016, pp. 4461–4468

    2016

  22. [30]

    Se (3)-poseflow: Estimating 6d pose distribu- tions for uncertainty-aware robotic manipulation,

    Y . Jin, N. Funk, V . Prasad, Z. Li, M. Franzius, J. Peters, and G. Chalvatzaki, “Se (3)-poseflow: Estimating 6d pose distribu- tions for uncertainty-aware robotic manipulation,”arXiv preprint arXiv:2511.01501, 2025

    2025

  23. [31]

    Keymatchnet: Zero-shot pose estimation in 3d point clouds by generalized keypoint matching,

    F. Hagelskjær and R. L. Haugaard, “Keymatchnet: Zero-shot pose estimation in 3d point clouds by generalized keypoint matching,” in 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE). IEEE, 2024, pp. 870–877

    2024

  24. [32]

    Proper likelihoods for bayesian analysis,

    J. F. Monahan and D. D. Boos, “Proper likelihoods for bayesian analysis,”Biometrika, vol. 79, no. 2, pp. 271–278, 1992

    1992

  25. [33]

    Advances in machine vision for flexible feeding of assembly parts,

    A. A. Malik, M. V . Andersen, and A. Bilberg, “Advances in machine vision for flexible feeding of assembly parts,”Procedia Manufacturing, vol. 38, pp. 1228–1235, 2019

    2019

  26. [34]

    A planning framework for non- prehensile manipulation under clutter and uncertainty,

    M. R. Dogar and S. S. Srinivasa, “A planning framework for non- prehensile manipulation under clutter and uncertainty,”Autonomous Robots, vol. 33, no. 3, pp. 217–236, 2012

    2012

  27. [35]

    Control of nonprehensile manipula- tion,

    K. M. Lynch and T. D. Murphey, “Control of nonprehensile manipula- tion,” inControl problems in robotics. Springer, 2003, pp. 39–57

    2003

  28. [36]

    Robot control for nonprehensile dynamic manipulation tasks,

    D. Serra, “Robot control for nonprehensile dynamic manipulation tasks,” inInternational Conference on Informatics in Control, Automation and Robotics, Doctoral Consortium, 2016, pp. 3–12

    2016

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.