pith. sign in

arxiv: 2505.18780 · v3 · submitted 2025-05-24 · 💻 cs.RO · cs.LG

DreamPolicy: A Unified World-model Policy for Scalable Humanoid Locomotion

Yahao Fan , Tianxiang Gui , Kaiyang Ji , Shutong Ding , Chixuan Zhang , Yifeng Xu , Ke Yang , Jiayuan Gu
show 3 more authors
Jingyi Yu Jingya Wang Ye Shi
This is my paper

Pith reviewed 2026-05-19 12:46 UTC · model grok-4.3

classification 💻 cs.RO cs.LG
keywords humanoid locomotiondiffusion world modelzero-shot transferunified policyterrain adaptationrobot controloffline datascalable locomotion
0
0 comments X

The pith

A single humanoid policy guided by an autoregressive diffusion world model can generalize to unseen composite terrains without distillation or manual rewards.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper seeks to replace the common practice of training separate policies for different terrains and then distilling them into one student policy. Instead it trains a terrain-aware autoregressive diffusion model on rollouts collected from those specialized policies, then uses the model to generate future state trajectories that become dynamic objectives for a single conditioned policy. If the generated trajectories remain physically plausible, the policy learns general locomotion skills that transfer to novel combinations of terrains never seen in training. This removes the need for hand-designed reward functions and lets performance improve simply by adding more offline data. Experiments show gains of up to 27 percent on unseen terrains and 38 percent on combined terrains relative to the strongest baseline.

Core claim

The central claim is that an autoregressive diffusion world model trained on aggregated rollouts from specialized policies can synthesize physically plausible future trajectories that serve as effective dynamic objectives for a single conditioned policy, thereby enabling robust zero-shot transfer to unseen composite terrains while naturally scaling with additional offline data.

What carries the argument

Terrain-aware autoregressive diffusion world model that generates future trajectories used as dynamic objectives for the conditioned policy.

If this is right

  • A single policy can master both previously trained and entirely novel terrain combinations without explicit skill composition.
  • Reward engineering is bypassed because the world model supplies the objectives directly from data.
  • Performance continues to rise as the offline dataset grows because the diffusion model acquires richer locomotion patterns.
  • The framework unifies world-model planning and policy learning in one loop, removing the one-task-one-policy pattern.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same world-model-plus-conditioned-policy structure could be tested on manipulation or navigation tasks where composite environments also appear.
  • If the diffusion model can be fine-tuned online from real-robot failures, the zero-shot transfer gap might shrink further without new simulation data.
  • The approach suggests that learning a generative model of future states may be more data-efficient than distilling explicit teacher policies for high-dimensional control.

Load-bearing premise

The diffusion model trained only on rollouts from existing specialized policies will still produce trajectories that remain physically valid and useful when the policy is deployed on terrain combinations never encountered during data collection.

What would settle it

Run the trained policy on a composite terrain whose height map and friction values were withheld from both the specialized policies and the world-model training set, then measure whether the generated trajectories diverge from actual robot dynamics enough to cause frequent falls or stuck states.

Figures

Figures reproduced from arXiv: 2505.18780 by Chixuan Zhang, Jiayuan Gu, Jingya Wang, Jingyi Yu, Kaiyang Ji, Ke Yang, Shutong Ding, Tianxiang Gui, Yahao Fan, Ye Shi, Yifeng Xu.

Figure 1
Figure 1. Figure 1: The policy learns to imagine terrain-aware humanoid motions (HMI), enabling zero-shot [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Framework of DreamPolicy. The system is decomposed into two parts: (1) Terrain-aware [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Scenarios visualization. These figures are all the terrain used for evaluation. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison in Single terrain 8 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Success rate on Slope-Bridge with expert [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
read the original abstract

Achieving versatile humanoid locomotion with a single policy presents a critical scalability challenge. Prevailing methods often rely on distilling multiple terrain-specific teacher policies into a unified student policy. However, while such distillation captures basic locomotion primitives, it struggles to organically compose these skills to adapt to complex environments, resulting in poor generalization to novel composite terrains unseen during training. To overcome this, we present DreamPolicy, a unified framework that integrates offline data with a diffusion-based world model, enabling a single policy to master both known and unseen terrains. Central to our approach is a terrain-aware world model, driven by an autoregressive diffusion world model trained on aggregated rollouts from specialized policies. This model synthesizes physically plausible future trajectories, which serve as dynamic objectives for a conditioned policy, thereby bypassing manual reward engineering. Unlike distillation, our world model captures generalizable locomotion skills, allowing for robust zero-shot transfer to unseen composite terrains. DreamPolicy naturally scales with data availability. As the offline dataset expands, the diffusion world model continuously acquires richer skills. Experiments demonstrate that DreamPolicy outperforms the strongest baseline by up to 27\% on unseen terrains and 38\% on combined terrains. By unifying world model-based planning and policy learning, DreamPolicy breaks the "one task, one policy" bottleneck and establishes a scalable, data-driven paradigm for generalist humanoid control.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript presents DreamPolicy, a unified framework for humanoid locomotion control that integrates offline rollouts with a terrain-aware autoregressive diffusion world model. The world model is trained on aggregated data from specialized single-terrain policies and generates future trajectories that serve as dynamic objectives for a conditioned policy, with the goal of achieving better composition of locomotion skills and robust zero-shot transfer to unseen composite terrains than distillation-based baselines. The abstract reports performance gains of up to 27% on unseen terrains and 38% on combined terrains.

Significance. If the generalization claims are substantiated, the approach could meaningfully advance scalable humanoid control by replacing per-task distillation with a data-driven world-model-plus-policy pipeline that improves with additional offline data and avoids manual reward engineering.

major comments (2)
  1. [Abstract] Abstract: the reported 27% and 38% gains are presented without any description of experimental protocols, baseline implementations, number of evaluation episodes, statistical significance testing, or terrain-generation procedures, preventing verification that the numbers support the central zero-shot transfer claim.
  2. [Method] Method section describing the diffusion world model: training occurs exclusively on aggregated rollouts from single-terrain specialized policies, yet no explicit mechanism (terrain embedding details, physics-informed loss terms, or training data containing terrain transitions) is provided to guarantee that synthesized trajectories remain physically plausible at boundaries of composite terrains.
minor comments (1)
  1. [Experiments] Figure captions and axis labels in the results section could be expanded to include the exact terrain parameters used for the composite test cases.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments, which highlight important aspects of clarity and substantiation. We address each major comment point by point below, indicating planned revisions where appropriate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the reported 27% and 38% gains are presented without any description of experimental protocols, baseline implementations, number of evaluation episodes, statistical significance testing, or terrain-generation procedures, preventing verification that the numbers support the central zero-shot transfer claim.

    Authors: We agree that the abstract would benefit from additional context to support the reported gains. Due to length constraints, we will revise the abstract to include a concise mention of the evaluation protocol (e.g., number of episodes per terrain, baseline comparisons, and terrain generation approach) while directing readers to the Experiments section for full protocols, statistical testing, and implementation details. This change will better substantiate the zero-shot transfer claims without altering the core results. revision: partial

  2. Referee: [Method] Method section describing the diffusion world model: training occurs exclusively on aggregated rollouts from single-terrain specialized policies, yet no explicit mechanism (terrain embedding details, physics-informed loss terms, or training data containing terrain transitions) is provided to guarantee that synthesized trajectories remain physically plausible at boundaries of composite terrains.

    Authors: The terrain-aware property is achieved by conditioning the autoregressive diffusion process on terrain embeddings derived from visual and proprioceptive observations, allowing the model to adapt predictions to upcoming terrain features. The aggregated training data includes sequences that implicitly capture transitions between terrain types encountered during single-terrain rollouts. We acknowledge that explicit discussion of boundary handling, embedding architecture, and any physics-informed regularizations was insufficient. We will expand the Method section with these details, including diagrams of the conditioning mechanism and clarification on how autoregressive generation promotes plausible transitions at composite boundaries. revision: yes

Circularity Check

0 steps flagged

No circularity: pipeline uses external offline rollouts and empirical training without definitional reduction

full rationale

The paper's core pipeline trains an autoregressive diffusion world model on aggregated rollouts collected from separate specialized policies, then conditions a policy on trajectories synthesized by that model. This structure depends on external data generation and standard supervised training rather than defining the target generalization performance in terms of the model's own fitted parameters or prior self-citations. No equation or claim reduces the zero-shot transfer result to a quantity that is true by construction; the reported gains on unseen terrains are presented as measured outcomes.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests primarily on the domain assumption that rollouts from terrain-specific policies contain the information needed for a diffusion world model to learn generalizable locomotion dynamics.

free parameters (1)
  • diffusion model training hyperparameters
    Parameters controlling the autoregressive diffusion process are chosen or tuned during training on the aggregated rollouts.
axioms (1)
  • domain assumption Aggregated rollouts from specialized policies contain sufficient information to train a generalizable world model for locomotion skills.
    Invoked as the foundation for the terrain-aware world model capturing composable skills.

pith-pipeline@v0.9.0 · 5804 in / 1282 out tokens · 75481 ms · 2026-05-19T12:46:28.525573+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

88 extracted references · 88 canonical work pages · 10 internal anchors

  1. [1]

    On the stability of anthropomorphic systems

    Miomir Vukobratovi´c and Juri Stepanenko. On the stability of anthropomorphic systems. Mathematical biosciences, 15(1-2):1–37, 1972

    work page 1972

  2. [2]

    Humanoid whole-body locomotion on narrow terrain via dynamic balance and reinforcement learning

    Weiji Xie, Chenjia Bai, Jiyuan Shi, Junkai Yang, Yunfei Ge, Weinan Zhang, and Xuelong Li. Humanoid whole-body locomotion on narrow terrain via dynamic balance and reinforcement learning. arXiv preprint arXiv:2502.17219, 2025

    work page arXiv 2025

  3. [3]

    Extreme parkour with legged robots

    Xuxin Cheng, Kexin Shi, Ananye Agarwal, and Deepak Pathak. Extreme parkour with legged robots. In 2024 IEEE International Conference on Robotics and Automation (ICRA) , pages 11443–11450. IEEE, 2024

    work page 2024

  4. [4]

    Robot parkour learning

    Ziwen Zhuang, Zipeng Fu, Jianren Wang, Christopher Atkeson, Soeren Schwertfeger, Chelsea Finn, and Hang Zhao. Robot parkour learning. arXiv preprint arXiv:2309.05665, 2023

    work page arXiv 2023

  5. [5]

    Robotkeyframing: Learning locomotion with high-level objectives via mixture of dense and sparse rewards

    Fatemeh Zargarbashi, Jin Cheng, Dongho Kang, Robert Sumner, and Stelian Coros. Robotkeyframing: Learning locomotion with high-level objectives via mixture of dense and sparse rewards. arXiv preprint arXiv:2407.11562, 2024

    work page arXiv 2024

  6. [6]

    Minimizing energy consumption leads to the emergence of gaits in legged robots

    Zipeng Fu, Ashish Kumar, Jitendra Malik, and Deepak Pathak. Minimizing energy consumption leads to the emergence of gaits in legged robots. In Conference on Robot Learning (CoRL), 2021

    work page 2021

  7. [7]

    Deep whole-body control: Learning a unified policy for manipulation and locomotion

    Zipeng Fu, Xuxin Cheng, and Deepak Pathak. Deep whole-body control: Learning a unified policy for manipulation and locomotion. In Conference on Robot Learning (CoRL), 2022

    work page 2022

  8. [8]

    Visual whole-body control for legged loco-manipulation

    Minghuan Liu, Zixuan Chen, Xuxin Cheng, Yandong Ji, Rizhao Qiu, Ruihan Yang, and Xiaolong Wang. Visual whole-body control for legged loco-manipulation. The 8th Conference on Robot Learning, 2024. 10

    work page 2024

  9. [9]

    Humanoid parkour learning

    Ziwen Zhuang, Shenzhe Yao, and Hang Zhao. Humanoid parkour learning. In 8th An- nual Conference on Robot Learning , 2024. URL https://openreview.net/forum?id= fs7ia3FqUM

    work page 2024

  10. [10]

    Learning humanoid locomotion over challenging terrain

    Ilija Radosavovic, Sarthak Kamat, Trevor Darrell, and Jitendra Malik. Learning humanoid locomotion over challenging terrain. arXiv preprint arXiv:2410.03654, 2024

    work page arXiv 2024

  11. [11]

    Humanoid locomotion as next token prediction

    Ilija Radosavovic, Bike Zhang, Baifeng Shi, Jathushan Rajasegaran, Sarthak Kamat, Trevor Darrell, Koushil Sreenath, and Jitendra Malik. Humanoid locomotion as next token prediction. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    work page 2024

  12. [12]

    Rhino: Learning real-time humanoid-human- object interaction from human demonstrations

    Jingxiao Chen, Xinyao Li, Jiahang Cao, Zhengbang Zhu, Wentao Dong, Minghuan Liu, Ying Wen, Yong Yu, Liqing Zhang, and Weinan Zhang. Rhino: Learning real-time humanoid-human- object interaction from human demonstrations. arXiv preprint arXiv:2502.13134, 2025

    work page arXiv 2025

  13. [13]

    Humanoid locomotion and manipulation: Current progress and challenges in control, planning, and learning,

    Zhaoyuan Gu, Junheng Li, Wenlan Shen, Wenhao Yu, Zhaoming Xie, Stephen McCrory, Xianyi Cheng, Abdulaziz Shamsah, Robert Griffin, C Karen Liu, et al. Humanoid locomotion and manipulation: Current progress and challenges in control, planning, and learning. arXiv preprint arXiv:2501.02116, 2025

    work page arXiv 2025

  14. [14]

    Open-television: Teleop- eration with immersive active visual feedback

    Xuxin Cheng, Jialong Li, Shiqi Yang, Ge Yang, and Xiaolong Wang. Open-television: Teleop- eration with immersive active visual feedback. arXiv preprint arXiv:2407.01512, 2024

    work page arXiv 2024

  15. [15]

    Expres- sive whole-body control for humanoid robots

    Xuxin Cheng, Yandong Ji, Junming Chen, Ruihan Yang, Ge Yang, and Xiaolong Wang. Expres- sive whole-body control for humanoid robots. arXiv preprint arXiv:2402.16796, 2024

    work page arXiv 2024

  16. [16]

    Humanplus: Humanoid shadowing and imitation from humans

    Zipeng Fu, Qingqing Zhao, Qi Wu, Gordon Wetzstein, and Chelsea Finn. Humanplus: Hu- manoid shadowing and imitation from humans. arXiv preprint arXiv:2406.10454, 2024

    work page arXiv 2024

  17. [17]

    Hover: Versatile neural whole-body controller for humanoid robots,

    Tairan He, Wenli Xiao, Toru Lin, Zhengyi Luo, Zhenjia Xu, Zhenyu Jiang, Changliu Liu, Guanya Shi, Xiaolong Wang, Linxi Fan, and Yuke Zhu. Hover: Versatile neural whole-body controller for humanoid robots. arXiv preprint arXiv:2410.21229, 2024

    work page arXiv 2024

  18. [18]

    A unified and general humanoid whole-body controller for fine-grained locomotion

    Yufei Xue, Wentao Dong, Minghuan Liu, Weinan Zhang, and Jiangmiao Pang. A unified and general humanoid whole-body controller for fine-grained locomotion. In Robotics: Science and Systems (RSS), 2025

    work page 2025

  19. [19]

    Flam: Foundation model-based body stabilization for humanoid locomotion and manipulation

    Xianqi Zhang, Hongliang Wei, Wenrui Wang, Xingtao Wang, Xiaopeng Fan, and Debin Zhao. Flam: Foundation model-based body stabilization for humanoid locomotion and manipulation. arXiv preprint arXiv:2503.22249, 2025

    work page arXiv 2025

  20. [20]

    Nil: No-data imitation learning by leveraging pre-trained video diffusion models

    Mert Albaba, Chenhao Li, Markos Diomataris, Omid Taheri, Andreas Krause, and Michael Black. Nil: No-data imitation learning by leveraging pre-trained video diffusion models. arXiv preprint arXiv:2503.10626, 2025

    work page arXiv 2025

  21. [21]

    Whole-body humanoid robot locomotion with human reference

    Qiang Zhang, Peter Cui, David Yan, Jingkai Sun, Yiqun Duan, Gang Han, Wen Zhao, Weining Zhang, Yijie Guo, Arthur Zhang, et al. Whole-body humanoid robot locomotion with human reference. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 11225–11231. IEEE, 2024

    work page 2024

  22. [22]

    Learning human-to-humanoid real-time whole-body teleoperation

    Tairan He, Zhengyi Luo, Wenli Xiao, Chong Zhang, Kris Kitani, Changliu Liu, and Guanya Shi. Learning human-to-humanoid real-time whole-body teleoperation. In 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 8944–8951. IEEE, 2024

    work page 2024

  23. [23]

    Harmon: Whole- body motion generation of humanoid robots from language descriptions

    Zhenyu Jiang, Yuqi Xie, Jinhan Li, Ye Yuan, Yifeng Zhu, and Yuke Zhu. Harmon: Whole- body motion generation of humanoid robots from language descriptions. arXiv preprint arXiv:2410.12773, 2024

    work page arXiv 2024

  24. [24]

    Dynamic locomotion on slippery ground

    Fabian Jenelten, Jemin Hwangbo, Fabian Tresoldi, C Dario Bellicoso, and Marco Hutter. Dynamic locomotion on slippery ground. IEEE Robotics and Automation Letters, 4(4):4170– 4176, 2019. 11

    work page 2019

  25. [25]

    BeamDojo: Learning agile humanoid locomotion on sparse footholds

    Huayi Wang, Zirui Wang, Junli Ren, Qingwei Ben, Tao Huang, Weinan Zhang, and Jiangmiao Pang. BeamDojo: Learning agile humanoid locomotion on sparse footholds. In Robotics: Science and Systems (RSS), 2025

    work page 2025

  26. [26]

    Vb-com: Learning vision-blind composite humanoid locomotion against deficient perception

    Junli Ren, Tao Huang, Huayi Wang, Zirui Wang, Qingwei Ben, Jiangmiao Pang, and Ping Luo. Vb-com: Learning vision-blind composite humanoid locomotion against deficient perception. arXiv preprint arXiv:2502.14814, 2025

    work page arXiv 2025

  27. [27]

    Advancing humanoid locomotion: Mastering challenging terrains with denoising world model learning

    Xinyang Gu, Yen-Jen Wang, Xiang Zhu, Chengming Shi, Yanjiang Guo, Yichen Liu, and Jianyu Chen. Advancing humanoid locomotion: Mastering challenging terrains with denoising world model learning. arXiv preprint arXiv:2408.14472, 2024

    work page arXiv 2024

  28. [28]

    Learning humanoid locomotion with perceptive internal model, 2024

    Junfeng Long, Junli Ren, Moji Shi, Zirui Wang, Tao Huang, Ping Luo, and Jiangmiao Pang. Learning humanoid locomotion with perceptive internal model, 2024. URL https://arxiv. org/abs/2411.14386

    work page arXiv 2024

  29. [29]

    Distillation-ppo: A novel two-stage reinforcement learning framework for humanoid robot perceptive locomotion

    Qiang Zhang, Gang Han, Jingkai Sun, Wen Zhao, Chenghao Sun, Jiahang Cao, Jiaxu Wang, Yijie Guo, and Renjing Xu. Distillation-ppo: A novel two-stage reinforcement learning framework for humanoid robot perceptive locomotion. arXiv preprint arXiv:2503.08299, 2025

    work page arXiv 2025

  30. [30]

    Learning perceptive humanoid locomotion over challenging terrain

    Wandong Sun, Baoshi Cao, Long Chen, Yongbo Su, Yang Liu, Zongwu Xie, and Hong Liu. Learning perceptive humanoid locomotion over challenging terrain. arXiv preprint arXiv:2503.00692, 2025

    work page arXiv 2025

  31. [31]

    Teacher motion priors: Enhancing robot locomotion over challenging terrain

    Fangcheng Jin, Yuqi Wang, Peixin Ma, Guodong Yang, Pan Zhao, En Li, and Zhengtao Zhang. Teacher motion priors: Enhancing robot locomotion over challenging terrain. arXiv preprint arXiv:2504.10390, 2025

    work page arXiv 2025

  32. [32]

    Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoperation and learning

    Tairan He, Zhengyi Luo, Xialin He, Wenli Xiao, Chong Zhang, Weinan Zhang, Kris Kitani, Changliu Liu, and Guanya Shi. Omnih2o: Universal and dexterous human-to-humanoid whole-body teleoperation and learning. arXiv preprint arXiv:2406.08858, 2024

    work page arXiv 2024

  33. [33]

    Amass: Archive of motion capture as surface shapes

    Naureen Mahmood, Nima Ghorbani, Nikolaus F Troje, Gerard Pons-Moll, and Michael J Black. Amass: Archive of motion capture as surface shapes. In Proceedings of the IEEE/CVF international conference on computer vision, pages 5442–5451, 2019

    work page 2019

  34. [34]

    Deep reinforcement learning for bipedal locomotion: A brief survey

    Lingfan Bao, Joseph Humphreys, Tianhu Peng, and Chengxu Zhou. Deep reinforcement learning for bipedal locomotion: A brief survey. arXiv preprint arXiv:2404.17070, 2024

    work page arXiv 2024

  35. [35]

    Scaling cross- embodied learning: One policy for manipulation, navigation, locomotion and aviation

    Ria Doshi, Homer Walke, Oier Mees, Sudeep Dasari, and Sergey Levine. Scaling cross- embodied learning: One policy for manipulation, navigation, locomotion and aviation. arXiv preprint arXiv:2408.11812, 2024

    work page arXiv 2024

  36. [36]

    Skill transfer in deep reinforcement learning under morpho- logical heterogeneity

    Yang Hu and Giovanni Montana. Skill transfer in deep reinforcement learning under morpho- logical heterogeneity. arXiv preprint arXiv:1908.05265, 2019

    work page arXiv 1908

  37. [37]

    Transfer deep reinforcement learning in 3d environments: An empirical study

    Devendra Singh Chaplot, Guillaume Lample, Kanthashree Mysore Sathyendra, and Ruslan Salakhutdinov. Transfer deep reinforcement learning in 3d environments: An empirical study. In NIPS deep reinforcemente leaning workshop, volume 138, 2016

    work page 2016

  38. [38]

    One policy to control them all: Shared modular policies for agent-agnostic control

    Wenlong Huang, Igor Mordatch, and Deepak Pathak. One policy to control them all: Shared modular policies for agent-agnostic control. In International Conference on Machine Learning, pages 4455–4464. PMLR, 2020

    work page 2020

  39. [39]

    Fay, Henrik I

    Bo Ai, Liu Dai, Nico Bohlinger, Dichen Li, Tongzhou Mu, Zhanxin Wu, K. Fay, Henrik I. Christensen, Jan Peters, and Hao Su. Towards embodiment scaling laws in robot locomotion,

    work page

  40. [40]

    URL https://arxiv.org/abs/2505.05753

    work page arXiv

  41. [41]

    One policy to run them all: an end-to-end learning approach to multi-embodiment locomotion

    Nico Bohlinger, Grzegorz Czechmanowski, Maciej Krupka, Piotr Kicki, Krzysztof Walas, Jan Peters, and Davide Tateo. One policy to run them all: an end-to-end learning approach to multi-embodiment locomotion. arXiv preprint arXiv:2409.06366, 2024

    work page arXiv 2024

  42. [42]

    Get-zero: Graph embodiment transformer for zero-shot embodi- ment generalization

    Austin Patel and Shuran Song. Get-zero: Graph embodiment transformer for zero-shot embodi- ment generalization. arXiv preprint arXiv:2407.15002, 2024. 12

    work page arXiv 2024

  43. [43]

    Genloco: Generalized locomotion controllers for quadrupedal robots

    Gilbert Feng, Hongbo Zhang, Zhongyu Li, Xue Bin Peng, Bhuvan Basireddy, Linzhu Yue, Zhitao Song, Lizhi Yang, Yunhui Liu, Koushil Sreenath, et al. Genloco: Generalized locomotion controllers for quadrupedal robots. In Conference on Robot Learning, pages 1893–1903. PMLR, 2023

    work page 1903

  44. [44]

    Octo: An Open-Source Generalist Robot Policy

    Octo Model Team, Dibya Ghosh, Homer Walke, Karl Pertsch, Kevin Black, Oier Mees, Sudeep Dasari, Joey Hejna, Tobias Kreiman, Charles Xu, et al. Octo: An open-source generalist robot policy. arXiv preprint arXiv:2405.12213, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  45. [45]

    GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

    Johan Bjorck, Fernando Castañeda, Nikita Cherniadev, Xingye Da, Runyu Ding, Linxi Fan, Yu Fang, Dieter Fox, Fengyuan Hu, Spencer Huang, et al. Gr00t n1: An open foundation model for generalist humanoid robots. arXiv preprint arXiv:2503.14734, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  46. [46]

    RT-1: Robotics Transformer for Real-World Control at Scale

    Anthony Brohan, Noah Brown, Justice Carbajal, Yevgen Chebotar, Joseph Dabis, Chelsea Finn, Keerthana Gopalakrishnan, Karol Hausman, Alex Herzog, Jasmine Hsu, et al. Rt-1: Robotics transformer for real-world control at scale. arXiv preprint arXiv:2212.06817, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  47. [47]

    RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

    Anthony Brohan, Noah Brown, Justice Carbajal, Yevgen Chebotar, Xi Chen, Krzysztof Choro- manski, Tianli Ding, Danny Driess, Avinava Dubey, Chelsea Finn, et al. Rt-2: Vision-language- action models transfer web knowledge to robotic control. arXiv preprint arXiv:2307.15818, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  48. [48]

    Score-Based Generative Modeling through Stochastic Differential Equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  49. [49]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

    work page 2020

  50. [50]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021

    work page 2021

  51. [51]

    Video diffusion models

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. Video diffusion models. Advances in Neural Information Processing Systems, 35:8633–8646, 2022

    work page 2022

  52. [52]

    Align your latents: High-resolution video synthesis with latent diffusion models

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dockhorn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. Align your latents: High-resolution video synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 22563–22575, 2023

    work page 2023

  53. [53]

    Diffusion policy: Visuomotor policy learning via action diffusion

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, page 02783649241273668, 2023

    work page 2023

  54. [54]

    3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations

    Yanjie Ze, Gu Zhang, Kangning Zhang, Chenyuan Hu, Muhan Wang, and Huazhe Xu. 3d diffusion policy: Generalizable visuomotor policy learning via simple 3d representations. In Proceedings of Robotics: Science and Systems (RSS), 2024

    work page 2024

  55. [55]

    Planning with diffusion for flexible behavior synthesis

    Michael Janner, Yilun Du, Joshua Tenenbaum, and Sergey Levine. Planning with diffusion for flexible behavior synthesis. In International Conference on Machine Learning , pages 9902–9915. PMLR, 2022

    work page 2022

  56. [56]

    AffordDP: Generalizable diffusion policy with transferable affordance

    Shijie Wu, Yihang Zhu, Yunao Huang, Kaizhen Zhu, Jiayuan Gu, Jingyi Yu, Ye Shi, and Jingya Wang. AffordDP: Generalizable diffusion policy with transferable affordance. arXiv preprint arXiv:2412.03142, 2024

    work page arXiv 2024

  57. [57]

    Diffusion policies as an expressive policy class for offline reinforcement learning

    Zhendong Wang, Jonathan J Hunt, and Mingyuan Zhou. Diffusion policies as an expressive policy class for offline reinforcement learning. In The Eleventh International Conference on Learning Representations, 2023

    work page 2023

  58. [58]

    Efficient diffusion policies for offline reinforcement learning

    Bingyi Kang, Xiao Ma, Chao Du, Tianyu Pang, and Shuicheng Yan. Efficient diffusion policies for offline reinforcement learning. Advances in Neural Information Processing Systems, 36: 67195–67212, 2023. 13

    work page 2023

  59. [59]

    Diffusion-dice: In-sample diffusion guidance for offline reinforcement learning

    Liyuan Mao, Haoran Xu, Xianyuan Zhan, Weinan Zhang, and Amy Zhang. Diffusion-dice: In-sample diffusion guidance for offline reinforcement learning. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    work page 2024

  60. [60]

    Policy representation via diffusion probability model for reinforcement learning

    Long Yang, Zhixiong Huang, Fenghao Lei, Yucun Zhong, Yiming Yang, Cong Fang, Shiting Wen, Binbin Zhou, and Zhouchen Lin. Policy representation via diffusion probability model for reinforcement learning. arXiv preprint arXiv:2305.13122, 2023

    work page arXiv 2023

  61. [61]

    Learning a diffusion model policy from rewards via q-score matching

    Michael Psenka, Alejandro Escontrela, Pieter Abbeel, and Yi Ma. Learning a diffusion model policy from rewards via q-score matching. In International Conference on Machine Learning, pages 41163–41182. PMLR, 2024

    work page 2024

  62. [62]

    Diffusion-based reinforcement learning via q-weighted variational policy optimization

    Shutong Ding, Ke Hu, Zhenhao Zhang, Kan Ren, Weinan Zhang, Jingyi Yu, Jingya Wang, and Ye Shi. Diffusion-based reinforcement learning via q-weighted variational policy optimization. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024

    work page 2024

  63. [63]

    Diffusion-vla: Scaling robot foundation models via unified diffusion and autoregression

    Junjie Wen, Minjie Zhu, Yichen Zhu, Zhibin Tang, Jinming Li, Zhongyi Zhou, Chengmeng Li, Xiaoyu Liu, Yaxin Peng, Chaomin Shen, et al. Diffusion-vla: Scaling robot foundation models via unified diffusion and autoregression. arXiv preprint arXiv:2412.03293, 2024

    work page arXiv 2024

  64. [64]

    Boosting continuous control with consistency policy

    Yuhui Chen, Haoran Li, and Dongbin Zhao. Boosting continuous control with consistency policy. In Proceedings of the 23rd International Conference on Autonomous Agents and Multiagent Systems, pages 335–344, 2024

    work page 2024

  65. [65]

    HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model

    Jiaming Liu, Hao Chen, Pengju An, Zhuoyang Liu, Renrui Zhang, Chenyang Gu, Xiaoqi Li, Ziyu Guo, Sixiang Chen, Mengzhen Liu, et al. Hybridvla: Collaborative diffusion and autoregression in a unified vision-language-action model. arXiv preprint arXiv:2503.10631, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  66. [66]

    CLoSD: Closing the loop between simulation and diffusion for multi-task character control

    Guy Tevet, Sigal Raab, Setareh Cohan, Daniele Reda, Zhengyi Luo, Xue Bin Peng, Amit H Bermano, and Michiel van de Panne. CLoSD: Closing the loop between simulation and diffusion for multi-task character control. arXiv preprint arXiv:2410.03441, 2024

    work page arXiv 2024

  67. [67]

    Dartcontrol: A diffusion-based autoregressive motion model for real-time text-driven motion control

    Kaifeng Zhao, Gen Li, and Siyu Tang. Dartcontrol: A diffusion-based autoregressive motion model for real-time text-driven motion control. In The Thirteenth International Conference on Learning Representations, 2024

    work page 2024

  68. [68]

    Motion planning diffusion: Learning and planning of robot motions with diffusion models

    Joao Carvalho, An T Le, Mark Baierl, Dorothea Koert, and Jan Peters. Motion planning diffusion: Learning and planning of robot motions with diffusion models. In 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 1916–1923. IEEE, 2023

    work page 2023

  69. [69]

    Dipper: Diffusion-based 2d path planner applied on legged robots

    Jianwei Liu, Maria Stamatopoulou, and Dimitrios Kanoulas. Dipper: Diffusion-based 2d path planner applied on legged robots. In 2024 IEEE International Conference on Robotics and Automation (ICRA), pages 9264–9270. IEEE, 2024

    work page 2024

  70. [70]

    DiffuseLoco: Real-time legged locomotion control with diffusion from offline datasets

    Xiaoyu Huang, Yufeng Chi, Ruofeng Wang, Zhongyu Li, Xue Bin Peng, Sophia Shao, Borivoje Nikolic, and Koushil Sreenath. DiffuseLoco: Real-time legged locomotion control with diffusion from offline datasets. In 8th Annual Conference on Robot Learning, 2024

    work page 2024

  71. [71]

    Birodiff: Diffusion policies for bipedal robot locomotion on unseen terrains

    GVS Mothish, Manan Tayal, and Shishir Kolathaya. Birodiff: Diffusion policies for bipedal robot locomotion on unseen terrains. arXiv preprint arXiv:2407.05424, 2024

    work page arXiv 2024

  72. [72]

    Discovery of skill switching criteria for learning agile quadruped locomotion

    Wanming Yu, Fernando Acero, Vassil Atanassov, Chuanyu Yang, Ioannis Havoutis, Dimitrios Kanoulas, and Zhibin Li. Discovery of skill switching criteria for learning agile quadruped locomotion. arXiv preprint arXiv:2502.06676, 2025

    work page arXiv 2025

  73. [73]

    Preference aligned diffusion planner for quadrupedal locomotion control

    Xinyi Yuan, Zhiwei Shang, Zifan Wang, Chenkai Wang, Zhao Shan, Meixin Zhu, Chenjia Bai, Xuelong Li, Weiwei Wan, and Kensuke Harada. Preference aligned diffusion planner for quadrupedal locomotion control. arXiv preprint arXiv:2410.13586, 2024

    work page arXiv 2024

  74. [74]

    Isaac Gym: High Performance GPU-Based Physics Simulation For Robot Learning

    Viktor Makoviychuk, Lukasz Wawrzyniak, Yunrong Guo, Michelle Lu, Kier Storey, Miles Macklin, David Hoeller, Nikita Rudin, Arthur Allshire, Ankur Handa, et al. Isaac gym: High performance gpu-based physics simulation for robot learning. arXiv preprint arXiv:2108.10470, 2021. 14

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  75. [75]

    Proximal Policy Optimization Algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

    work page internal anchor Pith review Pith/arXiv arXiv 2017

  76. [76]

    Amp: Adversarial motion priors for stylized physics-based character control

    Xue Bin Peng, Ze Ma, Pieter Abbeel, Sergey Levine, and Angjoo Kanazawa. Amp: Adversarial motion priors for stylized physics-based character control. ACM Transactions on Graphics (ToG), 40(4):1–20, 2021

    work page 2021

  77. [77]

    Is Conditional Generative Modeling all you need for Decision-Making?

    Anurag Ajay, Yilun Du, Abhi Gupta, Joshua Tenenbaum, Tommi Jaakkola, and Pulkit Agrawal. Is conditional generative modeling all you need for decision-making? arXiv preprint arXiv:2211.15657, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  78. [78]

    Classifier-free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance. In NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications, 2021

    work page 2021

  79. [79]

    Scheduled sampling for sequence prediction with recurrent neural networks

    Samy Bengio, Oriol Vinyals, Navdeep Jaitly, and Noam Shazeer. Scheduled sampling for sequence prediction with recurrent neural networks. Advances in neural information processing systems, 28, 2015

    work page 2015

  80. [80]

    Learning physically simulated tennis skills from broadcast videos

    Ye Yuan, Viktor Makoviychuk, Y Guo, S Fidler, XB Peng, and K Fatahalian. Learning physically simulated tennis skills from broadcast videos. ACM Trans. Graph, 42(4), 2023

    work page 2023

Showing first 80 references.