Learning to Optimize Package Picking for Large-Scale, Real-World Robot Induction

Azarakhsh Keipour; Charles Swan; Kostas E. Bekris; Shuai Li; Sicong Zhao; Srinath Rajagopalan

REVIEW 4 major objections 6 minor 17 references

Learning to Optimize Package Picking for Large-Scale, Real-World Robot Induction

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A learned optimizer trained on pick-success score differences reduced physical missed picks from 2.23% to 1.80% across two million robot picks.

desk verdict Large-scale A/B test shows a real 0.43 pp missed-pick reduction, but the paper under-documents the experiment and ignores multi-pick costs; worth peer review with revisions. read the letter →

arxiv 2506.09765 v1 pith:X62EMVAH submitted 2025-06-11 cs.RO cs.LG

Shuai Li , Azarakhsh Keipour , Sicong Zhao , Srinath Rajagopalan , Charles Swan , Kostas E. Bekris This is my paper · ORCID

classification cs.ROcs.LG

keywords roboticpickoptimizationsuctiongripperswarehouseautomationsuccesspredictionA/Btestingrobotinductionvectorfieldregressionlarge-scalerobotics

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

Pick-and-place robots in parcel induction must decide where to put a suction end effector on cluttered, deformable packages; bad choices mean a missed pick, a re-pick, or a costly manual intervention. This paper proposes training a lightweight optimizer that, given a candidate pick from any sampler, nudges its position and rotation to raise the probability of success, using an existing learned pick-success model as the teacher. The teacher labels about a thousand physical picks plus their Gaussian-perturbed variants (about 28,000 comparisons) with the direction that increases predicted success, and regression models learn that direction in the action space. Over one million physical picks per group, the optimized group missed 1.80% of packages versus 2.23% for the heuristic baseline, a 0.43-percentage-point drop and about a 19.25% relative improvement. The paper's claim is that directly optimizing sampled picks, not just ranking many heuristic samples, is the source of the gain and that the gain persists at industrial scale.

What carries the argument

The load-bearing object is the training signal constructed from perturbed-pick comparisons: for an executed action $a_i$ and each Gaussian-perturbed neighbor $a_{i+1}$, the pick-success model F returns probabilities $p_i$ and $p_{i+1}$, and the target $\delta_i$ is the vector from the lower-probability action to the higher-probability action, paired with the feature vector of the higher-probability action. Three regression models, $G_x$ for the X offset, $G_y$ for the Y offset given the corrected X, and $G_r$ for the rotation given corrected X and Y, learn this mapping from scene features to action-space corrections. At deployment the predicted corrections are applied to the initial pick and iterated several times, moving the suction cups across package surfaces and into configurations F scores higher, such as pressing against flat surfaces away from labels or activating an additional cup.

What would settle it

Run a matched-scene A/B test in which, for every optimized pick, the robot also attempts the original unoptimized pick on an identical or nearly identical package layout; if picks where F's predicted probability rose by a large margin do not physically succeed more often than their unoptimized counterparts, then F's landscape is not a faithful teacher and the proposed mechanism would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that a pick action for a suction-based induction robot can be treated as a point in a three-dimensional action space, namely X, Y, and rotation about the surface normal, and that a learned vector field in this space moves an initial sampled pick to a nearby configuration with higher physical success. The vector field is learned from the pick-success model F: for each executed pick, Gaussian perturbations are generated, F scores the original and perturbed actions, and the signed difference defines a target displacement $\delta_i$ pointing toward higher predicted probability. At inference the learned regressors $G_x$, $G_y$, and $G_r$ are applied autoregressively and iteratively, with the Z coordinate and surface normal recomputed from the geometry under the X, Y location. The paper reports that this optimization, integrated into production-style induction workcells, reduced physical missed picks from 2.23% to 1.80% over two million picks, with non-overlapping 95% confidence intervals, while leaving the infeasible-pick rate unchanged and slightly increasing the multi-pick rate.

Load-bearing premise

The whole training pipeline assumes that the pick-success model F's predicted probability surface points in the same direction as real physical success, so that moving a pick to where F's score is higher is actually moving it to where it succeeds more often; the two-million-pick A/B test is the paper's evidence that this held in the tested workcells.

Editorial extensions

If this is right

If deployed at production scale, the 0.43-percentage-point reduction in missed-pick rate would cut re-picks, throughput losses, and manual-amnesty interventions roughly in proportion to the 19.25% relative improvement.
Because training needs only about 1,000 physical picks plus a pick-success model F, the same recipe can bootstrap pick optimization for other suction-based induction systems that already have such a model.
Iterative refinement changes which suction cups engage on a package surface, so the optimizer can shift cups away from labels and poor-seal regions while increasing the number of active cups, which the paper ties to more stable transfers.
The infeasible-pick rate did not improve and the multi-pick rate rose slightly, indicating the gains come from pick quality rather than feasibility; a production deployment should monitor multi-pick behavior.

Reading between the lines

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

The teacher model F is the crux of the approach: in a setting where F is miscalibrated or locally flat, the learned vector field could inflate F's score without improving physical outcomes, so a calibration audit of F's local landscape is a cheap safeguard before transferring the method.
This is effectively score-function distillation from F into an action-space vector field, which suggests the same data-generation recipe could optimize for other objectives, such as human preference scores or damage avoidance, by changing what the labels reward.
The paper's closing note implies a closed improvement loop: picks the optimizer generates can later be used to retrain F, which would then relabel a new optimizer; if that loop converges, the long-term gain could exceed the one-time 20% reduction measured here.
A clean ablation would separate the benefit of learned regression from simply giving the baseline model more candidate samples to rank; comparing the optimizer against re-ranking of the same perturbed actions using F alone would isolate the contribution of the learned vector field.

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

Colleague,

Read Li et al. on package picking. The headline is right: over two million physical picks, the treatment group missed 18,015 vs 22,310 for the control, a 0.43 percentage point absolute and 19.25% relative reduction. That is a genuine operational gain and the experiment scale is far beyond typical robotics papers. The contribution is also real: they train a regression model to predict pick-parameter adjustments using a pseudo-gradient derived from a previously learned pick-success model F, and they do it autoregressively across position and rotation dimensions. I haven't seen that exact scheme in prior suction picking work.

What the paper does well: the evaluation is direct — physical A/B test, not just simulation — and they report secondary metrics (infeasible, multi-pick) rather than hiding them. The comparison between gradient boosting and a small MLP is honest.

The soft spots are in the experimental reporting, not the core idea. First, the A/B design is under-specified. There's no statement that the control and treatment groups ran concurrently on the same workcells, with randomized assignment or balanced package mix. If the control data predates the treatment or came from different cells, the 0.43 pp is confounded. Second, the CI reporting is implausible. For 1M Bernoulli draws, a 2.23% rate should have a CI around ±0.03 pp, not [2.23,2.23]. Either they are rounding to two decimals or the data is clustered and they ignored that. A cluster-robust standard error is needed. Third, the treatment group has a higher multi-pick rate (0.89% vs 0.84%), and the paper never tells you how multi-picks are counted in the missed-pick denominator or what they cost. If a multi-pick can lead to extra drops or amnesty events, the net benefit of the treatment is smaller than the headline number.

On the circularity worry: yes, the optimizer is trained on labels generated by F, so the objective is defined by F's predictions. But the A/B test is external and validates that optimizing F transfers to physical picking. In this setting the assumption held. I would flag F's fidelity as a transfer limitation, not a flaw in this paper.

Bottom line: this paper deserves a serious referee and is likely publishable after the authors document the A/B design, report cluster-robust CIs, and explain multi-pick accounting. I would bring it to a reading group focused on real-world robot learning.

Referee Report

4 major / 6 minor

Summary. This paper proposes an ML-based pick optimization framework for robotic induction. Given a scene and an initial pick action, a regression model G predicts adjustments to the pick's x/y position and rotation; G is trained on synthetic labels generated by comparing the predicted success probabilities of a pre-trained pick success model F for perturbed actions. The authors integrate G into a FANUC-based workcell resembling Amazon's Robin system and report an A/B test with one million picks per group: missed pick rate drops from 2.23% (22,310/1M) to 1.80% (18,015/1M), a 0.43 pp absolute and 19.25% relative reduction. They also report infeasible pick rates (~4.49%) and multi-pick rates (0.84% vs 0.89%). The central claim is that learned pick optimization, rather than heuristic sampling plus ranking, causes the missed-pick improvement.

Significance. If the A/B effect is causal, the paper provides rare large-scale physical evidence for learned pick optimization in industrial suction-based picking, with practical significance for warehouse automation. Strengths include the scale (2M physical picks), the external evaluation on real hardware, a clear comparison against a learned-ranking baseline, and the use of a held-out set for model selection (Table 1). The work is a useful existence proof that optimizing sampled picks with a learned vector field can transfer to physical outcomes, even though the training data are derived from F. The main caveats are methodological: the A/B design, outcome accounting, and confidence-interval construction must be documented for the quantitative claim to be fully credible.

major comments (4)

[§3, Quantitative Results] The A/B test design is not described. The text does not state whether groups C and T were run concurrently, on the same workcells, with randomly assigned inducts, or with a balanced package mix. If the control data were collected under different conditions, the 0.43 percentage point difference in Table 2 may be confounded. The authors should specify the experimental protocol, including time period, workcell assignment, randomization, and the exact rule by which inducts were allocated to C and T.
[§3, Table 2] The reported 95% confidence intervals are not consistent with a simple binomial model of missed-pick counts: for 22,310/1,000,000 the approximate 95% CI is 2.23% ± 0.03%, not [2.23%, 2.23%], and for 18,015/1,000,000 it is roughly 1.80% ± 0.03%. The printed intervals are either rounded aggressively, computed from a different denominator, or adjusted for clustering by workcell/day. The paper should state how the intervals were constructed and report the effective sample size, since the statistical significance claim is based on these intervals.
[§3, Table 3 and §4] The treatment group has a higher multi-pick rate (0.89% vs 0.84%) with non-overlapping CIs, and the paper hypothesizes this reflects better pick quality. However, it does not state how multi-pick events are treated in the missed-pick rate denominator. If a multi-pick produces an extra package that is dropped, placed in amnesty, or requires rehandling, the net operational benefit of the optimizer is smaller than the 0.43 pp reduction. The authors should define the outcome hierarchy and report missed picks conditional on single vs. multi-pick outcomes.
[§2, Training Data Generation] The supervision for G comes from F's own predictions: Eq. (1)-(2) define the target gradient solely by comparing F(phi(s_t,a_i)) and F(phi(s_t,a_{i+1})). This makes the method circular in the sense that G can only be as good as the local landscape of F. The physical A/B test is strong evidence that this loop works in the tested conditions, but the paper should (a) state this dependence explicitly as an assumption, and (b) report evidence of F's calibration or a comparison between F's predicted improvement and realized pick outcomes on a held-out set. This is load-bearing for the generality of the framework, though not for the specific observed A/B result.

minor comments (6)

[§3, Quantitative Results] There is a typo: the text says '1 million picks for group C and 1 million picks for group C'; the second group should be T.
[Abstract and §4] The abstract reports a '20% reduction' while §4 reports a '19.25% relative improvement'; please make the reported percentage consistent or explicitly state the rounding.
[§2, Training Data Generation] The paper reports 27,977 data points generated from 1,000 physical picks but does not specify N (the number of perturbed actions per pick) or the noise variances sigma_pos and sigma_rot; please provide these values.
[§3, Table 1] The RMSE comparison would benefit from units, standard errors or repeated-run variation, and a description of the gradient boosting hyperparameters used; this would strengthen the model-selection argument.
[§3, Control Group] The control baseline is described only as 'similar to Li et al. [8]'; please specify the exact components of the baseline and confirm that the only difference between C and T is the proposed optimizer.
[§4, Statistical Significance] The notation 'T1' appears in the discussion of Table 3 but the group is named T; please correct this for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the optimizer is trained on F-labeled gradients, but the headline missed-pick result is a 2M-pick physical A/B test that breaks the loop.

full rationale

The internal optimization objective is defined in terms of the pick-success model F: the training targets δ_i in Eq. (1)-(2) are generated by comparing F(φ(s_t,a_i)) and F(φ(s_t,a_{i+1})), and G is trained to predict these targets. If the paper's claim were that G improves F scores, this would be circular. But the central claim is the physical missed-pick reduction in Tables 2 and 3, measured as failed package transfers over 1M inducts per group, not as F outputs. That external benchmark makes the success of 'optimize F' an empirical physical question, not a definitional consequence. F itself is an assumed prior input (Section 2, 'Access to Learned Model of Pick Success'), cited to the authors' prior [8]; this is a modular assumption, not evidence for the physical result. The RMSE model comparison in Table 1 is internal (predicting F-derived deltas) and is not presented as the physical result. No fitted parameter or predicted quantity is shown to be its own input, and no load-bearing uniqueness or ansatz is imported via self-citation. The minor self-citations are not load-bearing for the central A/B claim.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The framework relies heavily on a pre-existing pick success prediction model F developed by the same group. The learned optimizer G is trained on labels synthesized from F, and its evaluation is physical. The main free parameters are perturbation noise variances, number of perturbations per pick, inference-time iteration count, and feature-extraction radii, none of which are reported. The central domain assumption is that F is a faithful proxy for physical success; the large-scale A/B test provides empirical support, but the assumption is structurally load-bearing.

free parameters (7)

sigma_pos (position noise standard deviation)
Gaussian noise variance applied to pick position to create training perturbations; value not reported, affects the scale of learned gradients.
sigma_rot (rotation noise standard deviation)
Gaussian noise variance applied to rotation when generating training pairs; value not reported.
N (number of perturbed actions per executed pick)
Number of noise samples per real pick; not reported, but determines the 27,977 total training points from 1,000 picks.
Number of optimization iterations at inference
The paper shows qualitative results after 2 and 3 iterations but does not state the iteration count used in the A/B test.
d (local height map neighborhood radius)
Radius in meters around the target package for computing the local height map; value not reported.
MLP hidden sizes and learning rate = 35, 2, 0.001
Reported for the 2-layer MLP, but the deployed model is gradient boosting, whose hyperparameters are not reported.
Gradient boosting hyperparameters
The selected model for deployment; hyperparameters such as number of trees, depth, and learning rate are not reported.

assumptions (6)

domain assumption The learned pick success model F faithfully approximates physical pick success probability.
The entire training signal for G comes from F; if F is miscalibrated, learned gradients may not correspond to true physical improvements. Invoked in Section 2, 'Access to Learned Model of Pick Success'.
domain assumption The local neighborhood argmax objective over F is aligned with physical success.
The method optimizes F's predictions locally; the physical A/B test provides empirical support, but this is an assumption about F's landscape. Invoked in Section 2, 'Optimization Objective'.
ad hoc to paper Gaussian perturbations provide informative gradient estimates.
The paper uses zero-mean Gaussian noise to construct training pairs; the choice of distribution and variances is not justified by theory. Invoked in Section 2, 'Training Data Generation'.
domain assumption The engineered features (package height, plane fitting, cup alignment, adjacency, local height map) suffice to predict optimal adjustments.
The optimizer G uses only these features; visual information is explicitly absent, noted as a limitation in Section 4. Invoked in Section 2, 'Features'.
ad hoc to paper The autoregressive prediction order x, then y, then rotation is a valid decomposition.
No theoretical justification is given for predicting x before y before rz; it is a design choice. Invoked in Section 2, 'Model Architecture and Training'.
domain assumption The test workcells resemble production Robin operations.
The results come from test workcells, not the live fleet; transferability to production is assumed. Invoked in Section 3.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Learning to Optimize Package Picking for Large-Scale, Real-World Robot Induction." pith.science (2026). https://pith.science/paper/X62EMVAH

@misc{pith2026250609765,
  author       = {Pith},
  title        = {Pith review of: Learning to Optimize Package Picking for Large-Scale, Real-World Robot Induction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X62EMVAH}},
  note         = {Machine review of arXiv:2506.09765}
}

read the original abstract

Warehouse automation plays a pivotal role in enhancing operational efficiency, minimizing costs, and improving resilience to workforce variability. While prior research has demonstrated the potential of machine learning (ML) models to increase picking success rates in large-scale robotic fleets by prioritizing high-probability picks and packages, these efforts primarily focused on predicting success probabilities for picks sampled using heuristic methods. Limited attention has been given, however, to leveraging data-driven approaches to directly optimize sampled picks for better performance at scale. In this study, we propose an ML-based framework that predicts transform adjustments as well as improving the selection of suction cups for multi-suction end effectors for sampled picks to enhance their success probabilities. The framework was integrated and evaluated in test workcells that resemble the operations of Amazon Robotics' Robot Induction (Robin) fleet, which is used for package manipulation. Evaluated on over 2 million picks, the proposed method achieves a 20\% reduction in pick failure rates compared to a heuristic-based pick sampling baseline, demonstrating its effectiveness in large-scale warehouse automation scenarios.

Figures

Figures reproduced from arXiv: 2506.09765 by the authors.

**Figure 1.** An example of a Robot induction (Robin) workcell. Robin performs automated package singulation by picking packages from unstructured piles on conveyors and transferring them to mobile robots. Robot induction systems, such as the one shown in [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗

**Figure 2.** Example of an adjacency graph for a cluster of items. Features: The features ϕ(st, a) can be shared between the model of pick success F(ϕ(st, a′ )) and the proposed optimization model that refines pick parameters. We compute a set of features for each induct using scene metadata as well as RGB, and depth images. Specifically, the camera data is processed to generate package segments and tag each segment with an as… view at source ↗

**Figure 3.** Examples of local height maps of packages. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: An illustration of the data generation approach. An initial action ai (blue) and the N generated perturbed actions (green) are shown along with their predicted success probability scores. The vectors correspond to the noise term δi that the proposed model is learning…

**Figure 5.** Figure 5: illustrates the design overview of the described system. Training Data Generation Optimization Model Inference Pick Success Prediction Model F Randomly Sampled Pick Actions Dataset (axi , ayi , azi , ϕi , δaxi , δayi , δazi ) X-axis Prediction Model Gx(ϕi) = δaxi Y -ax…

**Figure 6.** Figure 6: Amazon’s Robin configuration used in our experiments. (a) Arm. (b) Bottom view of the EoAT. Robotic System Architecture The test workcells resemble logistics operations, such as Amazon’s Robin robotic system [1], which has been developed for package manipulation ( […

**Figure 7.** Figure 7: Qualitative examples of pick optimization. The circles show the location of suction cups projected over the scene. Red is for retracted, deactivated suction cups. Green corresponds to extended, activated cups. The probabilities on top of each image correspond to the re…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

URL https://www.amazon.science/latest-news/ robin-deals-with-a-world-where-things-are-changing-all-around-it

Amazon Robotics: Robin (2022). URL https://www.amazon.science/latest-news/ robin-deals-with-a-world-where-things-are-changing-all-around-it

work page 2022
[2]

IEEE TRO30(2), 289–309 (2014)

Bohg, J., Morales, A., Asfour, T., Kragic, D.: Data-driven grasp synthesis—a sur- vey. IEEE TRO30(2), 289–309 (2014)

work page 2014
[3]

IEEE RAL6(4), 8718–8725 (2021)

Cao, H., Fang, H.S., Liu, W., Lu, C.: SuctionNet-1Billion: A large-scale bench- mark for suction grasping. IEEE RAL6(4), 8718–8725 (2021)

work page 2021
[4]

IEEE TASE15, 172–188 (2018)

Correll, N., Bekris, K.E., Berenson, D., Brock, O., Causo, A., Hauser, K., Okada, K., Rodriquez, A., Romano, J.M., Wurman, P.R.: Analysis and observations from the first amazon picking challenge. IEEE TASE15, 172–188 (2018)

work page 2018
[5]

IEEE TRO39(5), 3929–3945 (2023)

Fang, H.S., Wang, C., Fang, H., Gou, M., Liu, J., Yan, H., Liu, W., Xie, Y ., Lu, C.: AnyGrasp: Robust and efficient grasp perception in spatial and temporal domains. IEEE TRO39(5), 3929–3945 (2023)

work page 2023
[6]

In: CVPR, pp

Fang, H.S., Wang, C., Gou, M., Lu, C.: GraspNet-1Billion: A large-scale bench- mark for general object grasping. In: CVPR, pp. 11,441–11,450 (2020)

work page 2020
[7]

In: Robotics: Science and Systems (RSS), pp

Lenz, I., Lee, H., Saxena, A.: Deep learning for detecting robotic grasps. In: Robotics: Science and Systems (RSS), pp. 1–8. Berlin, Germany (2013)

work page 2013
[8]

In: Robotics: Science and Systems (RSS) (2023)

Li, S., Keipour, A., Jamieson, K., Hudson, N., Swan, C., Bekris, K.: Demonstrat- ing large-scale package manipulation via learned metrics of pick success. In: Robotics: Science and Systems (RSS) (2023)

work page 2023

Show all 17 references

[9]

In: Learning Meets Model-based Methods for Manipulation and Grasping Workshop, 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp

Li, S., Keipour, A., Jamieson, K., Hudson, N., Zhao, S., Swan, C., Bekris, K.: Pick planning strategies for large-scale package manipulation. In: Learning Meets Model-based Methods for Manipulation and Grasping Workshop, 2023 IEEE/RSJ International Conference on Intelligent Ro...

2023
[10]

In: ICLR (2023)

Lipman, Y ., Chen, R.T.Q., Ben-Hamu, H., Nickel, M., Le, M.: Flow matching for generative modeling. In: ICLR (2023)

2023
[11]

IEEE TII17(2), 830–838 (2021)

Liu, H., Deng, Y ., Guo, D., Fang, B., Sun, F., et al.: An interactive perception method for warehouse automation in smart cities. IEEE TII17(2), 830–838 (2021)

2021
[12]

In: ICRA, pp

Mahler, J., Matl, M., Liu, X., Li, A., Gealy, D., Goldberg, K.: Dex-Net 3.0: Com- puting robust vacuum suction grasp targets in point clouds using a new analytic model and deep learning. In: ICRA, pp. 5620–5627. Brisbane, Australia (2018)

2018
[13]

IEEE RAM11(4), 110–122 (2004)

Miller, A., Allen, P.K.: Graspit! a versatile simulator for robotic grasping. IEEE RAM11(4), 110–122 (2004)

2004
[14]

In: IROS, pp

Morales, A., Chinellato, E., Fagg, A., del Pobil, A.: Experimental prediction of the performance of grasp tasks from visual features. In: IROS, pp. 3423–3428 (2003)

2003
[15]

IJRR39(2-3), 183–201 (2020)

Morrison, D., Corke, P., Leitner, J.: Learning robust, real-time, reactive robotic grasping. IJRR39(2-3), 183–201 (2020)

2020
[16]

In: Robotics:Science & Systems (RSS) (2025)

Wang, C., Vanbaar, J., Mitash, C., Li, S., Randle, D., Wang, W., Sontakke, S., Bekris, K., Katyal, K.: Demonstrating multi-suction item picking at scale via multi-modal learning pick success. In: Robotics:Science & Systems (RSS) (2025)

2025
[17]

In: CoRL, pp

Yuan, W., Murali, A., Mousavian, A., Fox, D.: M2T2: Multi-task masked trans- former for object-centric pick and place. In: CoRL, pp. 1–12. Atlanta, GA (2023)

2023

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

URL https://www.amazon.science/latest-news/ robin-deals-with-a-world-where-things-are-changing-all-around-it

Amazon Robotics: Robin (2022). URL https://www.amazon.science/latest-news/ robin-deals-with-a-world-where-things-are-changing-all-around-it

work page 2022

[2] [2]

IEEE TRO30(2), 289–309 (2014)

Bohg, J., Morales, A., Asfour, T., Kragic, D.: Data-driven grasp synthesis—a sur- vey. IEEE TRO30(2), 289–309 (2014)

work page 2014

[3] [3]

IEEE RAL6(4), 8718–8725 (2021)

Cao, H., Fang, H.S., Liu, W., Lu, C.: SuctionNet-1Billion: A large-scale bench- mark for suction grasping. IEEE RAL6(4), 8718–8725 (2021)

work page 2021

[4] [4]

IEEE TASE15, 172–188 (2018)

Correll, N., Bekris, K.E., Berenson, D., Brock, O., Causo, A., Hauser, K., Okada, K., Rodriquez, A., Romano, J.M., Wurman, P.R.: Analysis and observations from the first amazon picking challenge. IEEE TASE15, 172–188 (2018)

work page 2018

[5] [5]

IEEE TRO39(5), 3929–3945 (2023)

Fang, H.S., Wang, C., Fang, H., Gou, M., Liu, J., Yan, H., Liu, W., Xie, Y ., Lu, C.: AnyGrasp: Robust and efficient grasp perception in spatial and temporal domains. IEEE TRO39(5), 3929–3945 (2023)

work page 2023

[6] [6]

In: CVPR, pp

Fang, H.S., Wang, C., Gou, M., Lu, C.: GraspNet-1Billion: A large-scale bench- mark for general object grasping. In: CVPR, pp. 11,441–11,450 (2020)

work page 2020

[7] [7]

In: Robotics: Science and Systems (RSS), pp

Lenz, I., Lee, H., Saxena, A.: Deep learning for detecting robotic grasps. In: Robotics: Science and Systems (RSS), pp. 1–8. Berlin, Germany (2013)

work page 2013

[8] [8]

In: Robotics: Science and Systems (RSS) (2023)

Li, S., Keipour, A., Jamieson, K., Hudson, N., Swan, C., Bekris, K.: Demonstrat- ing large-scale package manipulation via learned metrics of pick success. In: Robotics: Science and Systems (RSS) (2023)

work page 2023