Pith. sign in

REVIEW 5 major objections 6 minor 1 cited by

Grammarization-Based Grasping with Deep Multi-Autoencoder Latent Space Exploration by Reinforcement Learning Agent

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Compressing the gripper and target into a shared autoencoder latent space lets an RL agent re-adapt to changes in either one 35.8% faster than exploring the raw environment.

desk verdict Plausible latent-space RL architecture, but the 35.8% adaptation claim is unfalsifiable as reported. read the letter →

arxiv 2411.08566 v2 pith:HMCCESUY submitted 2024-11-13 cs.RO cs.LG

classification cs.ROcs.LG
keywords roboticgraspinglatentspaceexplorationmulti-autoencoderreinforcementlearningPoWERgrammarizationgraspadaptationvoxelrepresentation
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 sets out to show that a robot can adapt its grasps faster when the gripper and target are first compressed into a common latent space by three stacked autoencoders, and the reinforcement-learning agent explores that compressed space instead of the raw environment. The grammarization step encodes geometry plus physical properties such as mass, friction, and inertia, then fuses the target code and gripper code so that one perturbation in the fused space determines both the gripper pose and the target representation. If this holds, a robot facing a new gripper or object would need fewer trials to re-adapt, because the search space is smaller and the reconstructed outputs carry the physical quantities that matter for grasping. The reported headline is a 35.8% faster adaptation rate in simulation, measured as time to regain an 80% grasp-success rate after a gripper or target is changed.

What carries the argument

The carrying object is the fused latent vector $z_C = E_3([z_T, z_G])$, where $z_T$ comes from the target autoencoder AE1 and $z_G$ from the gripper autoencoder AE2, with pose encoded as quaternions inside $z_G$. The agent's action is a perturbation $\delta$ added to $z_C$; decoding through $D_3$ and then through $D_1$ and $D_2$ (Eqs. 21–27) produces the target representation and gripper pose that the simulator evaluates. Positional and pose constraints are imposed as reconstruction losses in AE3, and the reward function balances grasp quality against reconstruction error with reconstruction de-emphasized during exploration. This constraint structure is what is supposed to keep latent-space perturbations physically meaningful enough to execute.

What would settle it

Rerun the altered-gripper and altered-target trials in a physics simulator, decode the RL agent's proposed perturbations into gripper poses, and count how many decoded poses are kinematically feasible and collision-free; if most decoded perturbations produce impossible poses, the exploration is not physically meaningful and the 35.8% gain would not transfer to execution. Alternatively, retrain AE3 with $\alpha$ and $\beta$ set to zero; if the adaptation advantage disappears, the claimed speed-up is carried by an unquantified constraint rather than by latent-space compression.

Watch

Extended reading notes

Core claim

The central claim is that the agent should not explore over raw observations, nor over the separate latent spaces, but over the fused latent vector $z_C$ of a third autoencoder, AE3, whose input is the concatenation of the target code $z_T$ and gripper code $z_G$. A perturbation $\delta$ added to $z_C$ is decoded by AE3 back into separate target and gripper codes, and then by AE1 and AE2 into a reconstructed object shape and gripper pose (Eqs. 21–27). The policy updates use PoWER, a reward-weighted policy-search rule, so the agent learns which perturbations in the compressed space give high grasp rewards. The paper reports that this latent-space agent reaches an 80% grasp-success threshold 35.8% faster than a baseline exploring the observable environment after the same gripper or target alteration, with AE1, AE2, and AE3 reconstruction accuracies of about 90%, 85%, and 71% respectively.

Load-bearing premise

The load-bearing premise is that a perturbation $\delta$ applied to the fused latent vector $z_C$, when decoded back through AE3 and the two lower autoencoders, still yields a gripper pose and target representation that are physically realizable and graspable; the paper relies on reconstruction losses with weights $\alpha$ and $\beta$ that are never quantified, so nothing in the reported experiments directly verifies this physical feasibility.

Editorial extensions

If this is right

  • After a gripper or target is altered, the agent should regain an 80% grasp-success rate in fewer episodes than the baseline, giving the claimed 35.8% faster adaptation.
  • The autoencoders should not need retraining for each new object instance; only the RL policy in the fused space $z_C$ has to adapt.
  • Exploration in a low-dimensional latent code should carry lower computational overhead per trial than exploring raw voxel or image observations.
  • The same stack is intended to support grasp attempts made during execution, where the agent adjusts its grasp online in unpredictable settings.

Reading between the lines

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

  • The 35.8% result is measured against a baseline in the observable environment; comparing against an agent that explores the concatenated latent space $[z_T, z_G]$ before AE3 compression would isolate whether the fusion itself, rather than latent-space exploration in general, drives the gain.
  • If the result transfers to real robots, the computational bottleneck shifts from online RL samples to offline autoencoder training: a shared, pretrained grammarization stack could be reused across tasks and gripper designs.
  • The unquantified weights $\alpha$ and $\beta$ in the AE3 loss suggest a direct test: sweeping those weights, or replacing the soft reconstruction penalty with a hard projection that enforces valid poses after decoding, would show whether the adaptation advantage moves together with decoded-pose feasibility.
  • The paper's own future-work idea of generating a suitable gripper with masked autoencoders follows from the same logic: if the fused space encodes grasp-relevant physics, inpainting the missing gripper entries could propose new end effectors for a given target.
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

5 major / 6 minor

Summary. The paper proposes a robotic grasping framework that compresses high-dimensional target and gripper representations into a common latent space using three autoencoders (AE1 for the target, AE2 for the gripper with pose, and AE3 that fuses the two latent codes). A PoWER reinforcement learning agent then explores perturbations of the fused latent code z_C, and the decoded output is used to generate grasps in a Gazebo simulation. The central reported claim is that this latent-space exploration yields a 35.8% faster re-adaptation after the gripper or target is altered, compared with a baseline that explores the observable environment directly. The paper also reports reconstruction accuracies for the three autoencoders and frames the contribution as 'grammarization' of grasping components.

Significance. If the central claim were adequately supported, the idea of learning in a fused latent space that jointly encodes target geometry, physical properties, and gripper pose could be a useful direction for sample-efficient grasping, especially in domains where rapid adaptation is required. The paper has some strengths: it gives explicit loss functions for the three autoencoders, uses a well-known policy search algorithm (PoWER), and describes a concrete simulation protocol. However, the main empirical result is currently not substantiated: the baseline is not specified, no variance or seed information is reported, and there is an internal inconsistency in the reported AE3 accuracy. As written, the central quantitative claim is not reproducible or falsifiable from the manuscript alone, which limits the significance of the contribution in its present form.

major comments (5)
  1. [Section IV-B, Table I] The central claim of a 35.8% faster adaptation rate is not supported as reported. The baseline is described only as 'the baseline methodology exploring on the observable environment'; no algorithm, policy parameterization, reward function, episode budget, or environment conditions are specified for this baseline, and no absolute adaptation times are given. The result also lacks the number of seeds, error bars, and any statistical test. Because the paper's contribution is precisely this empirical speedup, the missing experimental specification makes the claim unfalsifiable from the paper alone. Please specify the baseline completely, report mean and variance over multiple seeds, and include absolute convergence times.
  2. [Section V vs Table I] The AE3 reconstruction accuracy is reported as 71.16% in Table I but as 79% in Section V ('AE3 achieved 79%'). This internal inconsistency suggests that the experimental pipeline was not carefully controlled and must be resolved. In addition, the Discussion states that 'the PoWER algorithm enhanced RL training efficiency by 35%' while Table I reports a 35.8% improvement; clarify whether these refer to the same measurement and, if so, why the numbers differ.
  3. [Section III-D, Eqs. (18), (23), (28)] Several parameters that directly determine the reported learning behavior are not reported: the reward weights α and β in Eq. (23), the AE3 loss weights α and β in Eq. (18), the stopping threshold Rsuccess in Eq. (28), and the latent dimensions m_T, m_G, and m_C. Without these values the experiments cannot be reproduced, and the claim that α and β are 'small weights' is unverifiable. Please report all hyperparameter values and, ideally, a sensitivity analysis for α and β.
  4. [Section III-D, Eqs. (25)-(27)] The mechanism that is supposed to ensure that decoded latent perturbations yield physically feasible grasps is only enforced through reconstruction penalties (Eq. (25)), but no evidence is provided that during RL exploration the decoded pose vectors (position and quaternion) remain within executable ranges, or that the reconstructed target and gripper representations are valid inputs for grasp execution. An analysis of pose reconstruction errors during exploration, or an ablation without the constraints, is needed to support the claim that exploration in AE3's latent space preserves pose integrity.
  5. [Abstract and Section IV] The abstract claims that the approach achieves 'higher learning rates at the initial stages of exploration of a new environment, as well as at non-zero shot grasp attempts', but the experiments in Section IV only measure re-adaptation after altering the gripper or target. No experimental evaluation of non-zero-shot grasp attempts or initial exploration in a new environment is reported. If this claim is retained, it needs a dedicated experiment with a clear protocol.
minor comments (6)
  1. [Section I-A] There are typos in the definition of grammarization: 'not not necessarily' and 'copmression' should be corrected, and the notation for the grammar rules θ_i is inconsistent with the subsequent use of θ as a policy parameter in Eq. (24).
  2. [Section III-A] The phrase 'constructed by 10cm from which assumed manufacturing proceedures have been applied according to [30]' in Section IV-A is unclear; please rewrite for precision.
  3. [Section III-C] The notation 'zG[i] − →ˆzG[i]' and similar arrows is nonstandard and should be replaced with a clear mathematical statement, e.g., zG[i] ≈ ˆzG[i].
  4. [References] The reference list contains duplicates: [20] duplicates [11], [21] duplicates [25], [23] duplicates [11], and [26] duplicates [22]. Please remove or merge duplicates.
  5. [Section V] In the last paragraph of the Discussion, 'A2' should read 'AE2' (the second autoencoder).
  6. [Figures 1-3] The figure captions only say 'Qualitative representation'; they do not provide network architecture details such as layer sizes, latent dimensions, or filter counts. Adding these details would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the claimed 35.8% adaptation improvement is measured against an external grasp simulator, not forced by the paper's loss equations or self-citations.

full rationale

The paper's derivation chain is not circular. The load-bearing empirical claim is that an RL agent exploring AE3's latent space adapts 35.8% faster than a baseline exploring the observable environment; this comparison is made in Gazebo and evaluated by grasp success, stability, and force, i.e., an external reward signal not defined by the autoencoder losses. The reconstruction losses in Eqs. (15)-(18) and the reconstruction penalty in Eq. (23) are training and regularization objectives; they constrain but do not generate the reported speedup. No self-citation is used to justify a central premise; the FeatureNet citation [30] is an external architecture and dataset source, and PoWER [31] is an external algorithm. 'Grammarization' is presented as a definition (Sec. I-A), not derived from the target result, so calling it a rename of autoencoder compression is a framing concern, not a circular reduction. The paper's weaknesses, including unquantified alpha and beta weights, training-set-only accuracy figures, the inconsistent AE3 accuracy values (71.16% vs 79%), and an underspecified baseline, are correctness and reproducibility risks rather than circularity.

Assumptions & free parameters 5 free parameters · 3 assumptions · 1 invented entities

The central claim is an empirical simulation result. The loss and reward equations introduce hand-chosen weights α and β; the stopping threshold Rsuccess and latent dimensions are unspecified. The main assumption, that latent-space perturbations decode to feasible grasps, is unvalidated. No code or data are provided.

free parameters (5)
  • alpha (reconstruction weight in AE3 loss and reward)
    Appears in Eq. (18) and Eq. (23); hand-chosen small weight, no value reported.
  • beta (pose weight in AE3 loss and reward)
    Appears in Eq. (18) and Eq. (23); hand-chosen, no value reported.
  • eta (PoWER learning rate)
    Eq. (24); no value reported.
  • Rsuccess (success threshold for stopping)
    Eq. (28); defined as a threshold but not quantified.
  • latent dimensions m_T, m_G, m_C
    Chosen by hand; the paper states higher dimensions were 'found to be optimal' for AE2 and AE3 without giving values.
assumptions (3)
  • domain assumption Autoencoder reconstruction loss preserves grasp-relevant information
    Central to AE1 and AE2 training (Section III-A, Eq. 1; Section III-B, Eq. 7); no independent validation that reconstruction accuracy correlates with grasp success.
  • domain assumption Perturbations in the fused latent space decode to physically feasible grasps
    Section III-D, Eqs. (21)-(27); assumes that decoding z_C + δ via AE3, AE1, and AE2 yields executable gripper poses and target states.
  • domain assumption Simulation in Gazebo is a faithful proxy for real grasping
    Section IV-B; all results are simulated, no real-robot validation.
invented entities (1)
  • Grammarization function g
    purpose: Abstract physical object and gripper features into a lower-dimensional metric space
    Defined in Section I-A; equivalent to standard autoencoder dimensionality reduction, no new physical content or falsifiable prediction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Grammarization-Based Grasping with Deep Multi-Autoencoder Latent Space Exploration by Reinforcement Learning Agent." pith.science (2026). https://pith.science/paper/HMCCESUY

@misc{pith2026241108566,
  author       = {Pith},
  title        = {Pith review of: Grammarization-Based Grasping with Deep Multi-Autoencoder Latent Space Exploration by Reinforcement Learning Agent},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HMCCESUY}},
  note         = {Machine review of arXiv:2411.08566}
}
read the original abstract

Grasping by a robot in unstructured environments is deemed a critical challenge because of the requirement for effective adaptation to a wide variation in object geometries, material properties, and other environmental factors. In this paper, we propose a novel framework for robotic grasping based on the idea of compressing high-dimensional target and gripper features in a common latent space using a set of autoencoders. Our approach simplifies grasping by using three autoencoders dedicated to the target, the gripper, and a third one that fuses their latent representations. This allows the RL agent to achieve higher learning rates at the initial stages of exploration of a new environment, as well as at non-zero shot grasp attempts. The agent explores the latent space of the third autoencoder for better quality grasp without explicit reconstruction of objects. By implementing the PoWER algorithm into the RL training process, updates on the agent's policy will be made through the perturbation in the reward-weighted latent space. The successful exploration efficiently constrains both position and pose integrity for feasible executions of grasps. We evaluate our system on a diverse set of objects, demonstrating the high success rate in grasping with minimum computational overhead. We found that approach enhances the adaptation of the RL agent by more than 35 % in simulation experiments.

Figures

Figures reproduced from arXiv: 2411.08566 by the authors.

Figure 1
Figure 1. Qualitative representation of AE1 The latent space produced by AE1 captures a compressed representation of the target object, retaining key features such as shape, surface contours, and physical-related properties. Let X ∈ R n×n×n represent the voxelized object shape, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Qualitative representation of AE2 gripper in a compact form. This vector will later be used as part of the input to AE3. Note that the optimal latent space has been found to be optimal in higher dimensions than AE1 because of the encoding of the extra not so correletable parameters of the pose vector. Latent Space and Reconstruction: The latent space of AE2 is designed to allow both the reconstruction of the grip￾pe… view at source ↗
Figure 3
Figure 3. Qualitative representation of AE3 3) Decoding and Reconstruction: The decoder of AE3 maps the compressed latent vector zC back to the concatenated latent space zˆGT , which is then split into the latent spaces zˆG and zˆT for the gripper and the target, respectively: zˆGT = D3(zC ) (13) Where D3 represents the decoder network of AE3. The latent spaces are then passed back to the decoders of AE1 and AE2 for the final… view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Continuous In-Situ and Remote Sun Observation for Space Weather Monitoring and Mitigation of Infrastructure Threats Through an Optimized Heliocentric Satellite Constellation

    physics.space-ph 2024-11 reject novelty 5.0 of 10

    A six-spacecraft heliocentric Elliptical Walker constellation is claimed to enable continuous 4π Sun coverage, in-situ measurements, and near-continuous Sun-Earth line monitoring, but the supporting demonstration is l...

Reference graph

Works this paper leans on

30 extracted references · 13 canonical work pages · cited by 1 Pith paper

  1. [1]

    Data-driven grasp synthesis—A survey,

    J. Bohg, A. Morales, T. Asfour, and D. Kragic, “Data-driven grasp synthesis—A survey,” Proceedings of the IEEE , vol. 102, no. 12, pp. 1920–1933, 2014, doi: 10.1109/JPROC.2014.2303973

  2. [2]

    Trends and challenges in robot manipula- tion,

    A. Billard and D. Kragic, “Trends and challenges in robot manipula- tion,” Autonomous Robots , vol. 43, no. 5, pp. 1071–1104, 2019, doi: 10.1007/s10514-019-09857-4

  3. [3]

    Prehensile pushing: In-hand manipulation with push-primitives,

    N. Chavan-Dafle and A. Rodriguez, “Prehensile pushing: In-hand manipulation with push-primitives,” in Proc. IEEE Int. Conf. Robotics and Automation (ICRA) , 2015, pp. 6215–6222, doi: 10.1109/ICRA.2015.7139801

  4. [4]

    Robot learning of manipulation actions with deep reinforcement learning based on object-oriented state representation,

    J. Chen, X. Lan, H. Ding, and J. Qian, “Robot learning of manipulation actions with deep reinforcement learning based on object-oriented state representation,” Robotics and Autonomous Systems , vol. 115, pp. 1–11, 2019, doi: 10.1016/j.robot.2018.09.012

  5. [5]

    A survey of multi-robot interaction research and its applications,

    X. Cheng, M. Otte, and E. Frazzoli, “A survey of multi-robot interaction research and its applications,” in Proc. IEEE Int. Conf. Robotics and Automation (ICRA) , 2019, pp. 8734–8741, doi: 10.1109/ICRA.2019.8794312

  6. [6]

    Robot grasping in cluttered environments: Towards robust 3D object recognition with noise resilience,

    A. Jain, J. Van Baar, M. Popovic, and A. Andonian, “Robot grasping in cluttered environments: Towards robust 3D object recognition with noise resilience,” in Proc. IEEE Int. Conf. Robotics and Automation (ICRA) , 2020, pp. 914–921, doi: 10.1109/ICRA.2020.9144917

  7. [7]

    Reinforcement learning in la- tent action sequence space,

    H. Kim, J. Park, and K. Kim, “Reinforcement learning in la- tent action sequence space,” in Proc. IEEE/RSJ Int. Conf. In- telligent Robots and Systems (IROS) , 2020, pp. 6765–6772, doi: 10.1109/IROS45743.2020.9341619

  8. [8]

    A review of robot learn- ing for manipulation: Challenges, representations, and algorithms,

    O. Kroemer, S. Niekum, and G. Konidaris, “A review of robot learn- ing for manipulation: Challenges, representations, and algorithms,” Proceedings of the IEEE , vol. 109, no. 4, pp. 873–890, 2021, doi: 10.1109/JPROC.2021.3050175

Show all 30 references
  1. [9]

    Robot skill learning in the latent space of a deep autoencoder,

    S. Pande, A. Mathew, and S. Gupta, “Robot skill learning in the latent space of a deep autoencoder,” in Proc. IEEE/RSJ Int. Conf. Intelligent Robots and Systems (IROS) , 2019, pp. 4089–4096, doi: 10.1109/IROS40897.2019.8967977

  2. [10]

    Generating multi-fingered robotic grasps via deep learning,

    J. Varley, J. Weisz, J. Weiss, and P. K. Allen, “Generating multi-fingered robotic grasps via deep learning,” IEEE Trans. Robotics , vol. 33, no. 5, pp. 1183–1194, 2017, doi: 10.1109/TRO.2017.2651055

  3. [12]

    Deep reinforcement learning based moving object grasping,

    Z. Chen, Z. Wei, and Y . Shi, “Deep reinforcement learning based moving object grasping,” IEEE Trans. Cognitive and Developmental Systems , 2021, doi: 10.1109/TCDS.2020.2968103

  4. [13]

    Visual-based robotic grasping under unstructured environment,

    M. Wang, Y . Zhang, and D. Jiang, “Visual-based robotic grasping under unstructured environment,” J. Visual Communication and Image Repre- sentation, vol. 71, p. 102800, 2020, doi: 10.1016/j.jvcir.2020.102800

  5. [14]

    Data-efficient deep rein- forcement learning for dexterous manipulation,

    I. Popov, N. Heess, T. Lillicrap, R. Hafner, G. Barth-Maron, M. Vecerik, T. Lampe, Y . Tassa, and M. Riedmiller, “Data-efficient deep rein- forcement learning for dexterous manipulation,” Proc. Conf. Robotics: Science and Systems (RSS) , 2017, doi: 10.15607/RSS.2017.XIII.022

  6. [15]

    Robotic grasping using deep rein- forcement learning,

    V . Joshi, Z. Xue, and D. Kumar, “Robotic grasping using deep rein- forcement learning,” Int. J. Advanced Robotic Systems , vol. 17, no. 1, pp. 1–12, 2020, doi: 10.1177/1729881420905726

  7. [16]

    Learning the latent space of robot dynamics for cutting interaction inference,

    M. Rezaei-Shoshtari, H. I. Bozma, and O. Korkmaz, “Learning the latent space of robot dynamics for cutting interaction inference,” IEEE Robotics and Automation Letters , vol. 5, no. 2, pp. 3484–3491, 2020, doi: 10.1109/LRA.2020.2976994

  8. [17]

    CVML-Pose: Convolutional AE based multi-level network for object 3D pose estimation,

    Y . Zhao, Q. Wang, M. Li, and H. Sun, “CVML-Pose: Convolutional AE based multi-level network for object 3D pose estimation,” Neurocomput- ing, vol. 507, pp. 413–426, 2023, doi: 10.1016/j.neucom.2022.07.020

  9. [18]

    PoseRBPF: A Rao–Blackwellized particle filter for 6-D object pose tracking,

    X. Deng, A. Mousavian, and D. Fox, “PoseRBPF: A Rao–Blackwellized particle filter for 6-D object pose tracking,” IEEE Robotics and Automation Letters , vol. 4, no. 4, pp. 4250–4257, 2019, doi: 10.1109/LRA.2019.2932904

  10. [19]

    Neighborhood geometric structure- preserving variational autoencoder for smooth and bounded data sources,

    S. Chen, H. Zhu, and J. Zhang, “Neighborhood geometric structure- preserving variational autoencoder for smooth and bounded data sources,” IEEE Trans. Neural Networks and Learning Systems , vol. 32, no. 7, pp. 3120–3133, 2021, doi: 10.1109/TNNLS.2020.3026338

  11. [23]

    6-DOF GraspNet: Variational Grasp Generation for Object Manipulation,

    A. Mousavian, C. Eppner, and D. Fox, “6-DOF GraspNet: Variational Grasp Generation for Object Manipulation,” 2019 IEEE/CVF Interna- tional Conference on Computer Vision (ICCV) , pp. 2901–2910, 2019, doi: 10.1109/ICCV .2019.00299

  12. [24]

    Contact- GraspNet: Efficient 6-DoF Grasp Generation in Cluttered Scenes,

    M. Sundermeyer, A. Mousavian, R. Triebel, and D. Fox, “Contact- GraspNet: Efficient 6-DoF Grasp Generation in Cluttered Scenes,” 2021 IEEE International Conference on Robotics and Automation (ICRA) , pp. 13438–13444, 2021, doi: 10.1109/ICRA48506.2021.9561877

  13. [25]

    A model-free 6- DOF grasp detection method based on point clouds of local sphere area,

    J. Sun, K. Zhang, G. Yang, and J. Chu, “A model-free 6- DOF grasp detection method based on point clouds of local sphere area,” Advanced Robotics , vol. 37, pp. 679–690, 2023, doi: 10.1080/01691864.2023.2197961

  14. [26]

    Efficient Heatmap-Guided 6-DoF Grasp Detection in Cluttered Scenes,

    S. Chen, W. N. Tang, P. Xie, W. Yang, and G. Wang, “Efficient Heatmap-Guided 6-DoF Grasp Detection in Cluttered Scenes,” IEEE Robotics and Automation Letters , vol. 8, pp. 4895–4902, 2023, doi: 10.1109/LRA.2023.3290513

  15. [27]

    LieGrasPFormer: Point Transformer- Based 6-DOF Grasp Detection with Lie Algebra Grasp Rep- resentation,

    J. Lin, M. Rickert, and A. Knoll, “LieGrasPFormer: Point Transformer- Based 6-DOF Grasp Detection with Lie Algebra Grasp Rep- resentation,” 2023 IEEE 19th International Conference on Au- tomation Science and Engineering (CASE) , pp. 1–7, 2023, doi: 10.1109/CASE56687.2023.10260543

  16. [28]

    Keypoint-GraspNet: Keypoint-based 6- DoF Grasp Generation from the Monocular RGB-D input,

    Y . Chen, Y . Lin, and P. Vela, “Keypoint-GraspNet: Keypoint-based 6- DoF Grasp Generation from the Monocular RGB-D input,” ArXiv, vol. abs/2209.08752, 2022, doi: 10.48550/arXiv.2209.08752

  17. [29]

    GraspNeRF: Multiview-based 6-DoF Grasp Detection for Transparent and Specular Objects Using Generalizable NeRF,

    Q. Dai, Y . Zhu, Y . Geng, C. Ruan, J. Zhang, and H. Wang, “GraspNeRF: Multiview-based 6-DoF Grasp Detection for Transparent and Specular Objects Using Generalizable NeRF,” ArXiv, vol. abs/2210.06575, 2022, doi: 10.48550/arXiv.2210.06575

  18. [30]

    FeatureNet: Machining feature recognition based on 3D Convolutional Neural Network,

    Z. Zhang, J. Li, and X. Li, “FeatureNet: Machining feature recognition based on 3D Convolutional Neural Network,” Computer-Aided Design , vol. 115, pp. 89–102, 2019, doi: 10.1016/j.cad.2019.05.003

  19. [31]

    Policy search for motor primitives in robotics,

    J. Kober and J. Peters, “Policy search for motor primitives in robotics,” Machine Learning , vol. 84, no. 1, pp. 171–203, 2011, doi: 10.1007/s10994-010-5223-6

  20. [32]

    Estimating the center of mass of an object with nonuniform density using robotic pushing,

    H. Gao, S. Zhang, and M. Tomizuka, “Estimating the center of mass of an object with nonuniform density using robotic pushing,” 2023 IEEE International Conference on Robotics and Automation (ICRA) , pp. 4512–4518, 2023, doi: 10.1109/ICRA.2023.4512

  21. [33]

    Vision-based moment of inertia estimation for noncooperative space objects,

    Z. Xu and X. Wang, “Vision-based moment of inertia estimation for noncooperative space objects,” IEEE Transactions on Aerospace and Electronic Systems , vol. 53, no. 4, pp. 1902–1914, 2017, doi: 10.1109/TAES.2017.2684760

  22. [34]

    Reducing the dimensionality of data with neural networks,

    G. E. Hinton and R. R. Salakhutdinov, “Reducing the dimensionality of data with neural networks,” Science, vol. 313, no. 5786, pp. 504-507, 2006, doi: 10.1126/science.1127647

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.