Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

Human-Humanoid Robots Cross-Embodiment Behavior-Skill Transfer Using Decomposed Adversarial Learning from Demonstration

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read A unified digital human lets one set of demonstrations transfer loco-manipulation skills to five humanoid robots.

desk verdict Solid systems paper with a plausible pipeline, but the headline efficiency claim rests on an incomplete cost count and there is no hardware validation. read the letter →

arxiv 2412.15166 v1 pith:GKZH2HN7 submitted 2024-12-19 cs.RO cs.AI

Junjia Liu , Zhuo Li , Minghao Yu , Zhipeng Dong , Sylvain Calinon , Darwin Caldwell , Fei Chen This is my paper · ORCID
classification cs.ROcs.AI
keywords cross-embodimentskilltransferhumanoidrobotsloco-manipulationdecomposedadversarialimitationlearningmotionretargetingbehaviorprimitivesinteractiongraphfromdemonstration
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 tries to establish that a single set of human motion demonstrations can be turned into reusable whole-body skills that deploy onto several differently built humanoid robots without retraining each robot from scratch. The authors propose a unified digital human model as a common prototype: human demonstrations are retargeted onto this 92-DoF body, decomposed adversarial imitation learning trains behavior primitives for separate functional parts (legs, arms, hands), and a human-object interaction graph guides a high-level policy that composes those primitives. Transfer to each concrete robot then requires only kinematic motion retargeting plus a short fine-tuning step. If the central claim is right, the data-collection bottleneck for new humanoid platforms is largely removed, because one human capture session seeds skills for many embodiments and the per-robot training cost drops to under an hour in simulation.

What carries the argument

The load-bearing machinery is the pairing of a unified digital human (UDH) with decomposed adversarial imitation learning (DAIL). The UDH is a 92-DoF common body that aggregates human demonstrations and defines an embodiment-independent motion space, so that a single demonstration set can be re-targeted to any robot. DAIL splits the body into functional components (lower limbs, upper limbs, hands), each with its own behavior encoder and style discriminator; during whole-body training the per-part style rewards are multiplied, which lets high-DoF coordination emerge while keeping each part's imitation signal clean. Around this core sit two smaller mechanisms: the interaction graph $G$, a contact-and-relative-distance representation that makes the high-level skill policy $\eta(z|s,G,g)$ robust to changing object poses and sizes, and the kinematic retargeting plus MLP fine-tuning layer that maps latent behavior plans onto a specific robot's joint commands while respecting its dynamics.

What would settle it

Deploy the fine-tuned policies on the physical NAVIAI, H1, Walker, Bruce, and CURI robots and run the box-carrying task: if the robots cannot maintain balance, reach the box, or grasp without falling under real torque limits and sensor noise, the cross-embodiment transfer claim fails. A cheaper simulation-side check is to randomize the simulated joint-torque limits and link inertias around the nominal URDF values and measure whether the under-an-hour fine-tuning still converges.

Watch

Extended reading notes

Core claim

The paper claims that cross-embodiment loco-manipulation skill transfer can be reduced to three separable steps: retarget, decompose, fine-tune. First, a unified digital human (92 DoFs) absorbs human demonstrations through kinematic retargeting that groups joints by function and solves partial inverse kinematics, normalizing root trajectories for body-size differences. Second, decomposed adversarial imitation learning (DAIL) trains a low-level behavior controller $\pi(a|s,z)$ on a latent behavior space $\mathcal{Z}$ (a unit hypersphere), with separate style discriminators per functional part whose rewards multiply, so the robot imitates behavioral characteristics rather than exact trajectories. Third, a high-level policy $\eta(z|s,G,g)$ plans latent behavior sequences under an interaction graph $G$ that encodes hand-object contact and relative distances, and an MLP fine-tuning layer converts retargeted kinematics into dynamics-level commands for the target robot. The paper validates this pipeline in a GPU-based physics simulator on five humanoid robots with 16-87 DoFs, reporting that the box-carrying task completes with normalized returns comparable to or better than adversarial-motion-prior baselines while cutting per-robot training time from roughly 2.2-5.8 hours to 0.75-0.97 hour.

Load-bearing premise

The central load-bearing premise is that the simulated copies of the five humanoid robots faithfully represent the real robots' joint limits, torque limits, and inertias, so that policies trained and fine-tuned in simulation will behave the same way on physical hardware; the paper itself states that the dynamics-motor gap to real robots remains unresolved.

Editorial extensions

If this is right

  • A newly introduced humanoid platform could acquire a demonstrated loco-manipulation skill after only kinematic retargeting and under an hour of fine-tuning, instead of several hours of task-specific reinforcement learning.
  • The same captured human demonstration can be amortized across all platforms, so the main data-collection cost is paid once per skill, not once per robot.
  • Decomposing the body into functional parts with independent discriminators consistently accelerates and improves imitation learning relative to whole-body adversarial imitation, with the largest gains on the highest-DoF robots.
  • The interaction graph makes the learned skill re-usable under different start and target object positions without re-training the low-level behavior controller, as demonstrated for H1 in the paper.
  • Because all experiments are in simulation, the quantitative claims about returns and training times are strictly about simulated copies of the five robots; real-robot deployment is identified by the paper itself as future work.

Reading between the lines

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

  • Extending beyond the paper: if the simulated-fidelity assumption holds on hardware, the framework implies that per-robot data collection for new humanoid skills could drop to near zero, shifting the bottleneck to capturing diverse human interactions once.
  • Extending beyond the paper: the part-wise decomposition principle is not humanoid-specific; it could apply to any high-DoF embodied agent (quadruped with arms, wheeled manipulator, surgical robot) whose functional components need different control precision, using the same multiplied style-reward training.
  • Extending beyond the paper: the interaction graph could be generalized beyond hand-object contact to include feet, torso, and environmental contacts, which would let the same skill-level planning handle climbing, carrying on the shoulder, or pushing against walls.
  • Extending beyond the paper: a testable extension would measure how fine-tuning time scales with robot DoF across a larger family of platforms; the reported numbers suggest the method is most advantageous for high-DoF robots, but the paper does not isolate that trend.
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

5 major / 5 minor

Summary. The paper presents a framework for transferring loco-manipulation skills from human demonstrations to multiple humanoid robots. The method combines a unified digital human (UDH) model as a common prototype, decomposed adversarial imitation learning (DAIL) over functional body parts, a human-object interaction graph for task-level guidance, and embodiment-specific kinematic retargeting followed by a short fine-tuning stage. The authors validate the approach in Isaac Gym on five humanoid robots (NAVIAI, H1, Bruce, Walker, CURI) for behavior imitation and for a box-carrying loco-manipulation task, reporting normalized training returns and wall-clock training times that compare favorably with AMP, ASE, and PMP baselines.

Significance. If the claims are supported, the framework would be a useful step toward reusing human demonstration data across heterogeneous humanoid platforms, addressing an important data-efficiency and transfer problem in humanoid robotics. The paper also makes a positive contribution by releasing retargeted motion datasets and the Rofunc-based code, which supports reproducibility. The decomposed training idea and the use of an interaction graph for task generalization are plausible and potentially valuable. However, the evidence as presented is not yet sufficient to establish the central cross-embodiment deployability claim: all results are from simulation, the headline metric is a learned training return rather than task success, no variance or seed information is reported, and the comparison against baselines is described only at a high level.

major comments (5)
  1. [Sec. IV-A and Sec. VI (Discussion)] All validation is performed in Isaac Gym, with robot models described as 'similar as hardware constraints' and motor torque limitations simulated by constraining PD controller output. The Discussion explicitly acknowledges that the disparity between simulated and real robot dynamics remains a significant challenge. Since the paper's headline benefit is deploying the same loco-manipulation skills on physical humanoid platforms without extensive retraining, the lack of any real-hardware validation, or at least a systematic sim-to-real study with domain randomization and system identification, leaves the central deployability claim unsupported. The authors should either provide hardware experiments or explicitly scope the contribution to simulation and provide evidence that the simulated dynamics are faithful.
  2. [Sec. IV-D, Table III] The quantitative comparison uses 'normalized average training returns' as the performance metric, but this is a learned objective combining task reward and style rewards rather than a measured task success rate. Table III reports no standard deviations, no number of seeds, and no success/failure counts, so the reported superiority over AMP, ASE, and PMP cannot be assessed statistically. The paper should report task success rates, return values with variance across seeds, and learning curves for at least the box-carrying task.
  3. [Sec. IV-D, Table III] The training-time comparison appears to undercount the cost of the proposed method. The 'Ours' times in Table III likely correspond to the fine-tuning stage described in Sec. III-D, but the framework also requires behavior-primitive pre-training on each embodiment (Sec. IV-C) and skill-policy training on the UDH, whose costs are not included in the per-robot 'under an hour' claim. Without reporting the full pipeline cost, including primitive pre-training for each robot, the efficiency advantage over baselines is not established.
  4. [Sec. III-B and Sec. IV-C] The abstract and introduction claim that the framework 'bypasses the need for re-training on every new robot platform,' yet Sec. IV-C states that decomposed behavior primitive pre-training is conducted independently on the UDH and on each of the five humanoid robots. This is a substantial per-embodiment training step. The claim of bypassing retraining is therefore overstated. The authors should clarify what exactly is transferred, what must be trained per robot, and report the total per-robot training time including primitive pre-training.
  5. [Sec. III-C and Sec. IV-D] The generalization claim rests on a single loco-manipulation task (box carrying) with one additional variation in start and target positions for one robot. There are no multiple task instances, no statistical evaluation across seeds, and no ablations on the interaction graph's contribution beyond the descriptive comparison with ASE. The evidence is too limited to support the general statement that task generalization is achieved through the interaction graph. Additional tasks, multiple seeds, and a quantitative analysis of the graph's effect are needed.
minor comments (5)
  1. [Table I and Table III] There is an inconsistency: Table I lists Bruce as having 1x20 finger DoFs, but Table III's footnote says Bruce was excluded from the loco-manipulation comparison due to the lack of hands. Please clarify whether Bruce has hands or not, and why the exclusion criterion is stated differently in the two tables.
  2. [Sec. IV-C] The text refers to 'NA VIAI, HI, and Walker' when listing full-sized humanoid robots; 'HI' should be 'H1'. Elsewhere the paper also uses inconsistent spacing in 'NA VIAI' (with or without a space). Please correct these typos.
  3. [Table II] The entry 'Thump up' should read 'Thumbs up'. Also, the table would benefit from a caption explaining whether the reported values are means over episodes, seeds, or both.
  4. [Sec. IV-A and Sec. III-B] Several important hyperparameters are not reported, including the latent behavior space dimension, the prior p(z), the reward weights wg, ws, wb in Eq. (3), network architectures, and the number of training iterations. Providing these details would improve reproducibility.
  5. [Sec. IV-B] The paper states that motion data were converted to FBX and retargeted using the Rofunc platform, but it does not describe the retargeting procedure for the UDH versus the five robots beyond the partial inverse kinematics sketch in Sec. III-A. A more detailed algorithm or pseudo-code would help readers judge the generality of the approach.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the framework is benchmarked against external baselines, and the self-citations are tool references rather than load-bearing premises.

full rationale

The paper's derivation chain is not circular. Human demonstrations are captured with an external motion-capture system, retargeted kinematically to a unified digital human model and then to each robot embodiment, and used to train behavior primitives via adversarial imitation. The high-level skill policy is trained with a combination of task reward and style rewards, and fine-tuning adapts the retargeted reference to each robot's dynamics. The main comparisons are against external baselines (AMP, ASE, PMP) in the Isaac Gym simulator, so the central efficiency and transfer claims are not forced by construction from the paper's own equations. The self-citations to the Rofunc platform and earlier imitation-learning works are tool and method references, not sources of a uniqueness theorem or of the target result; they do not reduce the derivation to itself. The normalized training-return metric is an internal performance measure rather than an independent task-completion metric, and the Discussion explicitly concedes that the sim-to-real gap remains a significant challenge. These are missing-support and external-validity concerns, not circular reductions. The reported per-robot fine-tuning times may also exclude per-robot behavior-primitive pre-training, but that is a reporting issue rather than a circularity. Overall, the paper does not exhibit a circular derivation, and any concerns about benchmark completeness or hardware transfer do not raise the circularity score beyond a low non-circular level.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The framework relies on a set of hand-chosen hyperparameters (reward weights, latent dimension, decomposition grouping) and several domain assumptions about simulation fidelity, retargeting fidelity, and the sufficiency of the interaction graph. These are not derived from first principles or independently validated beyond the paper's own experiments.

free parameters (3)
  • Reward weights wg, ws, wb = unspecified
    In Eq. 3, the task reward, skill style reward, and behavior style reward are combined with weights wg, ws, wb. Their values are not given in the paper and are presumably tuned by hand; the central results depend on them.
  • Latent behavior space dimension and prior = unspecified
    Section III-B defines a latent variable z on a unit hypersphere, but the dimension is not stated. This affects the capacity of the behavior primitive space.
  • Decomposed functional part grouping = listed in Table I
    The grouping of joints into functional parts is a design choice; different groupings would change the training dynamics. The paper does not ablate over this choice.
assumptions (4)
  • domain assumption The simulated URDF models for NAVIAI, H1, Bruce, Walker, and CURI faithfully represent their real dynamics, torque limits, and inertia
    Sec IV-A states joint limitations and inertia are based on URDF models 'similar as hardware constraints'. All validation is in Isaac Gym, so the sim-to-real transfer depends on this assumption.
  • domain assumption Kinematic motion retargeting via partial inverse kinematics preserves the task-relevant information in human demonstrations across embodiments
    Sec III-A assumes that solving joint values by functional-group IK and normalizing root trajectories retains the behavioral patterns needed for later imitation and fine-tuning.
  • ad hoc to paper The interaction graph G with binary contact edges and relative distances is a sufficient representation for task generalization
    Sec III-C introduces this graph as the sole high-level task representation; no evidence that richer interaction features would not be needed.
  • standard math Adversarial imitation with a per-part discriminator converges to a policy that matches human-like whole-body behavior
    This is the standard assumption of GAIL/AMP-style training: the discriminator is a good proxy for behavior similarity.
invented entities (2)
  • Unified Digital Human (UDH) model
    purpose: Common 92-DoF prototype embodiment that aggregates human demonstrations and provides a shared motion space for retargeting to different robots
    The UDH is constructed in this paper as a reference model; its kinematic and dynamic fidelity is not independently validated.
  • Human-object interaction graph G
    purpose: Graph whose nodes are body parts and objects, with binary contact edges and relative distances, used to guide the high-level skill policy
    The graph is an internal representation invented for this framework; its predictive validity is demonstrated only indirectly via the reported returns.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Human-Humanoid Robots Cross-Embodiment Behavior-Skill Transfer Using Decomposed Adversarial Learning from Demonstration." pith.science (2026). https://pith.science/paper/GKZH2HN7

@misc{pith2026241215166,
  author       = {Pith},
  title        = {Pith review of: Human-Humanoid Robots Cross-Embodiment Behavior-Skill Transfer Using Decomposed Adversarial Learning from Demonstration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GKZH2HN7}},
  note         = {Machine review of arXiv:2412.15166}
}
read the original abstract

Humanoid robots are envisioned as embodied intelligent agents capable of performing a wide range of human-level loco-manipulation tasks, particularly in scenarios requiring strenuous and repetitive labor. However, learning these skills is challenging due to the high degrees of freedom of humanoid robots, and collecting sufficient training data for humanoid is a laborious process. Given the rapid introduction of new humanoid platforms, a cross-embodiment framework that allows generalizable skill transfer is becoming increasingly critical. To address this, we propose a transferable framework that reduces the data bottleneck by using a unified digital human model as a common prototype and bypassing the need for re-training on every new robot platform. The model learns behavior primitives from human demonstrations through adversarial imitation, and the complex robot structures are decomposed into functional components, each trained independently and dynamically coordinated. Task generalization is achieved through a human-object interaction graph, and skills are transferred to different robots via embodiment-specific kinematic motion retargeting and dynamic fine-tuning. Our framework is validated on five humanoid robots with diverse configurations, demonstrating stable loco-manipulation and highlighting its effectiveness in reducing data requirements and increasing the efficiency of skill transfer across platforms.

Figures

Figures reproduced from arXiv: 2412.15166 by the authors.

Figure 1
Figure 1. Human can serve as the prototype of diverse humanoid robots, efficiently learning generalized loco-manipulation skills without [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Schematic overview of the cross-embodiment loco-manipulation skill transfer framework. 1) Human embodiment demonstration is captured by motion capture system and retargeted to the unified digital human, then retargeted to diverse humanoid robots. 2) High DoFs of humanoids are decomposed into functional parts p ∈ [1, P] and trained with partial demonstration separately via adversarial imitation to form the latent beh… view at source ↗
Figure 3
Figure 3. Kinematic motion retargeting. (a) Motions are retargeted by grouping DoFs belonging to same function parts and each joint value is solved by partial inverse kinematics. Orange ellipses refer to the group of shoulder DoFs, light blue ellipses refer to the group of hip DoFs. (b) The animation screenshots of kinematic motion retargeting on the unified digital human, NAVIAI and H1 humanoid robots. It is worth noting tha… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Whole-body functional decomposition and behavior primitive training. (a) Behavior primitive pre-training for the unified digital human and five humanoid robots. It shows the style imitation of the given motion dataset using pre-trained behavior primitives. (b) Hand-spe…
Figure 5
Figure 5. Figure 5: Loco-manipulation skill learning with interaction graph. An illustration of the detailed loco-manipulation skill learning process with skill-level interaction graph guidance and decomposed behavior primitive trajectories in latent behavior spaces. pose a method called …
Figure 6
Figure 6. Figure 6: Behavior primitive pre-training. Stability experiments for legged humanoid robots by hitting them with moving cubes. Tasks UDH NAVIAI H1 Bruce Walker CURI Walk 0.93 0.81 0.82 0.95 0.87 - Jump 0.91 0.77 0.79 0.91 0.73 - Squat 0.94 0.82 0.89 0.96 0.86 - Leg stretch 0.93 …
Figure 7
Figure 7. Figure 7: Behavior primitive pre-training. Ablation study for the effectiveness of decomposition. Solid lines refer to decomposed imitation learning, dashed lines refer to the learning without de￾composed imitation. promising results of imitating these behaviors are illustrated …
Figure 8
Figure 8. Figure 8: Loco-manipulation skill. Box carrying and putting task performed by unified digital human, NAVIAI and UnitreeH1 robots. For better visualization, the box is set to float at the start position, and the red sphere is the target position of the box. body parts using the u…

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. Krysalis Hand: A Lightweight, High-Payload, 18-DoF Anthropomorphic End-Effector for Robotic Learning and Dexterous Manipulation

    cs.RO 2025-04 conditional novelty 6.0 of 10

    The Krysalis Hand is a 790 g, 18-DoF five-finger robot hand with self-locking lead-screw joints that can passively hold over 10 lbs and actively push about 10 N per fingertip.

Reference graph

Works this paper leans on

21 extracted references · 11 canonical work pages · cited by 1 Pith paper

  1. [1]

    Versatile multicontact planning and control for legged loco-manipulation,

    J.-P. Sleiman, F. Farshidian, and M. Hutter, “Versatile multicontact planning and control for legged loco-manipulation,” Science Robotics, vol. 8, no. 81, p. eadg5014, 2023

    2023

  2. [2]

    Learning robust perceptive locomotion for quadrupedal robots in the wild,

    T. Miki, J. Lee, J. Hwangbo, L. Wellhausen, V . Koltun, and M. Hutter, “Learning robust perceptive locomotion for quadrupedal robots in the wild,” Science Robotics , vol. 7, no. 62, p. eabk2822, 2022

    2022

  3. [3]

    Real-world humanoid locomotion with reinforcement learning,

    I. Radosavovic, T. Xiao, B. Zhang, T. Darrell, J. Malik, and K. Sreenath, “Real-world humanoid locomotion with reinforcement learning,” Science Robotics , vol. 9, no. 89, p. eadi9579, 2024

    work page 2024

  4. [4]

    Combining learning-based locomotion policy with model-based manipulation for legged mobile manipulators,

    Y . Ma, F. Farshidian, T. Miki, J. Lee, and M. Hutter, “Combining learning-based locomotion policy with model-based manipulation for legged mobile manipulators,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 2377–2384, 2022

    2022

  5. [5]

    Sim-to-real learning for humanoid box loco-manipulation,

    J. Dao, H. Duan, and A. Fern, “Sim-to-real learning for humanoid box loco-manipulation,” arXiv preprint arXiv:2310.03191 , 2023

    arXiv 2023

  6. [6]

    Calinon, Robot programming by demonstration

    S. Calinon, Robot programming by demonstration . EPFL Press, 2009

    work page 2009

  7. [7]

    Flexible automation driven by demon- stration: Leveraging strategies that simplify robotics,

    A. Giusti, M. J. Zeestraten, E. Icer, A. Pereira, D. G. Caldwell, S. Calinon, and M. Althoff, “Flexible automation driven by demon- stration: Leveraging strategies that simplify robotics,” IEEE Robotics & Automation Magazine , vol. 25, no. 2, pp. 18–27, 2018

    work page 2018

  8. [8]

    Robot cooking with stir-fry: Bimanual non-prehensile manipulation of semi-fluid objects,

    J. Liu, Y . Chen, Z. Dong, S. Wang, S. Calinon, M. Li, and F. Chen, “Robot cooking with stir-fry: Bimanual non-prehensile manipulation of semi-fluid objects,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 5159–5166, 2022

    work page 2022

Show all 21 references
  1. [9]

    Softgpt: Learn goal-oriented soft object manipulation skills by generative pre- trained heterogeneous graph transformer,

    J. Liu, Z. Li, W. Lin, S. Calinon, K. C. Tan, and F. Chen, “Softgpt: Learn goal-oriented soft object manipulation skills by generative pre- trained heterogeneous graph transformer,” in 2023 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS) , pp. 4920...

    2023

  2. [10]

    Birp: Learning robot generalized bimanual coordination using relative parameterization method on human demonstration,

    J. Liu, H. Sim, C. Li, K. C. Tan, and F. Chen, “Birp: Learning robot generalized bimanual coordination using relative parameterization method on human demonstration,” in 2023 62nd IEEE Conference on Decision and Control (CDC) , pp. 8300–8305, IEEE, 2023

    2023

  3. [11]

    Amp: Adversarial motion priors for stylized physics-based character con- trol,

    X. B. Peng, Z. Ma, P. Abbeel, S. Levine, and A. Kanazawa, “Amp: Adversarial motion priors for stylized physics-based character con- trol,” ACM Transactions on Graphics (ToG) , vol. 40, no. 4, pp. 1–20, 2021

    2021

  4. [12]

    Ase: Large- scale reusable adversarial skill embeddings for physically simulated characters,

    X. B. Peng, Y . Guo, L. Halper, S. Levine, and S. Fidler, “Ase: Large- scale reusable adversarial skill embeddings for physically simulated characters,” ACM Transactions On Graphics (TOG) , vol. 41, no. 4, pp. 1–17, 2022

    2022

  5. [13]

    Pmp: Learning to physically interact with environments using part-wise motion priors,

    J. Bae, J. Won, D. Lim, C.-H. Min, and Y . M. Kim, “Pmp: Learning to physically interact with environments using part-wise motion priors,” in ACM SIGGRAPH 2023 Conference Proceedings , pp. 1–10, 2023

    2023

  6. [14]

    Xskill: Cross embod- iment skill discovery,

    M. Xu, Z. Xu, C. Chi, M. Veloso, and S. Song, “Xskill: Cross embod- iment skill discovery,” in Conference on Robot Learning , pp. 3536– 3555, PMLR, 2023

    2023

  7. [15]

    Cross-embodiment robot manipulation skill transfer using latent space alignment,

    T. Wang, D. Bhatt, X. Wang, and N. Atanasov, “Cross-embodiment robot manipulation skill transfer using latent space alignment,” arXiv preprint arXiv:2406.01968, 2024

    2024 arXiv

  8. [16]

    Xirl: Cross-embodiment inverse reinforcement learning,

    K. Zakka, A. Zeng, P. Florence, J. Tompson, J. Bohg, and D. Dwibedi, “Xirl: Cross-embodiment inverse reinforcement learning,” in Confer- ence on Robot Learning , pp. 537–546, PMLR, 2022

    2022

  9. [17]

    Pushing the limits of cross- embodiment learning for manipulation and navigation,

    J. Yang, C. Glossop, A. Bhorkar, D. Shah, Q. Vuong, C. Finn, D. Sadigh, and S. Levine, “Pushing the limits of cross- embodiment learning for manipulation and navigation,” arXiv preprint arXiv:2402.19432, 2024

    2024 arXiv

  10. [18]

    Mirage: Cross-embodiment zero-shot policy transfer with cross- painting,

    L. Y . Chen, K. Hari, K. Dharmarajan, C. Xu, Q. Vuong, and K. Gold- berg, “Mirage: Cross-embodiment zero-shot policy transfer with cross- painting,” arXiv preprint arXiv:2402.19249 , 2024

    2024 arXiv

  11. [19]

    SMPL: A skinned multi-person linear model,

    M. Loper, N. Mahmood, J. Romero, G. Pons-Moll, and M. J. Black, “SMPL: A skinned multi-person linear model,” ACM Trans. Graphics (Proc. SIGGRAPH Asia) , vol. 34, pp. 248:1–248:16, Oct. 2015

    2015

  12. [20]

    Rofunc: The full process python package for robot learning from demonstration and robot manipulation,

    J. Liu, Z. Dong, C. Li, Z. Li, M. Yu, D. Delehelle, and F. Chen, “Rofunc: The full process python package for robot learning from demonstration and robot manipulation,” Zenodo, https://github.com/Skylark0924/Rofunc, 2023

    2023

  13. [21]

    Isaac gym: High performance gpu-based physics simulation for robot learning,

    V . Makoviychuk, L. Wawrzyniak, Y . Guo, M. Lu, K. Storey, M. Mack- lin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, et al. , “Isaac gym: High performance gpu-based physics simulation for robot learning,” arXiv preprint arXiv:2108.10470 , 2021

    2021 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.