Learning Physical Interaction Skills from Human Demonstrations

Hengbo Ma; Kwonjoon Lee; Sehoon Ha; Tianyu Li

REVIEW 3 major objections 6 minor 70 references

Learning Physical Interaction Skills from Human Demonstrations

T0 review · 3 major / 6 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 This paper claims that a sparse, learned Embedded Interaction Graph can carry the semantics of two-person physical interactions from human demonstrations to robots with very different bodies, and that training policies to reproduce this…

desk verdict A genuinely useful framework for cross-embodiment interaction learning, with the main transfer claim under-evidenced by qualitative-only evaluation. read the letter →

arxiv 2507.20445 v2 pith:F2MD3BW4 submitted 2025-07-28 cs.RO

Tianyu Li , Hengbo Ma , Sehoon Ha , Kwonjoon Lee This is my paper · ORCID

classification cs.RO

keywords cross-embodimentimitationlearningfromdemonstrationinteractiongraphmotionretargetingphysics-basedreinforcementmulti-agentcoordinationsemanticssparseattention

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

Physical interaction skills such as handshaking, sparring, and dancing are difficult to transfer from human demonstrations to robots because a quadruped or mobile manipulator does not share the demonstrator's skeleton. This paper proposes BuddyImitation, a two-stage framework that compresses a two-person interaction into a sparse, time-varying Embedded Interaction Graph and then trains a reinforcement-learning policy to reproduce that graph in a physics-based simulator. The central claim is that this graph, rather than raw joint trajectories, is the right thing to imitate: it selects the few inter-agent edges that predict future motion, so a one-armed robot can switch between imitating the demonstrator's left and right arms and adapt handshake height to its partner. If the claim is correct, cross-embodiment imitation can proceed without handcrafted interaction objectives or matching skeletons.

What carries the argument

The load-bearing object is the Embedded Interaction Graph (EIG), a sparse, time-indexed subgraph of the Interaction Graph in which nodes are joints and edges carry a 6D feature describing the relative position and midpoint of paired joints between the two characters. The embedding stage selects one edge per attention head via hard attention between the character's current pose and all edge embeddings, so the chosen edges are those most predictive of the next pose when passed through a pretrained motion decoder. The transfer stage converts this graph into an imitation objective through an interaction consistency reward on normalized edge length, root-edge XY direction, and edge center-point height, which rewards a new character for reproducing the relational geometry of the reference interaction rather than its joint angles.

What would settle it

Retrain one transferred interaction, such as Go2Ar handshaking, with the automatic vertex correspondence replaced by a deliberately permuted mapping that assigns the demonstrator's right-hand graph edge to the robot's left manipulator or rear leg. If the resulting policy still produces a recognizable handshake, the correspondence assumption is not load-bearing; if the behavior loses its interaction semantics or collapses, the assumption is the critical link.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that interaction semantics can be distilled into a time-varying sparse graph $G_{\mathrm{emb}}^t$, a subset of the fully connected Interaction Graph, whose edges are chosen by multi-head cross-attention to minimize future-pose prediction error. The paper shows that this embedded graph predicts the next 120 frames of a demonstration more accurately than random sparse graphs or the full 484-edge graph, and that the selected edges shift with context: right-arm and root edges dominate handshaking, mirrored hand edges appear in circling, and alternating dominant-arm edges appear in sparring. It then uses the graph as the reward signal for a centralized hierarchical policy, with an interaction consistency reward that compares normalized edge lengths, root-edge direction, and edge center-point heights. Reported results show Go2Ar, Stretch, and humanoid agents producing recognizable interaction patterns while adapting limb use and posture to their own bodies.

Load-bearing premise

The framework assumes that the automatic vertex correspondence, computed by inner products of root-to-end-effector vectors in neutral pose, maps each semantic role of the demonstration graph onto the correct body part of the new character; if that mapping is wrong, the interaction consistency reward measures the wrong graph features and the policy is trained toward incorrect semantics.

Editorial extensions

If this is right

A robot with a single manipulator can learn bimanual human interactions by switching its reference limb to whichever demonstrator arm dominates the current interaction state.
Agents adapt the spatial details of an interaction to their partner: handshake height rises or falls with the relative sizes of the two characters.
Training in physics-based simulation makes the policy robust to physically infeasible or noisy demonstration frames, because such frames lead to falls and low rewards and are avoided.
A four-edge embedded graph predicts future interaction poses better than a fully connected graph, supporting the use of sparse representations as imitation objectives.

Reading between the lines

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

Not claimed by the paper: because the selected edges are semantically interpretable, the same graph could be used as an explanation of what an interaction is about, not only as a control objective for a robot.
The paper trains separate policies per interaction and character pair; a natural extension it leaves implicit is conditioning one policy on the interaction identity, which would require a shared observation space across embodiments.
One testable extension is to feed the time-varying graph attention into a downstream task planner, so that the interaction semantics can be re-targeted online when a partner changes body mid-interaction; the paper does not address online switching.
The vertex-correspondence method is based on neutral-pose geometry, so a stronger test would be to evaluate it on embodiments with asymmetric or redundant limbs, where the inner-product mapping has multiple plausible answers.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The one thing you should know: BuddyImitation is a real step forward for cross-embodiment interaction learning. It learns a sparse, interpretable interaction graph from human demonstrations and uses it as an RL reward to train policies for non-humanoid robots. The individual pieces (Interaction Graph, MVAE, CrossLoco-style RL) are known, but the combination—learning which edges matter and transferring that semantics across morphologies—is new and clearly useful.

The embedding stage is the strongest part. The paper quantifies pose prediction against a random graph of the same size and a fully connected graph, and shows the learned 4-edge graph beats both. The ablations (E0, E1, E4, E8, EF) are well designed and show the sparsity tradeoff. That part earns its keep.

The soft spot is exactly where the reader put it: the transfer stage is evaluated by videos and a user study without error bars or tests. The user study (50 participants) reports "significantly outperforms" but gives no standard deviations or p-values. That is fixable in revision. The automated vertex correspondence in §3.3.1 is the load-bearing heuristic for the whole transfer reward, and the paper never reports the actual mappings for Stretch or Go2Ar. If an arm vertex maps to a paw or base, the reward silently drives the wrong body part. The qualitative videos suggest it usually works—Go2Ar switching from left to right arm based on the dominant sparring hand is a nice piece of evidence—but "usually" is not a metric. The authors should provide quantitative interaction metrics (e.g., edge-feature errors over rollouts) and at least a table of correspondences.

The circularity concern is minor. The graph is learned on the demonstrations and then used as an objective; that is standard reward design, not a hidden derivation of the result.

Overall: this is a promising robotics/animation paper with a sound core idea and a clear gap in evaluation. It deserves a serious referee—the novelty is there and the problems are addressable. I'd like to see it in a reading group, and I'd cite it if I work on cross-embodiment imitation.

Recommendation: send to peer review, with a request for transfer-stage metrics and code/data release.

Referee Report

3 major / 6 minor

Summary. The paper presents BuddyImitation, a two-stage framework for learning physical interaction skills (dancing, handshaking, sparring, rock-paper-scissors, circling) from human-human demonstration data and transferring them to agents with substantially different morphologies (humanoid, child, Go2Ar legged manipulator, Stretch mobile manipulator). In the Interaction Embedding stage, a sparse Embedded Interaction Graph (EIG) is learned via multi-head cross-attention edge selection and future-pose prediction, and is evaluated quantitatively against random-graph and full-graph baselines. In the Interaction Transfer stage, the EIG is used to define an interaction-consistency reward that guides reinforcement learning of a centralized hierarchical control policy for new embodiments. The transfer results are presented primarily through qualitative video stills, descriptive analysis, and a user study that compares the method to an IK-based retargeting baseline.

Significance. If the claims are substantiated, the paper makes a valuable contribution to cross-embodiment imitation learning: it proposes a compact, interpretable, and sparse graph representation of interaction dynamics that is explicitly designed to be transferable across morphologically distinct agents. The learned EIG is shown to be more predictive than random or fully connected graphs in the embedding module, which is a concrete, reproducible quantitative result. The idea of using the learned graph as an imitation objective rather than raw joint positions is novel and potentially impactful for robotics and character animation. However, the central claim of the paper—semantic cross-embodiment transfer—currently rests on qualitative demonstrations and a user study lacking statistical details, and on a vertex-correspondence heuristic that is not adequately validated. The framework is well motivated and the two-stage architecture is sensible, but the evidence for the main claim needs strengthening.

major comments (3)

[§3.3.1] The automatic vertex correspondence procedure is underspecified and may not preserve the semantic roles of interaction graph vertices when the new embodiment has a different number or arrangement of end-effectors. For example, human demonstrations have two arm end-effectors and two leg end-effectors, while Go2Ar has four paws and one gripper and Stretch has one gripper and two wheels. The text states that 'each end-effector in the new character is assigned to a unique end-effector in the embedded graph' but does not specify what happens when the embedded graph contains end-effectors with no counterpart in the new character, nor how the 'relative sequence to the end-effectors' rule assigns non-end-effector vertices such as the head or spine. If a human arm vertex is mapped to a paw or base, then any edge incident to that vertex in the embedded graph will drive the reward (Eqs. 8–12) toward moving the wrong body part, directly undermining the claimed semantic transfer. The paper never reports the actual vertex correspondences used in the experiments. Please provide the full mappings for each embodiment, describe how unmatched graph vertices are handled, and ideally add an analysis or ablation that verifies the assigned correspondences lead to the intended interaction semantics.
[§1.2, Figure 4D] The user study is the only quantitative evaluation of the Interaction Transfer module, yet the paper reports no error bars, no sample sizes per condition, and no statistical tests. The claim that 'our method significantly outperforms the baseline in both activity recognition and semantic consistency' is not backed by any p-values, confidence intervals, or effect sizes. This is a load-bearing gap because the central contribution of the paper is cross-embodiment interaction transfer, not just the pose-prediction accuracy of the embedding module. Please provide a full statistical analysis of the user study results, and consider supplementing it with an objective measure of interaction consistency—for example, the time-varying distance between the reference and generated embedded graph features (dl_t, ded_t, dcp_t)—so that the transfer quality can be assessed without relying solely on subjective ratings.
[§3.3.2, Eqs. (9) and (11)] The normalization lengths L_hat and L in the length and center-point metrics are described only as 'morphology-dependent length values ... predefined parameters defined according to the morphology of the agent.' No values or derivation are given for any of the four embodiments, even though these quantities directly scale two of the three terms in the interaction consistency reward and therefore affect the learned behavior. If these are hand-tuned, please report the values and justify them; if they are computed from the morphology (e.g., a characteristic body dimension), state the formula. Without this information, the reward design—and hence the transfer results—cannot be reproduced or properly assessed.

minor comments (6)

[Abstract] Typo: 'wholebbody' should be 'whole-body'.
[§1.1] Typo: 'correspondingtod' should be 'corresponding to'.
[§3.3.1] Typos: 'neural poses' should be 'neutral poses'; 'crossponding' should be 'corresponding'; 'elaboration' should be 'elaborate'; 'consistancy' should be 'consistency'.
[Figure 3d] The quantitative comparison of prediction error across graph configurations is described only in text; adding exact error values and error bars to the figure or a table would improve clarity and reproducibility.
[§3.2.3] The KL divergence weight β is said to be tuned, and the chosen value β = 0.3 is reported, but no sensitivity results are shown; a brief statement of the range explored would be helpful.
[References] Reference [68] is attributed to 'Authors, G.'; the citation should be updated to the actual authors of the Genesis environment.

Circularity Check

0 steps flagged · score 0.0 of 10

The derivation is self-contained: the learned embedded interaction graph is used as a training objective for imitation, not as a hidden restatement of the transfer outcome, so no circularity is present.

full rationale

BuddyImitation's derivation chain is: (i) learn a sparse interaction graph by minimizing pose reconstruction error on demonstrations (Eq. 5); (ii) use that graph's feature trajectory as an interaction consistency reward for RL policies (Eqs. 8-12). This is a standard learn-an-objective-then-optimize-it pipeline. The learned graph is not defined in terms of the transfer outcome; it is defined by predictive accuracy on the demonstration data. The transfer result (policies reproducing interaction semantics) is not guaranteed by construction: the RL optimization could fail to maximize the reward, and the authors evaluate it with ablations, comparisons to a retargeting baseline, and a user study. The same holds for the graph analysis: the claims that the four-edge embedded graph outperforms random graphs and captures arm/root relationships are empirical comparisons with independent baselines, not identities. Self-citations to CrossLoco [20] for the reward formulation and regularization and to ACE [16] for the vertex assignment heuristic are used as building blocks, not as uniqueness or existence theorems that force the paper's conclusions. The vertex-mapping concern raised by the skeptic is a potential correctness risk about semantic preservation under different morphologies, not a circularity: even if the mapping fails, the reward would be optimizing the wrong objective, which is the opposite of a by-construction success. The paper's own limitation statement (sharp turning motions cannot be fully replicated by Go2Ar due to limited turning agility) describes an embodiment constraint, not a circular dependency. No equation is equivalent to its own input, no fitted parameter is renamed as a prediction, and no load-bearing claim rests on an unverified self-citation. Therefore the circularity score is 0.

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

The framework rests on an inherited pretrained motion decoder, a physics simulator, and several hand-set hyperparameters. The central claim requires that the learned graph be a sufficient semantic summary and that the automatic vertex correspondence preserve edge semantics across morphologies.

free parameters (5)

Number of edges in embedded graph (default 4) = 4
Chosen from ablation over 0,1,4,8,full edges (Fig. 3d); E4 balances accuracy and efficiency.
KL weight beta in MVAE motion decoder = 0.3
Hand-tuned in Section 3.2.3 to balance latent space usage and motion quality.
Graph consistency loss weight lambda = not specified
Used in Eq. (7) but its value is not reported.
Reward weights wl, wed, wcp, wfar = not specified
Used in Eq. (12); wfar is set 'significantly lower' but exact values are omitted.
Morphology normalization lengths Lhat, L = predefined per morphology
Used in Eqs. (9) and (11); described as predefined parameters without values.

assumptions (5)

domain assumption The pretrained MVAE motion decoder provides a valid generative prior for human pose transitions.
Section 3.2.3 adopts the MVAE architecture from [69]; the quality of the decoder is inherited, not validated in this paper.
domain assumption The Genesis physics simulator faithfully models contacts and dynamics for the simulated agents.
Section 1 uses Genesis [68] as the backend; no validation against real hardware is provided.
domain assumption Interaction dynamics are Markovian: P(q_{t+1}|q_1,...,q_T)=P(q_{t+1}|q_t).
Equation (3) in Section 3.2.1 introduces this modeling assumption.
domain assumption All characters' joint positions can be reconstructed from one character's pose and the interaction graph feature.
Equation (1) in Section 3.1.1 is taken from the Interaction Graph formulation [1].
domain assumption Pretrained low-level policies initialized by motion primitive pretraining generalize to interaction learning.
Section 3.3.4 invokes Imitate-and-Repurpose [70]; the character-specific dataset and pretraining details are not described.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Learning Physical Interaction Skills from Human Demonstrations." pith.science (2026). https://pith.science/paper/F2MD3BW4

@misc{pith2026250720445,
  author       = {Pith},
  title        = {Pith review of: Learning Physical Interaction Skills from Human Demonstrations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F2MD3BW4}},
  note         = {Machine review of arXiv:2507.20445}
}

read the original abstract

Learning physical interaction skills, such as dancing, handshaking, or sparring, remains a fundamental challenge for agents operating in human environments, particularly when the agent's morphology differs significantly from that of the demonstrator. Existing approaches often rely on handcrafted objectives or morphological similarity, limiting their capacity for generalization. Here, we introduce a framework that enables agents with diverse embodiments to learn wholebbody interaction behaviors directly from human demonstrations. The framework extracts a compact, transferable representation of interaction dynamics, called the Embedded Interaction Graph (EIG), which captures key spatiotemporal relationships between the interacting agents. This graph is then used as an imitation objective to train control policies in physics-based simulations, allowing the agent to generate motions that are both semantically meaningful and physically feasible. We demonstrate BuddyImitation on multiple agents, such as humans, quadrupedal robots with manipulators, or mobile manipulators and various interaction scenarios, including sparring, handshaking, rock-paper-scissors, or dancing. Our results demonstrate a promising path toward coordinated behaviors across morphologically distinct characters via cross embodiment interaction learning.

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 64 canonical work pages

[1]

1–11 (2023)

Zhang, Y., Gopinath, D., Ye, Y., Hodgins, J., Turk, G., Won, J.: Simulation and retargeting of complex multi-character interactions, pp. 1–11 (2023)

work page 2023
[2]

33–42 (1998) 20

Gleicher, M.: Retargetting motion to new characters, pp. 33–42 (1998) 20

work page 1998
[3]

The Journal of Visualization and Computer Animation 11(5), 223–235 (2000)

Choi, K.-J., Ko, H.-S.: Online motion retargetting. The Journal of Visualization and Computer Animation 11(5), 223–235 (2000)

work page 2000
[4]

ACM Transac- tions on Graphics (TOG) 24(1), 98–117 (2005)

Tak, S., Ko, H.-S.: A physically-based motion retargeting filter. ACM Transac- tions on Graphics (TOG) 24(1), 98–117 (2005)

work page 2005
[5]

4142–4149 (2017)

Delhaisse, B., Esteban, D., Rozo, L., Caldwell, D.: Transfer learning of shared latent spaces between robots with similar kinematic structure, pp. 4142–4149 (2017). IEEE

work page 2017
[6]

1–2 (2018)

Jang, H., Kwon, B., Yu, M., Kim, S.U., Kim, J.: A variational u-net for motion retargeting, pp. 1–2 (2018)

work page 2018
[7]

8639–8648 (2018)

Villegas, R., Yang, J., Ceylan, D., Lee, H.: Neural kinematic networks for unsupervised motion retargetting, pp. 8639–8648 (2018)

work page 2018
[8]

ACM Transactions on Graphics (TOG) 39(4), 62–1 (2020)

Aberman, K., Li, P., Lischinski, D., Sorkine-Hornung, O., Cohen-Or, D., Chen, B.: Skeleton-aware networks for deep motion retargeting. ACM Transactions on Graphics (TOG) 39(4), 62–1 (2020)

work page 2020

Show all 70 references

[9]

In: 2010 10th IEEE-RAS Interna- tional Conference on Humanoid Robots, pp

Yamane, K., Anderson, S.O., Hodgins, J.K.: Controlling humanoid robots with human motion data: Experimental validation. In: 2010 10th IEEE-RAS Interna- tional Conference on Humanoid Robots, pp. 504–510 (2010). IEEE

2010
[10]

Proceedings of Robotics: Science and Systems (R: SS) (2020)

Choi, S., Pan, M., Kim, J.: Nonparametric motion retargeting for humanoid robots on shared latent space. Proceedings of Robotics: Science and Systems (R: SS) (2020)

2020
[11]

ACM SIGGRAPH 2022 Conference Proceedings (2022)

Aigerman, N., Gupta, K., Kim, V.G., Chaudhuri, S., Saito, J., Groueix, T.: Neural jacobian fields: Learning intrinsic mappings of arbitrary meshes. ACM SIGGRAPH 2022 Conference Proceedings (2022)

2022
[12]

ACM Transactions on Graphics (TOG) 34(6), 1–12 (2015)

Rhodin, H., Tompkin, J., Kim, K.I., De Aguiar, E., Pfister, H., Seidel, H.-P., Theobalt, C.: Generalizing wave gestures from sparse examples for real-time character control. ACM Transactions on Graphics (TOG) 34(6), 1–12 (2015)

2015
[13]

In: Computer Graphics Forum, vol

Rhodin, H., Tompkin, J., In Kim, K., Varanasi, K., Seidel, H.-P., Theobalt, C.: Interactive motion mapping for real-time character control. In: Computer Graphics Forum, vol. 33, pp. 273–282 (2014)

2014
[14]

In: Proceedings of the 12th ACM SIGGRAPH/Euro- graphics Symposium on Computer Animation, pp

Seol, Y., O’Sullivan, C., Lee, J.: Creature features: online motion puppetry for non-human characters. In: Proceedings of the 12th ACM SIGGRAPH/Euro- graphics Symposium on Computer Animation, pp. 213–221 (2013)

2013
[15]

Robotics: Science and Systems (2022) 21

Kim, S., Sorokin, M., Lee, J., Ha, S.: Human motion control of quadrupedal robots using deep reinforcement learning. Robotics: Science and Systems (2022) 21

2022
[16]

In: SIGGRAPH Asia 2023 Conference Papers, pp

Li, T., Won, J., Clegg, A., Kim, J., Rai, A., Ha, S.: Ace: Adversarial corre- spondence embedding for cross morphology motion retargeting from human to nonhuman characters. In: SIGGRAPH Asia 2023 Conference Papers, pp. 1–11 (2023)

2023
[17]

ACM Transactions on Graphics (TOG) 42(4), 1–14 (2023)

Grandia, R., Farshidian, F., Knoop, E., Schumacher, C., Hutter, M., B¨ acher, M.: Doc: Differentiable optimal control for retargeting motions onto legged robots. ACM Transactions on Graphics (TOG) 42(4), 1–14 (2023)

2023
[18]

: Genloco: Generalized locomotion controllers for quadrupedal robots

Feng, G., Zhang, H., Li, Z., Peng, X.B., Basireddy, B., Yue, L., Song, Z., Yang, L., Liu, Y., Sreenath, K., et al. : Genloco: Generalized locomotion controllers for quadrupedal robots. In: Conference on Robot Learning, pp. 1893–1903 (2023). PMLR

2023
[19]

International Conference on Robotics and Automation(ICRA) (2024)

Shafiee, M., Bellegarda, G., Ijspeert, A.: Manyquadrupeds: Learning a single loco- motion policy for diverse quadruped robots. International Conference on Robotics and Automation(ICRA) (2024)

2024
[20]

International Conference on Learning Representations (2024)

Li, T., Jung, H., Gombolay, M., Cho, Y.K., Ha, S.: Crossloco: Human motion driven control of legged robots via guided unsupervised reinforcement learning. International Conference on Learning Representations (2024)

2024
[21]

Preprint at: https://arxiv.org/abs/2502.05271 (2025)

Li, T., Truong, J., Yang, J., Clegg, A., Rai, A., Ha, S., Puig, X.: Robotmover: Learning to move large objects by imitating the dynamic chain. Preprint at: https://arxiv.org/abs/2502.05271 (2025)

2025 arXiv
[22]

Conference on Robot Learning (2024)

Bohlinger, N., Czechmanowski, G., Krupka, M., Kicki, P., Walas, K., Peters, J., Tateo, D.: One policy to run them all: an end-to-end learning approach to multi-embodiment locomotion. Conference on Robot Learning (2024)

2024
[23]

215–222 (2006)

Liu, C.K., Hertzmann, A., Popovi´ c , Z.: Composition of complex optimal multi- character motions, pp. 215–222 (2006)

2006
[24]

IEEE transactions on visualization and computer graphics 14(3), 707– 720 (2008)

Kwon, T., Cho, Y.-S., Park, S.I., Shin, S.Y.: Two-character motion analysis and synthesis. IEEE transactions on visualization and computer graphics 14(3), 707– 720 (2008)

2008
[25]

IEEE Transactions on Visualization and Computer Graphics 18(5), 741–752 (2010)

Shum, H.P., Komura, T., Yamazaki, S.: Simulating multiple character interactions with collaborative and adversarial goals. IEEE Transactions on Visualization and Computer Graphics 18(5), 741–752 (2010)

2010
[26]

898–906 (2006)

Lee, K.H., Choi, M.G., Lee, J.: Motion patches: building blocks for virtual environments annotated with motion data, pp. 898–906 (2006)

2006
[27]

117–126 (2012) 22

Kim, M., Hwang, Y., Hyun, K., Lee, J.: Tiling motion patches, pp. 117–126 (2012) 22

2012
[28]

ACM Transactions on Graphics (TOG) 33(6), 1–12 (2014)

Won, J., Lee, K., O’Sullivan, C., Hodgins, J.K., Lee, J.: Generating and rank- ing diverse multi-character interactions. ACM Transactions on Graphics (TOG) 33(6), 1–12 (2014)

2014
[29]

207–214 (2009)

Yersin, B., Ma¨ ım, J., Pettr´ e , J., Thalmann, D.: Crowd patches: populating large- scale virtual environments for real-time applications, pp. 207–214 (2009)

2009
[30]

International Journal of Computer Vision (IJCV) (2024)

Liang, H., Zhang, W., Li, W., Yu, J., Xu, L.: Intergen: Diffusion-based multi- human motion generation under complex interactions. International Journal of Computer Vision (IJCV) (2024)

2024
[31]

International Conference on Learning Representations (2024)

Shafir, Y., Tevet, G., Kapon, R., Bermano, A.H.: Human motion diffusion as a generative prior. International Conference on Learning Representations (2024)

2024
[32]

45–53 (2013)

Al-Asqhar, R.A., Komura, T., Choi, M.G.: Relationship descriptors for interactive motion adaptation, pp. 45–53 (2013)

2013
[33]

ACM Transactions on Graphics (TOG) 33(4), 1–10 (2014)

Kim, J., Seol, Y., Kwon, T., Lee, J.: Interactive manipulation of large-scale crowd animation. ACM Transactions on Graphics (TOG) 33(4), 1–10 (2014)

2014
[34]

Computer Animation and Virtual Worlds 32(3-4), 2015 (2021)

Kim, J., Seol, Y., Kwon, T.: Interactive multi-character motion retargeting. Computer Animation and Virtual Worlds 32(3-4), 2015 (2021)

2021
[35]

97–106 (2014)

Ho, E.S., Wang, H., Komura, T.: A multi-resolution approach for adapting close character interaction, pp. 97–106 (2014)

2014
[36]

In: ACM SIGGRAPH 2010 Papers, pp

Ho, E.S., Komura, T., Tai, C.-L.: Spatial relationship preserving character motion adaptation. In: ACM SIGGRAPH 2010 Papers, pp. 1–8 (2010)

2010
[37]

Jin, T., Kim, M., Lee, S.-H.: Aura mesh: Motion retargeting to preserve the spatial relationships between skinned characters, vol. 37, pp. 311–320 (2018). Wiley Online Library

2018
[38]

ACM Transactions on Graphics (ToG)31(4), 1–8 (2012)

Mordatch, I., Todorov, E., Popovi´ c , Z.: Discovery of complex behaviors through contact-invariant optimization. ACM Transactions on Graphics (ToG)31(4), 1–8 (2012)

2012
[39]

446–453 (2017)

Otani, K., Bouyarmane, K.: Adaptive whole-body manipulation in human-to- humanoid multi-contact motion retargeting, pp. 446–453 (2017). IEEE

2017
[40]

IEEE transactions on visualization and computer graphics 23(6), 1650–1662 (2016)

Vaillant, J., Bouyarmane, K., Kheddar, A.: Multi-character physical and behav- ioral interactions controller. IEEE transactions on visualization and computer graphics 23(6), 1650–1662 (2016)

2016
[41]

ACM Transactions on Graphics (TOG) 40(4), 1–11 (2021) 23

Won, J., Gopinath, D., Hodgins, J.: Control strategies for physically simu- lated characters performing two-player competitive sports. ACM Transactions on Graphics (TOG) 40(4), 1–11 (2021) 23

2021
[42]

ACM Transactions on Graphics (TOG) 26(3), 105 (2007)

Yin, K., Loken, K., Panne, M.: Simbicon: Simple biped locomotion control. ACM Transactions on Graphics (TOG) 26(3), 105 (2007)

2007
[43]

1–7 (2008)

Yin, K., Coros, S., Beaudoin, P., Panne, M.: Continuation methods for adapting simulated skills, pp. 1–7 (2008)

2008
[44]

ACM Transactions On Graphics (TOG) 29(4), 1–9 (2010)

Coros, S., Beaudoin, P., Panne, M.: Generalized biped walking control. ACM Transactions On Graphics (TOG) 29(4), 1–9 (2010)

2010
[45]

ACM transactions on graphics (TOG) 29(4), 1–10 (2010)

De Lasa, M., Mordatch, I., Hertzmann, A.: Feature-based locomotion controllers. ACM transactions on graphics (TOG) 29(4), 1–10 (2010)

2010
[46]

ACM Transactions on Graphics (TOG) 31(6), 154–1 (2012)

Liu, L., Yin, K., Panne, M., Guo, B.: Terrain runner: control, parameterization, composition, and planning for highly dynamic motions. ACM Transactions on Graphics (TOG) 31(6), 154–1 (2012)

2012
[47]

ACM Transactions on Graphics (TOG) 33(4), 1–12 (2014)

Tan, J., Gu, Y., Liu, C.K., Turk, G.: Learning bicycle stunts. ACM Transactions on Graphics (TOG) 33(4), 1–12 (2014)

2014
[48]

Science Robotics 4(26), 5872 (2019)

Hwangbo, J., Lee, J., Dosovitskiy, A., Bellicoso, D., Tsounis, V., Koltun, V., Hutter, M.: Learning agile and dynamic motor skills for legged robots. Science Robotics 4(26), 5872 (2019)

2019
[49]

Science robotics 5(47), 5986 (2020)

Lee, J., Hwangbo, J., Wellhausen, L., Koltun, V., Hutter, M.: Learning quadrupedal locomotion over challenging terrain. Science robotics 5(47), 5986 (2020)

2020
[50]

Luo, Z., Cao, J., Winkler, A.W., Kitani, K., Xu, W.: Perpetual humanoid control for real-time simulated avatars. (2023)

2023
[51]

ACM Transactions on Graphics (TOG) 37(4), 143–114314 (2018) https://doi.org/10.1145/3197517

Peng, X.B., Abbeel, P., Levine, S., Panne, M.: Deepmimic: Example-guided deep reinforcement learning of physics-based character skills. ACM Transactions on Graphics (TOG) 37(4), 143–114314 (2018) https://doi.org/10.1145/3197517. 3201311

2018 doi
[52]

Peng, X.B., Coumans, E., Zhang, T., Lee, T.-W.E., Tan, J., Levine, S.: Learning agile robotic locomotion skills by imitating animals. (2020). https://doi.org/10. 15607/RSS.2020.XVI.064

2020
[53]

Smith, L.M., Kew, J.C., Li, T., Luu, L., Peng, X.B., Ha, S., Tan, J., Levine, S.: Learning and adapting agile locomotion skills by transferring experience. (2023). https://doi.org/10.15607/RSS.2023.XIX.051 . https://doi.org/10.15607/ RSS.2023.XIX.051

2023 doi
[54]

ACM Transactions on Graphics (TOG) 41(4), 1–12 (2022)

Won, J., Gopinath, D., Hodgins, J.: Physics-based character controllers using conditional vaes. ACM Transactions on Graphics (TOG) 41(4), 1–12 (2022)

2022
[55]

263–269 (2019)

Li, T., Geyer, H., Atkeson, C.G., Rai, A.: Using deep reinforcement learning to 24 learn high-level policies on the atrias biped, pp. 263–269 (2019). IEEE

2019
[56]

1241–1246 (2018)

Xie, Z., Berseth, G., Clary, P., Hurst, J., Panne, M.: Feedback control for cassie with deep reinforcement learning, pp. 1241–1246 (2018). IEEE

2018
[57]

ACM Transactions on Graphics (TOG) 40(4) (2021) https://doi.org/10.1145/3450626.3459670

Peng, X.B., Ma, Z., Abbeel, P., Levine, S., Kanazawa, A.: Amp: Adversarial motion priors for stylized physics-based character control. ACM Transactions on Graphics (TOG) 40(4) (2021) https://doi.org/10.1145/3450626.3459670

2021
[58]

ACM Transac- tions on Graphics (TOG) 41(4) (2022)

Peng, X.B., Guo, Y., Halper, L., Levine, S., Fidler, S.: Ase: Large-scale reusable adversarial skill embeddings for physically simulated characters. ACM Transac- tions on Graphics (TOG) 41(4) (2022)

2022
[59]

ACM Transactions on Graphics 42(4) (2023)

Xu, P., Shang, X., Zordan, V., Karamouzas, I.: Composite motion learning with task control. ACM Transactions on Graphics 42(4) (2023)

2023
[60]

Science Robotics 9(88), 7566 (2024)

Hoeller, D., Rudin, N., Sako, D., Hutter, M.: Anymal parkour: Learning agile navigation for quadrupedal robots. Science Robotics 9(88), 7566 (2024)

2024
[61]

In: ACM SIGGRAPH 2022 Conference Proceed- ings, pp

Xie, Z., Starke, S., Ling, H.Y., Panne, M.: Learning soccer juggling skills with layer-wise mixture-of-experts. In: ACM SIGGRAPH 2022 Conference Proceed- ings, pp. 1–9 (2022)

2022
[62]

: From motor control to team play in simulated humanoid football

Liu, S., Lever, G., Wang, Z., Merel, J., Eslami, S.A., Hennes, D., Czarnecki, W.M., Tassa, Y., Omidshafiei, S., Abdolmaleki, A., et al. : From motor control to team play in simulated humanoid football. Science Robotics 7(69), 0235 (2022)

2022
[63]

2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 1479–1486 (2022)

Ji, Y., Li, Z., Sun, Y., Peng, X.B., Levine, S., Berseth, G., Sreenath, K.: Hierarchi- cal reinforcement learning for precise soccer shooting skills using a quadrupedal robot. 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 1479–1486 (2022)

2022
[64]

Preprint at: https://arxiv.org/abs/2304.13653 (2023)

Haarnoja, T., Moran, B., Lever, G., Huang, S.H., Tirumala, D., Wulfmeier, M., Humplik, J., Tunyasuvunakool, S., Siegel, N.Y., Hafner, R., et al.: Learning agile soccer skills for a bipedal robot with deep reinforcement learning. Preprint at: https://arxiv.org/abs/2304.13653 (2023)

2023 arXiv
[65]

ACM Transactions on Graphics (TOG) 42(4) (2023) https://doi.org/10.1145/ 3592408

Zhang, H., Yuan, Y., Makoviychuk, V., Guo, Y., Fidler, S., Peng, X.B., Fata- halian, K.: Learning physically simulated tennis skills from broadcast videos. ACM Transactions on Graphics (TOG) 42(4) (2023) https://doi.org/10.1145/ 3592408

2023
[66]

ACM Transactions on Graphics (TOG) 39(4), 33–1 (2020)

Won, J., Gopinath, D., Hodgins, J.: A scalable approach to control diverse behav- iors for physically simulated characters. ACM Transactions on Graphics (TOG) 39(4), 33–1 (2020)

2020
[67]

IEEE Robotics and Automation Letters 6(2), 2682–2689 (2021)

Li, T., Calandra, R., Pathak, D., Tian, Y., Meier, F., Rai, A.: Planning in learned 25 latent action spaces for generalizable legged locomotion. IEEE Robotics and Automation Letters 6(2), 2682–2689 (2021)

2021
[68]

https://github.com/Genesis-Embodied-AI/Genesis

Authors, G.: Genesis: A Universal and Generative Physics Engine for Robotics and Beyond. https://github.com/Genesis-Embodied-AI/Genesis
[69]

ACM Transactions on Graphics (TOG) 39(4), 40–1 (2020)

Ling, H.Y., Zinno, F., Cheng, G., Van De Panne, M.: Character controllers using motion vaes. ACM Transactions on Graphics (TOG) 39(4), 40–1 (2020)

2020
[70]

Preprint at: https://arxiv.org/abs/2203.17138 (2022) 26

Bohez, S., Tunyasuvunakool, S., Brakel, P., Sadeghi, F., Hasenclever, L., Tassa, Y., Parisotto, E., Humplik, J., Haarnoja, T., Hafner, R., et al.: Imitate and repur- pose: Learning reusable robot movement skills from human and animal behaviors. Preprint at: https://arxiv.org/a...

2022 arXiv

Pith tools

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

[1] [1]

1–11 (2023)

Zhang, Y., Gopinath, D., Ye, Y., Hodgins, J., Turk, G., Won, J.: Simulation and retargeting of complex multi-character interactions, pp. 1–11 (2023)

work page 2023

[2] [2]

33–42 (1998) 20

Gleicher, M.: Retargetting motion to new characters, pp. 33–42 (1998) 20

work page 1998

[3] [3]

The Journal of Visualization and Computer Animation 11(5), 223–235 (2000)

Choi, K.-J., Ko, H.-S.: Online motion retargetting. The Journal of Visualization and Computer Animation 11(5), 223–235 (2000)

work page 2000

[4] [4]

ACM Transac- tions on Graphics (TOG) 24(1), 98–117 (2005)

Tak, S., Ko, H.-S.: A physically-based motion retargeting filter. ACM Transac- tions on Graphics (TOG) 24(1), 98–117 (2005)

work page 2005

[5] [5]

4142–4149 (2017)

Delhaisse, B., Esteban, D., Rozo, L., Caldwell, D.: Transfer learning of shared latent spaces between robots with similar kinematic structure, pp. 4142–4149 (2017). IEEE

work page 2017

[6] [6]

1–2 (2018)

Jang, H., Kwon, B., Yu, M., Kim, S.U., Kim, J.: A variational u-net for motion retargeting, pp. 1–2 (2018)

work page 2018

[7] [7]

8639–8648 (2018)

Villegas, R., Yang, J., Ceylan, D., Lee, H.: Neural kinematic networks for unsupervised motion retargetting, pp. 8639–8648 (2018)

work page 2018

[8] [8]

ACM Transactions on Graphics (TOG) 39(4), 62–1 (2020)

Aberman, K., Li, P., Lischinski, D., Sorkine-Hornung, O., Cohen-Or, D., Chen, B.: Skeleton-aware networks for deep motion retargeting. ACM Transactions on Graphics (TOG) 39(4), 62–1 (2020)

work page 2020