Pith. sign in

REVIEW 3 major objections 5 minor 29 references

PROBE: Proprioceptive Obstacle Detection and Estimation while Navigating in Clutter

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A quadruped can reconstruct a 2D map of box obstacles, including fully occluded ones, from joint torques, joint states, and body pose alone.

desk verdict Useful new system for contact-based obstacle mapping on a legged robot, but the SE(2) pose claim is only tested for axis-aligned boxes and the evaluation needs more rigor. read the letter →

arxiv 2505.11848 v1 pith:6I6K3P2H submitted 2025-05-17 cs.RO

Dhruv Metha Ramesh , Aravind Sivaramakrishnan , Shreesh Keskar , Kostas E. Bekris , Jingjin Yu , Abdeslam Boularias This is my paper · ORCID
classification cs.RO
keywords proprioceptiveperceptionobstaclereconstructionleggedrobotnavigationtransformercontactsensingmovableobstaclesocclusionscene
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

PROBE claims that a quadruped robot can reconstruct a partial 2D map of box-shaped obstacles — each box's center, dimensions, rotation, and whether it is fixed or can be pushed — using nothing but its own body as the sensor: a history of joint positions, joint velocities, applied torques, and body pose, with no camera or other exteroceptive input. The strongest part of the claim is occlusion handling: when the robot pushes a movable box into a static one, the network reports the hidden static box and its dimensions, something vision alone cannot do. To back this up, the paper trains a causal Transformer on 180,000 simulated exploration trajectories and evaluates it on a real quadruped, with final-contact rotated intersection-over-union values between roughly 0.33 and 0.50 in simulation and between 0.20 and 0.45 on the physical robot. If the claim stands, legged robots gain a mapping modality that works exactly where cameras fail: darkness, smoke, and scenes where the object of interest sits behind another object.

What carries the argument

The central object is the Obstacle Reconstruction Module (ORM): a causal Transformer encoder followed by a two-layer MLP decoder that reads a history of proprioception vectors — joint positions, joint velocities, commanded joint torques, and the robot's SE(2) pose — and outputs one parameterized box per obstacle, (I_static, x, y, θ, w, l). Two design decisions carry the argument. First, the contact window: ground-truth labels for each obstacle are masked except over the interval during which the robot or a box pushed by the robot is physically touching it, and the training loss is a weighted sum of binary cross-entropy for the contact and static flags plus mean squared error for pose and dimensions, so the network is supervised exactly on the information-rich phase of each interaction. Second, the data-generation stack: a PPO-trained high-level navigation policy commands a student-teacher locomotion policy, and the resulting 180,000 curated trajectories span contact modes from no contact to direct contact with a movable box that indirectly pushes a hidden static box, which is what lets the ORM learn nested occlusion.

What would settle it

Place a single movable box at a 45-degree angle to the robot's starting heading in the real corridor and let the robot navigate and push it: the ORM, trained only on zero-rotation boxes, either tracks the true angle with IoU near the reported movable-obstacle range or the reconstruction collapses, which settles whether the SE(2) claim generalizes or was carried by axis-aligned training. A second check replays one recorded real trajectory through the ORM and counts how often the hidden static box behind the pushed movable one is recovered, since that occlusion capability is the paper's strongest claim.

Watch

Extended reading notes

Core claim

The paper's central discovery is that a supervised Transformer can map a quadruped's proprioception history into an actionable scene description with no visual input. The Obstacle Reconstruction Module (ORM) outputs, for each obstacle, a tuple giving a static-or-movable flag, an SE(2) pose (a 2D position plus a rotation angle), and the box's width and length, and it is trained to make each prediction only during that obstacle's contact window — the interval in which the robot or a pushed neighboring box is mechanically pressing on it. At the end of the contact window the final estimate is scored with rotated intersection-over-union between the predicted and ground-truth box geometry, and the reported means land at 0.33-0.50 for simulated scenarios of one to three obstacles and at 0.20-0.45 on the real quadruped. The paper also demonstrates the nested-interaction case in which a static box fully hidden behind a movable one is reconstructed only after the movable box has been pushed against it, which is the capability that most distinguishes proprioceptive reconstruction from any camera-based map.

Load-bearing premise

The result currently rests on the assumption that every obstacle is an axis-aligned rectangle: the training environments fix all obstacle orientations to zero and no rotated box is ever evaluated, so the claimed full pose estimation is unverified for any other angle.

Editorial extensions

If this is right

  • A quadruped can keep a working 2D obstacle map in darkness, smoke, or rubble where cameras are useless, using only its own joint torques, joint states, and body pose.
  • Because a pushed movable box reveals obstacles hidden behind it, the robot can plan around what it has not yet seen directly, which is the information required for navigation among movable obstacles.
  • The ORM produces estimates online during navigation and refines them as contact accumulates, so a planner can act on each box as soon as its contact window ends rather than waiting for a post-run map.
  • Sim-to-real transfer is achieved without retraining: real-robot reconstructions follow the same accuracy trends as simulation, with the degradation attributed by the paper to mass, friction, and shape differences between simulated and physical boxes.

Reading between the lines

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

  • All training environments fix obstacle orientation to zero, so the SE(2) claim is only demonstrated for axis-aligned boxes; the immediate test is to rotate a box 15-45 degrees and check whether the network's rotation estimate holds, and if it collapses, orientation augmentation of the training set becomes necessary.
  • The contact-window design suggests reconstruction quality tracks the amount of sustained contact the navigation policy produces, so an exploration policy that slides along box faces rather than bumping corners should raise final IoU more than any network change.
  • The rectangular parameterization caps what the network can express; swapping the decoder for an occupancy grid or an implicit shape representation would carry the same proprioception pipeline to arbitrary obstacle shapes, at the cost of the compact interpretable box output.
  • The proprioceptive map is best viewed as a complement to vision rather than a replacement: a fused system could trust the ORM exactly when the camera is occluded or the scene is dark, and cross-modal disagreement could itself flag which obstacles are movable.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes PROBE, a Transformer-based obstacle reconstruction module that takes a history of proprioceptive signals (joint positions, velocities, torques, and robot pose) from a Unitree Go1 quadruped and predicts the presence, dimensions, pose, and static/movable status of rectangular obstacles in a 2D workspace, including obstacles occluded by movable obstacles. The method is trained in Isaac Gym on trajectories generated by learned navigation policies and evaluated in simulation (1000 trials per difficulty) and on a real robot (20 Easy, 20 Medium, 5 Hard trials). The paper reports IoU and absolute error metrics, and an ablation on input modalities.

Significance. If the claims hold, this is a novel contribution: using only proprioception to reconstruct a 2D obstacle map, including nested occluded obstacles, on a legged robot without vision. The paper provides a clear problem formulation, a scalable simulation pipeline, and real-robot validation. The architecture is straightforward, and the ablation gives evidence about which inputs matter. However, the evaluation has gaps—most notably the orientation issue and the lack of baselines—that limit the strength of the central claims.

major comments (3)
  1. [V, Table I] The abstract and introduction claim that PROBE predicts obstacle poses in SE(2), but Section V states that during training obstacles are placed in random SE(2) configurations with their orientations fixed to 0 with respect to the robot's initial orientation. Table I then omits the orientation error for static obstacles with the note that their orientation is fixed across benchmarks. This is an explicit acknowledgment that static obstacle orientation is never varied or evaluated. The reported theta errors for movable obstacles (0.198 to 0.214 in Table I) come only from dynamic reorientation during pushing, which does not exercise arbitrary initial orientations. The manuscript should either add experiments with rotated obstacles (both static and movable) and report static orientation error, or restrict the claims to axis-aligned obstacles. As written, the general SE(2) pose-estimation claim is not supported by the evidence.
  2. [VI.B, Table II] The real-robot evaluation of the Hard benchmark rests on only 5 trials (Table II), and no variance, confidence intervals, or per-trial results are reported for any of the simulation or real-robot results. The claim of real-world effectiveness under the Hard scenario is therefore fragile. Please report standard deviations or ranges, and ideally increase the number of Hard trials.
  3. [VI] No baseline or comparison method is presented. While the ablation in Fig. 6 studies input modalities and is useful, it is conducted only for Nmax=2 and the evaluation set size is not specified. Without an external baseline (e.g., a contact-window heuristic or a non-transformer sequence model), it is hard to judge whether the learned Transformer mapping contributes beyond the information available at the final contact time step. The authors should either provide such a baseline or explicitly justify why no existing method is directly comparable.
minor comments (5)
  1. [IV.E] The phrase 'provides an novel method' should read 'provides a novel method'.
  2. [IV.D] The notation for the output set, \hat{O}_t = \{\hat{O}^t_i\}_{i=N}^{i=1}, uses N for the number of obstacles while the surrounding text uses n. Please unify the notation.
  3. [V] The environment dimensions wenv and lenv appear in Fig. 2 but are not defined in the text; please define them explicitly.
  4. [VI.A] The sentence 'The reported reconstruction for the movable obstacle is surprisingly more accurate in the Medium and Hard scenarios' is speculative; please provide a more precise explanation or remove the word 'surprisingly'.
  5. [IV.C] The dataset curation section mentions pruning trajectories based on contact mode frequency, but the pruning rule and the resulting dataset composition are not described; please add details.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ORM is a supervised network trained on simulator ground-truth obstacle states and evaluated on held-out environments; the zero-orientation setup is a generalization limitation rather than a circular derivation.

full rationale

The paper's central claim is that a Transformer (ORM) can infer planar box obstacle parameters from proprioceptive histories. This is an empirical supervised-learning claim, not a derivation whose conclusion is already contained in its inputs. The ground-truth labels (static/movable flag, x, y, theta, width, length) are simulator states independent of the network's inputs (joint positions, joint velocities, joint torques, body pose); nothing in the input representation or the loss function algebraically forces the predicted obstacle parameters. Held-out evaluation in Tables I and II uses newly sampled environments and a real robot, so the reported IoU and absolute errors are not quantities used to fit any network parameter or constant. The only notable scope restriction is that static obstacles are always initialized with orientation 0 and static orientation error is omitted; however, this is an evaluation/generalization limitation, not circularity, because movable obstacles acquire nonzero orientation through pushing and their theta error is reported, and the SE(2) parameterization is not defined in terms of the predictions. The self-referential element that training trajectories are generated by the authors' own navigation policies is a standard data-collection pipeline; those policies provide only proprioceptive inputs, not the obstacle labels being predicted. Citations to prior work by co-authors (e.g., [7], [10]) are related-work background and are not load-bearing for the PROBE result. The paper's own stated limitations (box-shaped objects only, no other physical properties tested) further confirm that the claims are bounded empirical findings rather than conclusions entailed by the setup.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No invented physical entities; the system is a learned mapping. The main external commitments are the rectangle and axis-aligned world model, and the sufficiency of proprioceptive contact data.

free parameters (2)
  • loss weights alpha_1 to alpha_4
    The ORM loss in Section IV-D combines four terms with scalar weights alpha_1..alpha_4, hand-chosen and not reported; these control the balance between contact classification, static/movable classification, pose error, and shape error, and therefore shape the learned reconstruction.
  • nav policy reward weights
    The navigation policy's reward (Section IV-B) blends goal reaching, distance, time, wall collision, and heading penalties with unspecified weights; these affect which trajectories and contact modes appear in the ORM training set.
assumptions (3)
  • domain assumption Obstacles are planar rectangles with orientation fixed to 0 during training.
    Section V states obstacle orientations are fixed to 0 relative to the robot's initial orientation. The SE(2) representation and learned mapping assume this; the paper does not evaluate rotated rectangles.
  • domain assumption Proprioceptive signals during contact windows are sufficient to infer obstacle geometry, including occluded obstacles.
    Section IV-D defines contact windows as the only prediction segments, assuming the force and motion history within these windows carries all information needed for reconstruction; no identifiability or information-theoretic analysis is provided.
  • domain assumption Simulation to real transfer is achieved via domain randomization of physical parameters.
    Section V says domain randomization aims to reduce sim2real gap; the real-world results rely on this generalization without explicit sim-real alignment metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PROBE: Proprioceptive Obstacle Detection and Estimation while Navigating in Clutter." pith.science (2026). https://pith.science/paper/6I6K3P2H

@misc{pith2026250511848,
  author       = {Pith},
  title        = {Pith review of: PROBE: Proprioceptive Obstacle Detection and Estimation while Navigating in Clutter},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6I6K3P2H}},
  note         = {Machine review of arXiv:2505.11848}
}
read the original abstract

In critical applications, including search-and-rescue in degraded environments, blockages can be prevalent and prevent the effective deployment of certain sensing modalities, particularly vision, due to occlusion and the constrained range of view of onboard camera sensors. To enable robots to tackle these challenges, we propose a new approach, Proprioceptive Obstacle Detection and Estimation while navigating in clutter PROBE, which instead relies only on the robot's proprioception to infer the presence or absence of occluded rectangular obstacles while predicting their dimensions and poses in SE(2). The proposed approach is a Transformer neural network that receives as input a history of applied torques and sensed whole-body movements of the robot and returns a parameterized representation of the obstacles in the environment. The effectiveness of PROBE is evaluated on simulated environments in Isaac Gym and with a real Unitree Go1 quadruped robot.

Figures

Figures reproduced from arXiv: 2505.11848 by the authors.

Figure 1
Figure 1. The setup considered in PROBE involves a Go1 robot dog and obstacles that can potentially obstruct its path. Some planar obstacles, such as the long frontal box in the image, are movable, while others are fixed to the ground. A transformer network reconstructs the locations and sizes of the obstacles, including the occluded ones, from a history of proprioceptive data that the robot receives while exploring the scene… view at source ↗
Figure 2
Figure 2. (Left) Environment setup in reality (top) and simulation (bottom), best viewed in color. The robot’s workspace is bounded in a box of dimensions [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Given as input a sequence of the robot’s joint positions, velocities, applied torques, and poses (Top), the Obstacle Reconstruction Module (ORM, [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Examples of Easy (Left), Medium (Middle) and Hard (Right) simu [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: An example execution of PROBE with a real Unitree Go1, best viewed in color. (Top) Snapshots of the experiment at different timestamps. (Bottom) Obstacle reconstruction returned by PROBE. As the experiment progresses, the robot (green) comes into direct contact with th…
Figure 6
Figure 6. Figure 6: A: {q}, B: {q, q˙}, C: {q, q, τ ˙ }, D: {q, q, τ, ρ ˙ }, E: {τ, ρ}. The plots show an ablation study on the ORM. The ORM trained with different inputs affects the reconstruction performance measured with the IoU metric. Ablation Studies [PITH_FULL_IMAGE:figures/full_f…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 23 canonical work pages

  1. [1]

    Efficient optimization for autonomous robotic manipulation of natural objects,

    A. Boularias, J. A. Bagnell, and A. Stentz, “Efficient optimization for autonomous robotic manipulation of natural objects,” inProceedings of the Twenty-Eighth AAAI Conference on Artificial Intelligence, July 27 -31, 2014, Qu ´ebec City, Qu ´ebec, Canada., 2014, pp. 2520–2526. [Online]. Available: http://www.aaai.org/ocs/index.php/ AAAI/AAAI14/paper/view/8414

    work page 2014

  2. [2]

    Developing a Robust Disaster Response Robot: CHIMP and the Robotics Challenge,

    G. C. Haynes, D. Stager, A. Stentz, J. M. Vande Weghe, B. Zajac, H. Herman, A. Kelly, E. Meyhofer, D. Anderson, D. Bennington, J. Brindza, D. Butterworth, C. Dellin, M. George, J. Gonzalez-Mora, M. Jones, P. Kini, M. Laverne, N. Letwin, E. Perko, C. Pinkston, D. Rice, J. Scheifflee, K. Strabala, M. Waldbaum, and R. Warner, “Developing a Robust Disaster Re...

    work page 2017

  3. [3]

    What Happened at the DARPA Robotics Challenge Finals,

    C. Atkeson, B. P. Wisely Babu, N. Banerjee, D. Berenson, C. P. Bove, X. Cui, M. Dedonato, R. Du, S. Feng, P. Franklin, M. Gennert, J. P. Graff, P. He, A. Jaeger, J. Kim, K. Knoedler, L. Li, C. Liu, X. Long, and X. Xinjilefu, “What Happened at the DARPA Robotics Challenge Finals,”Springer Tracts in Advanced Robotics, pp. 667– 684, 04 2018

    work page 2018

  4. [4]

    Team IHMC’s lessons learned from the DARPA robotics challenge trials,

    M. Johnson, B. Shrewsbury, S. Bertrand, T. Wu, D. Duran, M. Floyd, P. Abeles, D. Stephen, N. Mertins, A. Lesman,et al., “Team IHMC’s lessons learned from the DARPA robotics challenge trials,”Journal of Field Robotics, vol. 32, no. 2, pp. 192–208, 2015

    work page 2015

  5. [5]

    Robust ladder-climbing with a humanoid robot with application to the darpa robotics chal- lenge,

    J. Luo, Y . Zhang, K. Hauser, H. A. Park, M. Paldhe, C. G. Lee, M. Grey, M. Stilman, J. H. Oh, J. Lee,et al., “Robust ladder-climbing with a humanoid robot with application to the darpa robotics chal- lenge,” inRobotics and Automation (ICRA), 2014 IEEE International Conference on. IEEE, 2014, pp. 2792–2798

    work page 2014

  6. [6]

    The DARPA Robotics Challenge,

    G. Pratt and J. Manzo, “The DARPA Robotics Challenge,”IEEE Robotics & Automation Magazine, vol. 20, no. 2, pp. 10–12, 2013

    work page 2013

  7. [7]

    Learning to slide unknown objects with differentiable physics simulations,

    C. Song and A. Boularias, “Learning to slide unknown objects with differentiable physics simulations,” inProceedings of Robotics: Science and Systems (RSS), Corvallis, Oregon, 2020

    work page 2020

  8. [8]

    Navigation among movable obstacles,

    M. Stilman, “Navigation among movable obstacles,” Ph.D. disserta- tion, Carnegie Mellon University, Pittsburgh, PA, October 2007

    work page 2007

Show all 29 references
  1. [9]

    Shape completion enabled robotic grasping,

    J. Varley, C. DeChant, A. Richardson, J. Ruales, and P. K. Allen, “Shape completion enabled robotic grasping,”2017 IEEE/RSJ Inter- national Conference on Intelligent Robots and Systems (IROS), pp. 2442–2447, 2016

    2017

  2. [10]

    Inferring 3d shapes of unknown rigid objects in clutter through inverse physics reasoning,

    C. Song and A. Boularias, “Inferring 3d shapes of unknown rigid objects in clutter through inverse physics reasoning,”IEEE Robotics and Automation Letters (RA-L), vol. 4, no. 2, 2019

    2019

  3. [11]

    Shapeformer: Transformer-based shape completion via sparse representation,

    X. Yan, L. Lin, N. J. Mitra, D. Lischinski, D. Cohen-Or, and H. Huang, “Shapeformer: Transformer-based shape completion via sparse representation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022

    2022

  4. [12]

    Force-based simultaneous mapping and object reconstruction for robotic manipulation,

    J. Bimbo, A. S. Morgan, and A. M. Dollar, “Force-based simultaneous mapping and object reconstruction for robotic manipulation,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 4749–4756, 2022

    2022

  5. [13]

    Contact sensing from force measurements,

    A. Bicchi, J. K. Salisbury, and D. L. Brock, “Contact sensing from force measurements,”The International Journal of Robotics Research, vol. 12, pp. 249 – 262, 1990

    1990

  6. [14]

    Whole-body force sensation by force sensor with end-effector of arbitrary shape,

    N. Kurita, S. Sakaino, and T. Tsuji, “Whole-body force sensation by force sensor with end-effector of arbitrary shape,” in2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2012, Vilamoura, Algarve, Portugal, October 7- 12, 2012. IEEE, 2012, pp. 542...

    2012

  7. [15]

    Finger contact sensing and the application in dexterous hand manipulation,

    H. Liu, K. Nguyen, V . Perdereau, J. Bimbo, J. Back, M. Godden, L. D. Seneviratne, and K. Althoefer, “Finger contact sensing and the application in dexterous hand manipulation,”Auton. Robots, vol. 39, no. 1, pp. 25–41, 2015. [Online]. Available: https://doi.org/10.1007/s10514-...

    2015 doi

  8. [16]

    Collision detection and safe reaction with the dlr-iii lightweight ma- nipulator arm,

    A. D. Luca, A. O. Albu-Sch ¨affer, S. Haddadin, and G. Hirzinger, “Collision detection and safe reaction with the dlr-iii lightweight ma- nipulator arm,”2006 IEEE/RSJ International Conference on Intelligent Robots and Systems, pp. 1623–1630, 2006

    2006

  9. [17]

    Simultaneous localization and mapping: part i,

    H. Durrant-Whyte and T. Bailey, “Simultaneous localization and mapping: part i,”IEEE robotics & automation magazine, vol. 13, no. 2, pp. 99–110, 2006

    2006

  10. [18]

    Simultaneous localization and mapping,

    C. Stachniss, J. J. Leonard, and S. Thrun, “Simultaneous localization and mapping,”Springer Handbook of Robotics, pp. 1153–1176, 2016

    2016

  11. [19]

    Visual simultaneous localization and mapping: a survey,

    J. Fuentes-Pacheco, J. Ruiz-Ascencio, and J. M. Rend ´on-Mancha, “Visual simultaneous localization and mapping: a survey,”Artificial intelligence review, vol. 43, pp. 55–81, 2015

    2015

  12. [20]

    The blindfolded robot: A bayesian approach to planning with contact feedback,

    B. Saund, S. Choudhury, S. S. Srinivasa, and D. Berenson, “The blindfolded robot: A bayesian approach to planning with contact feedback,” inInternational Symposium of Robotics Research, 2019

    2019

  13. [21]

    Exploiting dis- tributed tactile sensors to drive a robot arm through obstacles,

    A. Albini, F. Grella, P. Maiolino, and G. Cannata, “Exploiting dis- tributed tactile sensors to drive a robot arm through obstacles,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 4361–4368, 2021

    2021

  14. [22]

    A stretchable tactile sleeve for reaching into cluttered spaces,

    A. M. Gruebele, M. A. Lin, D. Brouwer, S. Yuan, A. C. Zerbe, and M. R. Cutkosky, “A stretchable tactile sleeve for reaching into cluttered spaces,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 5308–5315, 2021

    2021

  15. [23]

    Clusternav: Learning-based robust navigation operating in cluttered environments,

    G. S. Martins, R. P. Rocha, F. J. Pais, and P. Menezes, “Clusternav: Learning-based robust navigation operating in cluttered environments,” in2019 International Conference on Robotics and Automation (ICRA), 2019, pp. 9624–9630

    2019

  16. [24]

    Rma: Rapid motor adaptation for legged robots,

    A. Kumar, Z. Fu, D. Pathak, and J. Malik, “Rma: Rapid motor adaptation for legged robots,” 2021. [Online]. Available: https://arxiv.org/abs/2107.04034

    2021 arXiv

  17. [25]

    Legged locomotion in challenging terrains using egocentric vision,

    A. Agarwal, A. Kumar, J. Malik, and D. Pathak, “Legged locomotion in challenging terrains using egocentric vision,” 2022. [Online]. Available: https://arxiv.org/abs/2211.07638

    2022 arXiv

  18. [26]

    Rapid locomotion via reinforcement learning,

    G. B. Margolis, G. Yang, K. Paigwar, T. Chen, and P. Agrawal, “Rapid locomotion via reinforcement learning,” 2022. [Online]. Available: https://arxiv.org/abs/2205.02824

    2022 arXiv

  19. [27]

    Walk these ways: Tuning robot control for generalization with multiplicity of behavior,

    G. B. Margolis and P. Agrawal, “Walk these ways: Tuning robot control for generalization with multiplicity of behavior,” 2022

    2022

  20. [28]

    Proximal policy optimization algorithms,

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” 2017

    2017

  21. [29]

    Bert: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “Bert: Pre-training of deep bidirectional transformers for language understanding,” 2019

    2019

Pith tools

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