Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation

T0 review · 4 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read SYMDEX claims a robot's bilateral symmetry can be encoded as an equivariance constraint, letting one arm's experience transfer to the other and yielding a single ambidextrous policy.

desk verdict SYMDEX is a genuinely useful framework—per-subtask equivariant policies distilled into a hand-agnostic policy—but the 80% success claim doesn't match its own Table 1, and the four-arm result is too thin to carry much weight. read the letter →

arxiv 2505.05287 v2 pith:4SRJK37D submitted 2025-05-08 cs.RO cs.LG

Zechu Li , Yufeng Jin , Daniel Ordonez Apraez , Claudio Semini , Puze Liu , Georgia Chalvatzaki This is my paper · ORCID
classification cs.ROcs.LG
keywords bimanualdexterousmanipulationreinforcementlearningequivariantneuralnetworksmorphologicalsymmetrybilateralsim-to-realtransfermulti-armpolicydistillation
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

This paper asks whether a bimanual robot's inherent bilateral symmetry can be turned into a learning advantage: instead of training one monolithic controller for both hands, each hand learns its subtask with a symmetry-constrained policy, so experience gained by one arm is automatically reused by the mirrored other. On six simulated dexterous manipulation tasks the paper reports success rates above 80% for all tasks, with the largest gains on tasks where the two hands play distinct roles, and demonstrates zero-shot transfer to a real two-arm platform on two tasks. The significance is practical: if the claim holds, bilateral symmetry becomes a structural prior that roughly halves exploration, removes fixed hand-to-task assignment, and extends the same recipe to multi-arm systems with rotational rather than reflectional symmetry.

What carries the argument

The central object is the group action of the morphological symmetry group on the multi-task multi-agent POMDP that models the bimanual system. For a bimanual robot the group is the reflection group $C_2=\{e,g_r\}$, acting on states, actions, observations, and on the set of agent–task pairs. The load-bearing identity is G-equivariance of the optimal policy, $g\triangleright_A\pi^*(\sigma(s))=\pi^*(\sigma(g\triangleright_S s))$, together with G-invariance of the optimal value function, which holds when the transition kernel, reward, and initial-state distribution are all G-invariant. Subtask policies are parameterized as G-equivariant neural networks, which makes the action of one arm in the reflected workspace equal to the transformed action of the opposite arm; teacher–student distillation then produces a global equivariant policy that is independent of hand-to-task assignment.

What would settle it

Measure the success rate on a task whose two hands are deliberately made asymmetric, for instance by attaching a small extra mass to one arm only, while keeping the reward symmetric. If the equivariant policy succeeds on the unperturbed arm's configuration but systematically fails on the mirrored configuration, the $G$-invariance of the transition kernel in Eq. (2) is violated and the equivariance guarantee breaks.

Watch

Extended reading notes

Core claim

The paper claims that a robot's bilateral morphological symmetry can be used as a structural inductive bias to learn ambidextrous bimanual manipulation without fixed left/right role assignment. In SYMDEX, a bimanual task is first decomposed into per-hand subtasks, each trained by a separate policy whose neural network is G-equivariant under the reflection group $C_2$, so experience from one arm transfers to the mirrored arm by construction. The per-subtask policies are then distilled into a single global equivariant policy that is ambidextrous by design and trained on non-privileged observations for zero-shot real-world deployment. The paper reports success rates above 80% on all six simulated tasks, outperforming five baselines, successful real-world transfer on box-lift and table-clean, and extension to a four-arm setup with $C_4$ symmetry.

Load-bearing premise

The load-bearing premise is that swapping left and right hands, that is mirroring the workspace, leaves the task's physics, rewards, and starting conditions unchanged, while in reality the arms are identical rather than true mirror copies, so mass distribution and collision avoidance may not follow the reflection exactly.

Editorial extensions

If this is right

  • Each arm's collected experience automatically populates the mirrored arm's training data through $C_2$-equivariance, roughly halving the exploration burden.
  • Fixed hand-to-task assignments become unnecessary: the distilled global policy selects which arm performs which subtask from the scene, which is what ambidextrous means here.
  • Decomposing a bimanual task into per-hand subtask policies reduces each agent's action space from 44 to 22 DoF and gives each policy a single reward, sidestepping the credit-assignment failure of monolithic policies.
  • The same equivariance recipe carries over to larger finite symmetry groups, as demonstrated by a four-arm task under $C_4$ rotation symmetry.
  • With curriculum-based domain randomization and non-privileged observations, the distilled policy transfers to the physical robot without additional fine-tuning on two tasks.

Reading between the lines

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

  • Editorial inference: the recipe should transfer to any robot whose dynamics are approximately symmetric, including humanoid left/right limbs, as long as the task rewards are pose- and kinematic-dominated; the hard part is checking whether the dynamics are symmetric enough.
  • Editorial inference: the paper's comparison suggests that architectural equivariance beats on-policy data augmentation for morphological symmetry; a natural test is whether this ordering persists with much larger replay buffers or longer augmentation schedules.
  • Editorial inference: because the student is trained on teacher-generated data, the equivariant Gaussian student's edge over the diffusion student in the real world implies the teacher's distribution is narrow; collecting more diverse or on-policy student data could recover diffusion-level robustness.
  • Editorial inference: the failure mode identified in real-world runs is pose tracking, not control; an equivariant vision encoder, the paper's stated future direction, could make the pipeline robust under occlusion and rapid motion.
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.

Referee Report

4 major / 5 minor

Summary. The paper proposes SYMDEX, a reinforcement learning framework for bimanual and multi-arm manipulation that exploits bilateral symmetry as an inductive bias. The method decomposes a bimanual task into per-hand subtasks, learns a G-equivariant policy per subtask, and distills these into a single global equivariant policy intended to be ambidextrous and deployable zero-shot. The authors evaluate the approach on six simulated bimanual tasks, report real-world deployment on two tasks, and extend the framework to a four-arm C4-symmetric setup. The central claims are that SYMDEX consistently exceeds 80% success on all six tasks, outperforms several baselines, and that the distilled global policy is ambidextrous by design.

Significance. If confirmed, the paper would provide a practically useful demonstration that morphological symmetry can serve as an inductive bias for high-dimensional, contact-rich bimanual manipulation, reducing the need for fixed hand-task assignment and improving sample efficiency and robustness. The paper has notable strengths: it builds on standard POMDP symmetry theory, includes five controlled baselines and several ablations, and reports real-world trials, which is more than many simulation-only manipulation papers provide. The central risk is whether the physical robot actually satisfies the symmetry assumptions on which the guarantees rest, and whether the reported numbers support the strongest claims in the text.

major comments (4)
  1. [Section 4.1 / Table 1] The Abstract and Section 4.1 state that SYMDEX 'consistently learns all six tasks with success rates exceeding 80%.' Table 1 (top) reports the distilled global policy—the component claimed to be ambidextrous by design and zero-shot deployable—at 0.63 for the equivariant Gaussian policy and 0.60 for the equivariant diffusion policy on Threading, both below 80%. The same paragraph also states that both equivariant Gaussian and diffusion policies outperform the vanilla Gaussian policy across all six tasks, but on Threading the vanilla Gaussian policy achieves 0.62 while the equivariant diffusion policy achieves 0.60. Please specify which policy the 80% claim refers to, reconcile the Table 1 numbers with the text, and clarify the discrepancy.
  2. [Section 6 / Eqs. (1)-(3)] The equivariance guarantees in Eqs. (1)-(3) and Eq. (9) require the transition kernel, reward, and initial-state distribution to be G-invariant. Section 6 states that the arms are 'identical in construction rather than mirrored,' so mass distribution and collision avoidance may not follow reflectional symmetry. This means the central symmetry assumption is approximate for the physical system, and the assertion that this 'does not significantly impact its effectiveness in practice' is not backed by a direct measurement. Please add a quantitative check, such as comparing left- and right-arm transition dynamics under the reflection map or measuring the symmetry error in collision geometry, or explicitly qualify the equivariance guarantee as approximate.
  3. [Section 4.3] The four-arm C4 experiment is presented as evidence of scalability, but Fig. 9 reports only success rates for SYMDEX across the four symmetry elements, with no comparison against a non-equivariant baseline, no error bars, and no statement of the number of rollouts or seeds. Since 'scalability to multi-arm tasks' is one of the three stated contributions, the evidence should meet the same standard as the bimanual experiments. Please add a baseline comparison and report variance.
  4. [Section 4.2 / Table 1] The real-world evaluation is described in Section 4 as reporting 'both overall task and per-hand subtask success over 30 independent trials,' but the caption of Table 1 says the results were 'assessed through qualitative evaluations by human operators.' These statements describe different protocols. Please state the actual evaluation protocol, the number of trials, how the reported standard deviations were computed, and whether the evaluation was automated or operator-assessed.
minor comments (5)
  1. [Section 3, Eq. (4)] The text says the reflection leads to a 'permutation of agents and tasks,' but the displayed group action on agent-task pairs acts only on the task component and leaves the agent label unchanged; please clarify the intended group action on the agent-task pair set.
  2. [Abstract and Section 4.1] The Abstract contains the grammatical error 'complex task' where 'complex tasks' is intended, and Section 4.1 uses 'deglobal value functions,' which appears to be a typo for 'decentralized' or 'decomposed' value functions.
  3. [Appendix H] The text says all methods share identical hyperparameters but then lists a per-task entropy coefficient; please state explicitly whether the same per-task entropy values were used for all baselines.
  4. [Section 4.3] The four-arm system is described as 'each equipped with a right dexterous hand'; under a 90-degree rotational symmetry, please clarify how handedness transforms under the C4 action, since the Section 6 concern about non-mirrored construction may apply.
  5. [Appendix F, Eq. (20)] The confidence score in Eq. (20) uses a normalization by the sum of the foreground mask but the summation variable and the meaning of N are not fully specified; please define all symbols.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the equivariance constraints are imposed by construction or imported from external MDP-symmetry theory, and all performance claims are measured against external baselines.

full rationale

The paper's derivation chain is self-contained rather than circular. The symmetric-POMDP formalism and the optimal-policy equivariance property in Eqs. (1)-(3) are standard results taken from external literature [20, 32], and the paper explicitly states the invariance conditions on transition, reward, and initial-state distribution that must hold for the equivariance guarantees to apply. The distilled global policy's equivariance constraint in Eq. (9) is a design constraint imposed on the policy architecture, not a quantity fitted to data and then repackaged as a prediction; the claim that the policy is "ambidextrous by design" is therefore definitional in a legitimate sense. Empirical results are reported as measured success rates against five external baselines with shared hyperparameters, and no fitted constant is renamed as a prediction. The citation to the authors' prior morphological-symmetry framework [1] is used to supply group-action construction details, but this is a methodological implementation detail, the cited work is externally published, and no load-bearing conclusion reduces to an unverified self-citation. The Section 6 limitation that the arms are "identical in construction rather than mirrored" weakens the symmetry assumption and is a correctness risk, but it is not a circularity: the paper acknowledges the approximation instead of using the assumption to define the result.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The method rests on the standard assumption that the manipulation MDP is invariant under the bilateral reflection group, plus an equivariant observation function. Real-world arms are not exact dynamic mirrors, so the symmetry assumption is approximate. Training also uses per-task entropy coefficients, reward weights, curriculum thresholds, and penalty coefficients that are chosen by hand rather than derived. No new physical entities are introduced.

free parameters (5)
  • Per-task entropy coefficient = 0.0, 0.005, 0.01, 0.01, 0.01, 0.005 for Box-lift, Table-clean, Drawer-insert, Threading, Bowl-stir, Handover
    Tuned per task in Table 4; all methods share the same value per task, but it is still a hand-chosen number rather than a derived constant.
  • Reward term weights = not reported
    Appendix E describes reward components qualitatively, such as alignment, distance, and success bonus, without giving the numerical weights used in the PPO loss; these weights are free parameters of the training objective.
  • Curriculum success threshold = 0.7
    The curriculum advances only when success rate exceeds 0.7, chosen by hand in Table 3.
  • Safety penalty coefficients = collision -1000.0, energy -0.001
    Phased in late during curriculum learning, as shown in Table 5; these are hand-chosen penalty scales.
  • Domain randomization ranges = object mass, position, orientation, friction, restitution, noise ranges in Table 5
    Selected empirically to make sim-to-real transfer work; not derived from any first-principles criterion.
assumptions (5)
  • domain assumption The manipulation MDP is G-symmetric: transition kernel, reward, and initial-state distribution are invariant under the bilateral reflection group C2 (Eq. 2).
    Required for the equivariance constraints in Eq. (3) and for Eq. (7) through Eq. (9); real robot arms are not exact dynamic mirrors, as acknowledged in the Limitations section.
  • domain assumption Each subtask reward is G-invariant (Section 3, before Eq. 8).
    The paper asserts this holds naturally because rewards depend on hand-object pose errors; if a reward term depended on absolute body coordinates, the equivariant value function would not be justified.
  • domain assumption The observation function sigma is G-equivariant (Note 2.1 and Proposition B.1).
    Policy equivariance only follows if observations transform equivariantly; the construction is delegated to reference [1] and is not verified for the real-world perception stack.
  • domain assumption Kinematic-level symmetry is sufficient for transfer even though full dynamic symmetry is absent.
    The Limitations section states that arms are identical rather than mirrored, so mass distribution and collision avoidance do not fully follow reflectional symmetry; the method assumes this does not matter in practice.
  • domain assumption The simulator plus curriculum approximates real-world dynamics closely enough for zero-shot transfer.
    Real-world deployment relies on domain randomization and safety penalties from the curriculum rather than real-world fine-tuning.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation." pith.science (2026). https://pith.science/paper/4SRJK37D

@misc{pith2026250505287,
  author       = {Pith},
  title        = {Pith review of: Morphologically Symmetric Reinforcement Learning for Ambidextrous Bimanual Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4SRJK37D}},
  note         = {Machine review of arXiv:2505.05287}
}
read the original abstract

Humans naturally exhibit bilateral symmetry in their gross manipulation skills, effortlessly mirroring simple actions between left and right hands. Bimanual robots-which also feature bilateral symmetry-should similarly exploit this property to perform tasks with either hand. Unlike humans, who often favor a dominant hand for fine dexterous skills, robots should ideally execute ambidextrous manipulation with equal proficiency. To this end, we introduce SYMDEX (SYMmetric DEXterity), a reinforcement learning framework for ambidextrous bi-manipulation that leverages the robot's inherent bilateral symmetry as an inductive bias. SYMDEX decomposes complex bimanual manipulation tasks into per-hand subtasks and trains dedicated policies for each. By exploiting bilateral symmetry via equivariant neural networks, experience from one arm is inherently leveraged by the opposite arm. We then distill the subtask policies into a global ambidextrous policy that is independent of the hand-task assignment. We evaluate SYMDEX on six challenging simulated manipulation tasks and demonstrate successful real-world deployment on two of them. Our approach strongly outperforms baselines on complex task in which the left and right hands perform different roles. We further demonstrate SYMDEX's scalability by extending it to a four-arm manipulation setup, where our symmetry-aware policies enable effective multi-arm collaboration and coordination. Our results highlight how structural symmetry as inductive bias in policy learning enhances sample efficiency, robustness, and generalization across diverse dexterous manipulation tasks.

Figures

Figures reproduced from arXiv: 2505.05287 by the authors.

Figure 1
Figure 1. Overview of SYMDEX: (Left) Digital twin of our bimanual robot. (Middle) The task is [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. A comparison of action execution between (a) subtask policies and (b) global policy, where [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Our benchmark of six bimanual dexterous manipulation tasks with diverse levels of coop [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Performance of SYMDEX and baseline methods on six benchmark tasks. SYMDEX [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: The four-arm system setup. We demonstrate the scalability of SYMDEX on a multi-robot task in￾volving a system of four arms, each equipped with a right dexterous hand. The objective is for two arms to hold the flaps of a cardboard box while the other two arms pick up ob…
Figure 6
Figure 6. Figure 6: Overview of Perception bottle pose reward to encourage lifting it to the correct mid-air position; (3) A releasing reward that penalizes excessive holding force, encouraging proper release during handover. Reward functions for subtask 2: (4) A hand alignment reward to …
Figure 7
Figure 7. Figure 7: Performance comparison of curriculum learning, curriculum w/o safety penalty (SP), and [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: Environment-policy rollout for the multi-arm task starting from state [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]
Figure 9
Figure 9. Figure 9: Performance of SYMDEX on the multi-arm task. [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Snapshots from the real-world experiments. [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]

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. EquiBim: Learning Symmetry-Equivariant Policy for Bimanual Manipulation

    cs.RO 2026-03 conditional novelty 6.0 of 10

    Adding a loss that enforces left-right equivariance between observations and actions improves average bimanual imitation policy success by +2.7 to +9.5 points across four observation/action settings.

Reference graph

Works this paper leans on

48 extracted references · 27 canonical work pages · cited by 1 Pith paper

  1. [1]

    Ordo ˜nez Apraez, G

    D. Ordo ˜nez Apraez, G. Turrisi, V . Kostic, M. Martin, A. Agudo, F. Moreno-Noguer, M. Pontil, C. Semini, and C. Mastalli. Morphological symmetries in robotics. The International Journal of Robotics Research, 2025. doi:10.1177/02783649241282422

    doi:10.1177/02783649241282422 2025

  2. [2]

    Z. Su, X. Huang, D. Ordo ˜nez Apraez, Y . Li, Z. Li, Q. Liao, G. Turrisi, M. Pontil, C. Semini, Y . Wu, et al. Leveraging symmetry in rl-based legged locomotion control. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 6899–6906. IEEE, 2024

    work page 2024

  3. [3]

    Ordo ˜nez Apraez, M

    D. Ordo ˜nez Apraez, M. Martin, A. Agudo, and F. Moreno-Noguer. On discrete symmetries of robotics systems: A group-theoretic and data-driven analysis. In Proceedings of Robotics: Science and Systems, Daegu, Republic of Korea, July 2023. doi:10.15607/RSS.2023.XIX.053

    work page doi:10.15607/rss.2023.xix.053 2023

  4. [4]

    Mittal, N

    M. Mittal, N. Rudin, V . Klemm, A. Allshire, and M. Hutter. Symmetry considerations for learning task symmetric robot policies. In 2024 IEEE International Conference on Robotics and Automation (ICRA), pages 7433–7439. IEEE, 2024

    2024

  5. [5]

    K. Bao, C. Li, Y . As, A. Krause, and M. Hutter. Toward task generalization via memory augmentation in meta-reinforcement learning. arXiv preprint arXiv:2502.01521, 2025. Meta- RL with data-augmentation of the replay bufffer, for locomotion

    arXiv 2025

  6. [6]

    MI-HGNN: Morphology-Informed Heterogeneous Graph Neural Network for Legged Robot Contact Perception

    D. Butterfield, S. S. Garimella, N.-J. Cheng, and L. Gan. MI-HGNN: Morphology-Informed Heterogeneous Graph Neural Network for Legged Robot Contact Perception. arXiv preprint arXiv:2409.11146, 2024. URL https://arxiv.org/abs/2409.11146

    work page Pith review arXiv 2024

  7. [7]

    F. Xie, S. Wei, Y . Song, Y . Yue, and L. Gan. Morphological-symmetry-equivariant heteroge- neous graph neural network for robotic dynamics learning. arXiv preprint arXiv:2412.01297, 2024. 9

    arXiv 2024

  8. [8]

    H. Zhao, D. Wang, Y . Zhu, X. Zhu, O. L. Howell, L. Zhao, Y . Qian, R. Walters, and R. Platt. Hierarchical equivariant policy via frame transfer. In Forty-second International Conference on Machine Learning, 2025

    work page 2025

Show all 48 references
  1. [9]

    Amadio, A

    F. Amadio, A. Colom ´e, and C. Torras. Exploiting symmetries in reinforcement learning of bimanual robotic tasks. IEEE Robotics and Automation Letters, 4(2):1838–1845, 2019

    2019

  2. [10]

    Lin, Z.-H

    T. Lin, Z.-H. Yin, H. Qi, P. Abbeel, and J. Malik. Twisting lids off with two hands. In 8th Annual Conference on Robot Learning (CoRL) , 2024. URL https://openreview.net/ forum?id=3wBqoPfoeJ

    2024

  3. [11]

    Huang, Y

    B. Huang, Y . Chen, T. Wang, Y . Qin, Y . Yang, N. Atanasov, and X. Wang. Dynamic handover: Throw and catch with bimanual hands. In Conference on Robot Learning, pages 1887–1902. PMLR, 2023

    1902

  4. [12]

    T. Lin, K. Sachdev, L. Fan, J. Malik, and Y . Zhu. Sim-to-real reinforcement learning for vision-based dexterous manipulation on humanoids. arXiv preprint arXiv:2502.20396, 2025

    2025 arXiv

  5. [13]

    F. Lan, S. Wang, Y . Zhang, H. Xu, O. O. Oseni, Z. Zhang, Y . Gao, and T. Zhang. Dexcatch: Learning to catch arbitrary objects with dexterous hands. In 8th Annual Conference on Robot Learning

  6. [14]

    Jiang, X.-M

    J.-J. Jiang, X.-M. Wu, Y .-X. He, L.-A. Zeng, Y .-L. Wei, D. Zhang, and W.-S. Zheng. Rethink- ing bimanual robotic manipulation: Learning with decoupled interaction framework. arXiv preprint arXiv:2503.09186, 2025

    2025

  7. [15]

    L. Wang, Y . Zhang, Y . Hu, W. Wang, C. Zhang, Y . Gao, J. Hao, T. Lv, and C. Fan. Individual reward assisted multi-agent reinforcement learning. In International conference on machine learning, pages 23417–23432. PMLR, 2022

    2022

  8. [16]

    T. Chen, M. Tippur, S. Wu, V . Kumar, E. Adelson, and P. Agrawal. Visual dexterity: In-hand dexterous manipulation from depth. In Icml workshop on new frontiers in learning, control, and dynamical systems, 2023

    2023

  9. [17]

    Tiboni, P

    G. Tiboni, P. Klink, J. Peters, T. Tommasi, C. D’Eramo, and G. Chalvatzaki. Domain ran- domization via entropy maximization. In The Twelfth International Conference on Learning Representations, 2024. URL https://openreview.net/forum?id=GXtmuiVrOM

    2024

  10. [18]

    Akkaya, M

    I. Akkaya, M. Andrychowicz, M. Chociej, M. Litwin, B. McGrew, A. Petron, A. Paino, M. Plappert, G. Powell, R. Ribas, et al. Solving rubik’s cube with a robot hand. arXiv preprint arXiv:1910.07113, 2019

    1910 arXiv

  11. [19]

    M. M. Bronstein, J. Bruna, T. Cohen, and P. Veli ˇckovi´c. Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv preprint arXiv:2104.13478, 2021

    2021 arXiv

  12. [20]

    Zinkevich and T

    M. Zinkevich and T. Balch. Symmetry in markov decision processes and its implications for single agent and multi agent learning. Citeseer, 2001

    2001

  13. [21]

    Schulman, F

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  14. [22]

    Higgins, S

    I. Higgins, S. Racani `ere, and D. Rezende. Symmetry-based representations for artificial and biological general intelligence. Frontiers in Computational Neuroscience, page 28, 2022

    2022

  15. [23]

    D. Wang, M. Jia, X. Zhu, R. Walters, and R. Platt. On-robot learning with equivariant models. In 6th Annual Conference on Robot Learning, 2022

    2022

  16. [24]

    Weiler and G

    M. Weiler and G. Cesa. General e (2)-equivariant steerable cnns. Advances in neural informa- tion processing systems, 32, 2019. 10

    2019

  17. [25]

    T. Chen, E. Cousineau, N. Kuppuswamy, and P. Agrawal. Vegetable peeling: A case study in constrained dexterous manipulation. arXiv preprint arXiv:2407.07884, 2024

    2024 arXiv

  18. [26]

    Mittal, C

    M. Mittal, C. Yu, Q. Yu, J. Liu, N. Rudin, D. Hoeller, J. L. Yuan, R. Singh, Y . Guo, H. Mazhar, A. Mandlekar, B. Babich, G. State, M. Hutter, and A. Garg. Orbit: A unified simulation framework for interactive robot learning environments.IEEE Robotics and Automation Letters, 8...

    2023

  19. [27]

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, page 02783649241273668, 2023

    2023

  20. [28]

    W ¨uthrich, F

    M. W ¨uthrich, F. Widmaier, F. Grimminger, J. Akpo, S. Joshi, V . Agrawal, B. Hammoud, M. Khadiv, M. Bogdanovic, V . Berenz, et al. Trifinger: An open-source robot for learning dexterity. arXiv preprint arXiv:2008.03596, 2020

    2008 arXiv

  21. [29]

    Y . Ze, Z. Chen, W. Wang, T. Chen, X. He, Y . Yuan, X. B. Peng, and J. Wu. Generalizable humanoid manipulation with improved 3d diffusion policies.arXiv preprint arXiv:2410.10803, 2024

    2024 arXiv

  22. [30]

    C. Deng, O. Litany, Y . Duan, A. Poulenard, A. Tagliasacchi, and L. J. Guibas. Vector neu- rons: A general framework for so (3)-equivariant networks. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 12200–12209, 2021

    2021

  23. [31]

    Weiler, P

    M. Weiler, P. Forr ´e, E. Verlinde, and M. Welling. Equivariant and Coordinate Independent Convolutional Networks. 2023. URL https://maurice-weiler.gitlab.io/cnn_book/ EquivariantAndCoordinateIndependentCNNs.pdf

    2023

  24. [32]

    Van der Pol, D

    E. Van der Pol, D. Worrall, H. van Hoof, F. Oliehoek, and M. Welling. Mdp homomorphic networks: Group symmetries in reinforcement learning. Advances in Neural Information Pro- cessing Systems, 33:4199–4210, 2020

    2020

  25. [33]

    Ordo ˜nez Apraez, V

    D. Ordo ˜nez Apraez, V . Kostic, G. Turrisi, P. Novelli, C. Mastalli, C. Semini, and M. Pontil. Dy- namics harmonic analysis of robotic systems: Application in data-driven koopman modelling. In 6th Annual Learning for Dynamics & Control Conference, pages 1318–1329. PMLR, 2024

    2024

  26. [34]

    H. Ryu, J. Kim, H. An, J. Chang, J. Seo, T. Kim, Y . Kim, C. Hwang, J. Choi, and R. Horowitz. Diffusion-edfs: Bi-equivariant denoising generative modeling on se (3) for visual robotic ma- nipulation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Reco...

    2024

  27. [35]

    Brehmer, J

    J. Brehmer, J. Bose, P. De Haan, and T. S. Cohen. Edgi: Equivariant diffusion for planning with embodied agents. Advances in Neural Information Processing Systems, 36:63818–63834,

  28. [36]

    Huang, D

    H. Huang, D. Wang, R. Walters, and R. Platt. Equivariant transporter network. In Proceedings of Robotics: Science and Systems, 2022

    2022

  29. [37]

    X. Zhu, D. Wang, O. Biza, G. Su, R. Walters, and R. Platt. Sample efficient grasp learning using equivariant models. In Robotics: Science and Systems, 2022. Manipulation of objects in 3D space with SO(3) grasp pose generation

    2022

  30. [38]

    D. Wang, R. Walters, and R. Platt. SO(2)-equivariant reinforcement learning. In International Conference on Learning Representations, 2022

    2022

  31. [39]

    R. Wang, R. Walters, and R. Yu. Incorporating symmetry into deep dynamics models for improved generalization. arXiv preprint arXiv:2002.03061, 2020. 11

    2002 arXiv

  32. [40]

    D. Wang, S. Hart, D. Surovik, T. Kelestemur, H. Huang, H. Zhao, M. Yeatman, J. Wang, R. Walters, and R. Platt. Equivariant diffusion policy.arXiv preprint arXiv:2407.01812, 2024

    2024 arXiv

  33. [41]

    B. Hu, H. Tian, D. Wang, H. Huang, X. Zhu, R. Walters, and R. Platt. Push-grasp policy learn- ing using equivariant models and grasp score optimization. arXiv preprint arXiv:2504.03053, 2025

    2025 arXiv

  34. [42]

    Ordo ˜nez Apraez, A

    D. Ordo ˜nez Apraez, A. Agudo, F. Moreno-Noguer, and M. Martin. An adaptable approach to learn realistic legged locomotion without examples. In 2022 international conference on Robotics and automation (ICRA), pages 4671–4678. IEEE, 2022

    2022

  35. [43]

    Abdolhosseini, H

    F. Abdolhosseini, H. Y . Ling, Z. Xie, X. B. Peng, and M. Van de Panne. On learning symmetric locomotion. In Proceedings of the 12th ACM SIGGRAPH Conference on Motion, Interaction and Games, pages 1–10, 2019

    2019

  36. [44]

    H. Qi, A. Kumar, R. Calandra, Y . Ma, and J. Malik. In-hand object rotation via rapid motor adaptation. In Conference on Robot Learning, pages 1722–1732. PMLR, 2023

    2023

  37. [45]

    B. Wen, W. Yang, J. Kautz, and S. Birchfield. Foundationpose: Unified 6d pose estimation and tracking of novel objects. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 17868–17879, 2024

    2024

  38. [46]

    N. Ravi, V . Gabeur, Y .-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. R ¨adle, C. Rolland, L. Gustafson, et al. Sam 2: Segment anything in images and videos. arXiv preprint arXiv:2408.00714, 2024

    2024 arXiv

  39. [47]

    V . E. Kremer. Quaternions and slerp. InEmbots. dfki. de/doc/seminar ca/Kremer Quaternions. pdf, 2008. 12 Appendix A Background on group and representation theory Group actions and representations This section provides a brief overview of the fundamental concepts in group and ...

    2008

  40. [2023]

    Diffusion policy encoding SO(3) and O(n) (permutation of multiple instances of the same object)

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.