pith. machine review for the scientific record. sign in

arxiv: 2604.07084 · v1 · submitted 2026-04-08 · 💻 cs.RO · cs.AI

Recognition: no theorem link

Flow Motion Policy: Manipulator Motion Planning with Flow Matching Models

Davood Soleymanzadeh, Minghui Zheng, Xiao Liang

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:26 UTC · model grok-4.3

classification 💻 cs.RO cs.AI
keywords motion planningflow matchingrobotic manipulatorsend-to-end planningneural policiesgenerative modelsbest-of-N samplingcollision avoidance
0
0 comments X

The pith

Flow matching models a distribution over feasible manipulator paths to support best-of-N sampling that selects the first collision-free solution from sensor data.

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

The paper presents Flow Motion Policy to fix the single-path output problem in open-loop neural motion planners for robot arms. It trains a flow matching model on planning data to represent the range of valid trajectories instead of one fixed path. During use the system draws several candidate paths at once, checks them for collisions after they are generated, and executes the first safe one. A reader would care because this keeps planning direct from observations and fast while raising the odds of success through simple sampling at inference time.

Core claim

Flow Motion Policy is an open-loop end-to-end neural motion planner that uses the stochastic formulation of flow matching to capture the multi-modality of planning datasets, thereby modeling a distribution over feasible paths; this enables efficient inference-time best-of-N sampling in which multiple candidate paths are generated, their collision status is evaluated after planning, and the first collision-free solution is executed.

What carries the argument

Flow matching model that learns the conditional distribution of feasible manipulator trajectories from sensor observations and supports fast stochastic sampling of multiple diverse candidates for post-generation collision checks.

If this is right

  • Multiple candidate paths can be produced from the same observation and filtered by a simple post-planning collision check.
  • Planning success improves compared with single-output neural planners and some sampling-based baselines.
  • The planner remains fully open-loop and does not require a privileged collision checker while generating paths.
  • Efficiency gains appear in benchmarks when the first valid sample is selected without further optimization.

Where Pith is reading between the lines

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

  • The same distribution-modeling idea could be tested on other robot types where path multi-modality is also high.
  • If sampling cost stays low, the method might reduce dependence on expensive collision checkers during real-time operation.
  • The approach points to a general pattern in which generative models replace deterministic predictors to gain inference-time robustness.

Load-bearing premise

A flow matching model trained on planning datasets will accurately capture the multi-modality of feasible paths so that best-of-N sampling finds a collision-free solution with only a modest number of samples.

What would settle it

An experiment in which raising the number of samples fails to increase the fraction of workspaces that receive a collision-free path or in which the required number of samples grows large enough to erase the reported efficiency gains.

Figures

Figures reproduced from arXiv: 2604.07084 by Davood Soleymanzadeh, Minghui Zheng, Xiao Liang.

Figure 1
Figure 1. Figure 1: Method Overview. Flow Motion Policy is a point cloud–conditioned flow-based motion policy that enables inference￾time optimization for motion planning in complex environments. The framework leverages the generative formulation of flow matching to capture the inherent multi-modality of motion planning datasets and supports efficient inference-time best-of-N sampling. Compared to neural informed samplers suc… view at source ↗
Figure 2
Figure 2. Figure 2: Flow Motion Policy architecture. This framework leverages current time-step t planning information (i.e., planning scene (Pw) point cloud, robot (Pr) point cloud, robotic manipulator configuration (qt)) together with robotic manipulator goal configuration (qgoal) for end-to-end open-loop motion planning in complex environments. ht, hr, and hw are the current time-step t planning information corresponding e… view at source ↗
Figure 3
Figure 3. Figure 3: Best-of-N Sampling: Planning success rate and planning time of Flow Motion Policy and Neural MP [23] adaptation across the held-out planning tasks. “GMM-1” and “GMM-100” are Neural MP [23] adaptation with Gaussian Mixture Model (GMM) without inference-time optimization (N = 1) and inference-time optimization (N = 100), respectively. “FMP-1” and “FMP-100” are Flow Motion Policy without inference-time optimi… view at source ↗
Figure 4
Figure 4. Figure 4: Policy Head Architecture Ablation: Average planning success rate and planning time of Flow Motion Policy and Diffusion Motion Policy with various head architectures across all held-out evaluation tasks. “DMP-1” and “DMP-100” are Diffusion Motion Policy without inference-time optimization (N = 1) and inference-time optimization (N = 100), re￾spectively. “FMP-1” and “FMP-100” are Flow Motion Policy without i… view at source ↗
Figure 5
Figure 5. Figure 5: Diffusion Timestep Ablation: Average planning suc￾cess rate and planning time of Diffusion Motion Policy with various head architectures, and diffusion timesteps (k) across all held-out evaluation tasks. “DMP-1” and “DMP-100” are Diffusion Motion Policy without inference-time optimization (N = 1) and inference-time optimization (N = 100), respec￾tively. “DiT” denotes DiT-Block Policy [35]. Ablation study: … view at source ↗
Figure 6
Figure 6. Figure 6: Diffusion Motion Policy inference Time Optimization. success rate and planning time (mean ± standard deviation) of Diffusion Motion Policy with various head architectures across a range of trajectories (N ∈ [10, 200]) in inference time optimization. “DiT” denotes DiT-Block Policy [35]. 50 70 90 Success Rate TableTop 60 66 72 Box 90 95 100 Bins 65 78 91 Shelf1 40 55 70 Shelf2 40 50 60 Shelf3 10 200 0 2 Time… view at source ↗
Figure 7
Figure 7. Figure 7: Flow Motion Policy inference Time Optimization. Success rate and planning time (mean ± standard deviation) of Flow Motion Policy with various head architectures across a range of trajectories (N ∈ [10, 200]) in inference time optimization. “DiT” denotes DiT-Block Policy [35]. Motion Policy with a diffusion-based policy head [34] while keeping the remaining components of the framework un￾changed. This diffu… view at source ↗
Figure 8
Figure 8. Figure 8: Inference Flow Steps: Success rate and average planning time of Flow Motion Policy with various head architectures across a range of inference flow steps (step ∈ [5, 90]) without inference time optimization (N = 1). “DiT” denotes DiT-Block Policy [35]. 70 80 90 Success Rate TableTop 64 67 70 Box 94 96 98 Bins 75 82 89 Shelf1 50 57 64 Shelf2 40 48 56 Shelf3 5 15 40 60 80 1 2 3 Time (s) 5 15 40 60 80 1 2 3 5… view at source ↗
Figure 9
Figure 9. Figure 9: Inference Flow Steps: Success rate and average planning time of Flow Motion Policy with various head architectures across a range of inference flow steps (step ∈ [5, 90]) with inference time optimization (N = 100). “DiT” denotes DiT-Block Policy [35]. time increases with the number of solver steps based on the network sizes provided in Table III. This observation suggests that the policy heads are able to … view at source ↗
Figure 10
Figure 10. Figure 10: AprilTag markers were used to calibrate the camera and estimate the rigid transformation between the camera frame and the robot base frame. TABLE IV: Flow Motion Policy Deployment in Real-world. Planning success rate of Flow Motion Policy performance in real-world evaluation tasks. “FMP-1” and “FMP-100” are Flow Motion Policy without inference-time optimization (N = 1) and with inference-time optimization… view at source ↗
Figure 11
Figure 11. Figure 11: Flow Motion Policy Deployment in Real-world. The path for the given motion planning problem is visualized for each environment, illustrating the planned motion execution across subsequent frames. all planning instances, whereas the base policy achieves only 50% success (5/10). Similarly, in articulated environments, the policy with inference-time optimization also achieves a 100% (10/10) success rate, whi… view at source ↗
read the original abstract

Open-loop end-to-end neural motion planners have recently been proposed to improve motion planning for robotic manipulators. These methods enable planning directly from sensor observations without relying on a privileged collision checker during planning. However, many existing methods generate only a single path for a given workspace across different runs, and do not leverage their open-loop structure for inference-time optimization. To address this limitation, we introduce Flow Motion Policy, an open-loop, end-to-end neural motion planner for robotic manipulators that leverages the stochastic generative formulation of flow matching methods to capture the inherent multi-modality of planning datasets. By modeling a distribution over feasible paths, Flow Motion Policy enables efficient inference-time best-of-$N$ sampling. The method generates multiple end-to-end candidate paths, evaluates their collision status after planning, and executes the first collision-free solution. We benchmark the Flow Motion Policy against representative sampling-based and neural motion planning methods. Evaluation results demonstrate that Flow Motion Policy improves planning success and efficiency, highlighting the effectiveness of stochastic generative policies for end-to-end motion planning and inference-time optimization. Experimental evaluation videos are available via this \href{https://zh.engr.tamu.edu/wp-content/uploads/sites/310/2026/03/FMP-Website.mp4}{link}.

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 paper proposes Flow Motion Policy, an open-loop end-to-end neural motion planner for robotic manipulators based on flow matching models. By learning a distribution over feasible paths from planning datasets, the method supports inference-time best-of-N sampling: multiple candidate paths are generated from the model, their collision status is checked after generation using an external verifier, and the first collision-free path is executed. The authors benchmark the approach against sampling-based planners and other neural methods, claiming improvements in planning success rate and efficiency due to the ability to capture multi-modality in the path distribution.

Significance. If the empirical claims hold, the work would be significant for end-to-end neural motion planning in robotics. It demonstrates a practical way to leverage the stochastic nature of generative models for inference-time optimization without requiring privileged collision information during path generation, potentially improving upon deterministic single-shot neural planners while retaining their open-loop advantages.

major comments (2)
  1. [Abstract] Abstract: The central claim of improved planning success and efficiency is asserted without any quantitative results, baseline comparisons, dataset descriptions, or ablation studies. This leaves the empirical contribution unevaluated and prevents assessment of whether the best-of-N strategy delivers the stated gains.
  2. [Evaluation] The method's effectiveness depends on the flow matching model placing non-negligible probability mass on multiple distinct feasible trajectories per observation. No supporting analysis is provided, such as the fraction of collision-free samples per scene, path diversity metrics, or success-rate curves as a function of N, which are required to substantiate that low-N sampling reliably yields valid solutions.
minor comments (1)
  1. [Abstract] The video link in the abstract should be verified for accessibility and permanence.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed review and valuable comments on our paper. We have carefully considered the feedback and made revisions to address the concerns raised regarding the abstract and evaluation sections. Below, we provide point-by-point responses to the major comments.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of improved planning success and efficiency is asserted without any quantitative results, baseline comparisons, dataset descriptions, or ablation studies. This leaves the empirical contribution unevaluated and prevents assessment of whether the best-of-N strategy delivers the stated gains.

    Authors: We acknowledge that the abstract, as a concise summary, did not include specific quantitative results. However, the full manuscript provides detailed benchmarks, baseline comparisons, and dataset descriptions in the Experiments section. To better highlight the contributions upfront, we have revised the abstract to incorporate key quantitative findings, such as the improvement in success rates and reduction in planning time achieved by Flow Motion Policy compared to baselines. revision: yes

  2. Referee: [Evaluation] The method's effectiveness depends on the flow matching model placing non-negligible probability mass on multiple distinct feasible trajectories per observation. No supporting analysis is provided, such as the fraction of collision-free samples per scene, path diversity metrics, or success-rate curves as a function of N, which are required to substantiate that low-N sampling reliably yields valid solutions.

    Authors: We agree that additional analysis on the multi-modality captured by the model would strengthen the paper. While our experiments demonstrate the benefits of best-of-N sampling through overall performance metrics, we have added new supporting analyses in the revised manuscript. These include statistics on the fraction of collision-free samples generated per scene, quantitative path diversity metrics (e.g., mean pairwise Hausdorff distance between samples), and plots of success rate versus N to show how low-N sampling achieves high reliability. This directly substantiates the effectiveness of the stochastic generative approach. revision: yes

Circularity Check

0 steps flagged

No significant circularity; direct application of flow matching with external post-processing

full rationale

The paper applies existing flow matching to learn a distribution over paths from planning datasets and performs standard best-of-N sampling followed by an independent collision check. No equations, definitions, or claims reduce the claimed gains to fitted parameters renamed as predictions, self-citations that bear the central load, or ansatzes smuggled from prior author work. The multi-modality assumption is an empirical modeling claim evaluated via benchmarks rather than a self-referential construction. The derivation chain is self-contained against external baselines and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim depends on the flow matching model learning a useful distribution over paths and on the post-generation collision check being both reliable and cheap; no new physical entities are introduced.

free parameters (1)
  • best-of-N sample count
    Inference-time hyperparameter controlling the number of paths drawn before collision filtering; trades compute for success probability.
axioms (1)
  • domain assumption Flow matching models trained on planning datasets can capture the inherent multi-modality of feasible manipulator paths
    Invoked to justify why best-of-N sampling improves over single-path generation.

pith-pipeline@v0.9.0 · 5521 in / 1336 out tokens · 62256 ms · 2026-05-10T18:26:43.277994+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

45 extracted references · 10 canonical work pages · 3 internal anchors

  1. [1]

    Motion planning networks: Bridging the gap between learning-based and classical motion planners,

    A. H. Qureshi, Y . Miao, A. Simeonov, and M. C. Yip, “Motion planning networks: Bridging the gap between learning-based and classical motion planners,”IEEE Transactions on Robotics, vol. 37, no. 1, pp. 48–66, 2020

    2020

  2. [2]

    Perfact: Motion policy with llm-powered dataset synthesis and fusion action-chunking trans- formers,

    D. Soleymanzadeh, X. Liang, and M. Zheng, “Perfact: Motion policy with llm-powered dataset synthesis and fusion action-chunking trans- formers,”arXiv preprint arXiv:2512.03444, 2025

    work page arXiv 2025

  3. [3]

    Conventional, heuristic and learning-based robot motion planning: Reviewing frameworks of current practical significance,

    F. Noroozi, M. Daneshmand, and P. Fiorini, “Conventional, heuristic and learning-based robot motion planning: Reviewing frameworks of current practical significance,”Machines, vol. 11, no. 7, p. 722, 2023

    2023

  4. [4]

    Toward generalist neural motion planners for robotic manipulators: Challenges and opportunities,

    D. Soleymanzadeh, I. Lopez-Sanchez, H. Su, Y . Li, X. Liang, and M. Zheng, “Toward generalist neural motion planners for robotic manipulators: Challenges and opportunities,”IEEE Transactions on Automation Science and Engineering, 2026

    2026

  5. [5]

    Rapidly-exploring random trees: A new tool for path planning,

    S. LaValle, “Rapidly-exploring random trees: A new tool for path planning,”Research Report 9811, 1998

    1998

  6. [6]

    Chomp: Covariant hamiltonian optimization for motion planning,

    M. Zucker, N. Ratliff, A. D. Dragan, M. Pivtoraiko, M. Klingensmith, C. M. Dellin, J. A. Bagnell, and S. S. Srinivasa, “Chomp: Covariant hamiltonian optimization for motion planning,”The International jour- nal of robotics research, vol. 32, no. 9-10, pp. 1164–1193, 2013

    2013

  7. [7]

    Rapidly-exploring random trees: Progress and prospects: Steven m. lavalle, iowa state university, a james j. kuffner, jr., university of tokyo, tokyo, japan,

    S. M. LaValle and J. J. Kuffner, “Rapidly-exploring random trees: Progress and prospects: Steven m. lavalle, iowa state university, a james j. kuffner, jr., university of tokyo, tokyo, japan,”Algorithmic and computational robotics, pp. 303–307, 2001

    2001

  8. [8]

    Sampling-based algorithms for optimal motion planning,

    S. Karaman and E. Frazzoli, “Sampling-based algorithms for optimal motion planning,”The international journal of robotics research, vol. 30, no. 7, pp. 846–894, 2011

    2011

  9. [9]

    Forward kinematics kernel for improved proxy collision checking,

    N. Das and M. C. Yip, “Forward kinematics kernel for improved proxy collision checking,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 2349–2356, 2020

    2020

  10. [10]

    pRRTC: GPU-Parallel RRT-Connect for Fast, Consistent, and Low-Cost Mo- tion Planning

    C. H. Huang, P. Jadhav, B. Plancher, and Z. Kingston, “prrtc: Gpu- parallel rrt-connect for fast, consistent, and low-cost motion planning,” arXiv preprint arXiv:2503.06757, 2025

    work page arXiv 2025

  11. [11]

    Motions in microsec- onds via vectorized sampling-based planning,

    W. Thomason, Z. Kingston, and L. E. Kavraki, “Motions in microsec- onds via vectorized sampling-based planning,” in2024 IEEE interna- tional conference on robotics and automation (ICRA). IEEE, 2024, pp. 8749–8756

    2024

  12. [12]

    Flexible informed trees (fit*): Adaptive batch-size approach in informed sampling-based path planning,

    L. Zhang, Z. Bing, K. Chen, L. Chen, K. Cai, Y . Zhang, F. Wu, P. Krumbholz, Z. Yuan, S. Haddadinet al., “Flexible informed trees (fit*): Adaptive batch-size approach in informed sampling-based path planning,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, pp. 3146–3152

    2024

  13. [13]

    Batch informed trees (bit*): Informed asymptotically optimal anytime search,

    J. D. Gammell, T. D. Barfoot, and S. S. Srinivasa, “Batch informed trees (bit*): Informed asymptotically optimal anytime search,”The International Journal of Robotics Research, vol. 39, no. 5, pp. 543– 567, 2020

    2020

  14. [14]

    Stomp: Stochastic trajectory optimization for motion planning,

    M. Kalakrishnan, S. Chitta, E. Theodorou, P. Pastor, and S. Schaal, “Stomp: Stochastic trajectory optimization for motion planning,” in IEEE international conference on robotics and automation, 2011, pp. 4569–4574

    2011

  15. [15]

    Curobo: Parallelized collision-free robot motion generation,

    B. Sundaralingam, S. K. S. Hari, A. Fishman, C. Garrett, K. Van Wyk, V . Blukis, A. Millane, H. Oleynikova, A. Handa, F. Ramoset al., “Curobo: Parallelized collision-free robot motion generation,” inIEEE International Conference on Robotics and Automation, 2023, pp. 8112– 8119

    2023

  16. [16]

    Simpnet: Spatial-informed motion planning network,

    D. Soleymanzadeh, X. Liang, and M. Zheng, “Simpnet: Spatial-informed motion planning network,”IEEE Robotics and Automation Letters, 2025

    2025

  17. [17]

    Learning sampling dictionaries for efficient and generalizable robot motion planning with transformers,

    J. J. Johnson, A. H. Qureshi, and M. C. Yip, “Learning sampling dictionaries for efficient and generalizable robot motion planning with transformers,”IEEE Robotics and Automation Letters, vol. 8, no. 12, pp. 7946–7953, 2023

    2023

  18. [18]

    Gaide: Graph-based attention masking for spatial-and embodiment-aware mo- tion planning,

    D. Soleymanzadeh, X. Liang, and M. Zheng, “Gaide: Graph-based attention masking for spatial-and embodiment-aware motion planning,” arXiv preprint arXiv:2603.04463, 2026

    work page arXiv 2026

  19. [19]

    Pairwisenet: Pairwise collision distance learning for high-dof robot systems,

    J. Kim and F. C. Park, “Pairwisenet: Pairwise collision distance learning for high-dof robot systems,” inConference on Robot Learning. PMLR, 2023, pp. 2863–2877

    2023

  20. [20]

    Graph-based 3d collision- distance estimation network with probabilistic graph rewiring,

    M. Song, Y . Kim, M. J. Kim, and D. Park, “Graph-based 3d collision- distance estimation network with probabilistic graph rewiring,” in IEEE International Conference on Robotics and Automation, 2024, pp. 10 939–10 945

    2024

  21. [21]

    Motion planning diffusion: Learning and adapting robot motion planning with diffusion models,

    J. Carvalho, A. T. Le, P. Kicki, D. Koert, and J. Peters, “Motion planning diffusion: Learning and adapting robot motion planning with diffusion models,”IEEE Transactions on Robotics, 2025

    2025

  22. [22]

    M 2 diffuser: Diffusion-based trajectory optimization for mobile manipulation in 3d scenes,

    S. Yan, Z. Zhang, M. Han, Z. Wang, Q. Xie, Z. Li, Z. Li, H. Liu, X. Wang, and S.-C. Zhu, “M 2 diffuser: Diffusion-based trajectory optimization for mobile manipulation in 3d scenes,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    2025

  23. [23]

    Neural mp: A generalist neural motion planner,

    M. Dalal, J. Yang, R. Mendonca, Y . Khaky, R. Salakhutdinov, and D. Pathak, “Neural mp: A generalist neural motion planner,”arXiv preprint arXiv:2409.05864, 2024

    work page arXiv 2024

  24. [24]

    Deep reactive policy: Learning reactive manipulator motion planning for dynamic environments,

    J. Yang, J. J. Liu, Y . Li, Y . Khaky, K. Shaw, and D. Pathak, “Deep reactive policy: Learning reactive manipulator motion planning for dynamic environments,”arXiv preprint arXiv:2509.06953, 2025

    work page arXiv 2025

  25. [25]

    Flow Matching for Generative Modeling

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le, “Flow matching for generative modeling,”arXiv preprint arXiv:2210.02747, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  26. [26]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space,

    C. R. Qi, L. Yi, H. Su, and L. J. Guibas, “Pointnet++: Deep hierarchical feature learning on point sets in a metric space,”Advances in neural information processing systems, vol. 30, 2017

    2017

  27. [27]

    Advanced bit*(abit*): Sampling- based planning with advanced graph-search techniques,

    M. P. Strub and J. D. Gammell, “Advanced bit*(abit*): Sampling- based planning with advanced graph-search techniques,” in2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2020, pp. 130–136

    2020

  28. [28]

    Neural manipulation planning on constraint manifolds,

    A. H. Qureshi, J. Dong, A. Choe, and M. C. Yip, “Neural manipulation planning on constraint manifolds,”IEEE Robotics and Automation Letters, vol. 5, no. 4, pp. 6089–6096, 2020

    2020

  29. [29]

    Continuous- time gaussian process motion planning via probabilistic inference,

    M. Mukadam, J. Dong, X. Yan, F. Dellaert, and B. Boots, “Continuous- time gaussian process motion planning via probabilistic inference,”The International Journal of Robotics Research, vol. 37, no. 11, pp. 1319– 1340, 2018

    2018

  30. [30]

    Non-euclidean motion planning with graphs of geodesically convex sets,

    T. Cohn, M. Petersen, M. Simchowitz, and R. Tedrake, “Non-euclidean motion planning with graphs of geodesically convex sets,”The Interna- tional Journal of Robotics Research, vol. 44, no. 10-11, pp. 1840–1862, 2025

    2025

  31. [31]

    Flowmp: Learning motion fields for robot planning with conditional flow matching,

    K. Nguyen, A. T. Le, T. Pham, M. Huber, J. Peters, and M. N. Vu, “Flowmp: Learning motion fields for robot planning with conditional flow matching,” in2025 IEEE/RSJ International Conference on Intelli- gent Robots and Systems (IROS). IEEE, 2025, pp. 11 291–11 297

    2025

  32. [32]

    Motion policy networks,

    A. Fishman, A. Murali, C. Eppner, B. Peele, B. Boots, and D. Fox, “Motion policy networks,” inconference on Robot Learning. PMLR, 2023, pp. 967–977

    2023

  33. [33]

    Avoid everything: Model-free collision avoidance with expert-guided fine-tuning,

    A. Fishman, A. Walsman, M. Bhardwaj, W. Yuan, B. Sundaralingam, B. Boots, and D. Fox, “Avoid everything: Model-free collision avoidance with expert-guided fine-tuning,” inCoRL Workshop on Safe and Robust Robot Learning for Operation in the Real World, 2024

    2024

  34. [34]

    Diffusion policy: Visuomotor policy learning via action diffusion,

    C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song, “Diffusion policy: Visuomotor policy learning via action diffusion,”The International Journal of Robotics Research, vol. 44, no. 10-11, pp. 1684–1704, 2025

    2025

  35. [35]

    The ingre- dients for robotic diffusion transformers,

    S. Dasari, O. Mees, S. Zhao, M. K. Srirama, and S. Levine, “The ingre- dients for robotic diffusion transformers,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 15 617–15 625

    2025

  36. [36]

    GR00T N1: An Open Foundation Model for Generalist Humanoid Robots

    J. Bjorck, F. Casta ˜neda, N. Cherniadev, X. Da, R. Ding, L. Fan, Y . Fang, D. Fox, F. Hu, S. Huanget al., “Gr00t n1: An open foundation model for generalist humanoid robots,”arXiv preprint arXiv:2503.14734, 2025

    work page internal anchor Pith review arXiv 2025

  37. [37]

    $\pi_0$: A Vision-Language-Action Flow Model for General Robot Control

    K. Black, N. Brown, D. Driess, A. Esmail, M. Equi, C. Finn, N. Fusai, L. Groom, K. Hausman, B. Ichteret al., “π 0: A vision-language-action flow model for general robot control,”arXiv preprint arXiv:2410.24164, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  38. [38]

    Diffusion model predictive control,

    G. Zhou, S. Swaminathan, R. V . Raju, J. S. Guntupalli, W. Lehrach, J. Ortiz, A. Dedieu, M. L ´azaro-Gredilla, and K. Murphy, “Diffusion model predictive control,”arXiv preprint arXiv:2410.05364, 2024

    work page arXiv 2024

  39. [39]

    Neural ordinary differential equations,

    R. T. Chen, Y . Rubanova, J. Bettencourt, and D. K. Duvenaud, “Neural ordinary differential equations,”Advances in neural information pro- cessing systems, vol. 31, 2018

    2018

  40. [40]

    Rrt-connect: An efficient approach to single-query path planning,

    J. J. Kuffner and S. M. LaValle, “Rrt-connect: An efficient approach to single-query path planning,” inProceedings 2000 ICRA. Millennium conference. IEEE international conference on robotics and automation. Symposia proceedings (Cat. No. 00CH37065), vol. 2. IEEE, 2000, pp. 995–1001

    2000

  41. [41]

    Automatic differentiation in pytorch,

    A. Paszke, S. Gross, S. Chintala, G. Chanan, E. Yang, Z. DeVito, Z. Lin, A. Desmaison, L. Antiga, and A. Lerer, “Automatic differentiation in pytorch,” 2017

    2017

  42. [42]

    The open motion planning library,

    I. A. Sucan, M. Moll, and L. E. Kavraki, “The open motion planning library,”IEEE Robotics & Automation Magazine, vol. 19, no. 4, pp. 72–82, 2012

    2012

  43. [43]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems, vol. 33, pp. 6840– 6851, 2020

    2020

  44. [44]

    Rieman- nian motion policies,

    N. D. Ratliff, J. Issac, D. Kappler, S. Birchfield, and D. Fox, “Rieman- nian motion policies,”arXiv preprint arXiv:1801.02854, 2018

    work page arXiv 2018

  45. [45]

    Geometric fabrics: Generalizing classical mechanics to capture the physics of behavior,

    K. Van Wyk, M. Xie, A. Li, M. A. Rana, B. Babich, B. Peele, Q. Wan, I. Akinola, B. Sundaralingam, D. Foxet al., “Geometric fabrics: Generalizing classical mechanics to capture the physics of behavior,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3202–3209, 2022

    2022