Pith. sign in

REVIEW 3 major objections 5 minor 45 references

Bridging Adaptivity and Safety: Learning Agile Collision-Free Locomotion Across Varied Physics

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · 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 stay fast and safe when physics change, if both its controller and its safety guard are told the current payload and friction.

desk verdict A credible incremental extension of ABS with real-world demos, but the 'collision-free' claim outruns the evidence and the theory appendix doesn't cover the learned safety function. read the letter →

arxiv 2501.04276 v3 pith:FOMRKKS6 submitted 2025-01-08 cs.RO cs.LG

Yichao Zhong , Chong Zhang , Tairan He , Guanya Shi This is my paper · ORCID
classification cs.ROcs.LG
keywords reinforcementlearningadaptivesafecontrolleggedlocomotionreach-avoidvaluenetworkphysicalparameterestimationonlineadaptationquadrupedrobotpolicyswitching
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

The paper claims that a legged robot can stay both fast and collision-free when the physics it walks on change, if the control policy and the safety guard both receive live estimates of the physical parameters. To do this, the authors extend the Agile But Safe (ABS) pipeline with an estimator that reads a short history of the robot's own motion and infers payload mass, center-of-mass shift, and ground friction. The agile policy and the reach-avoid value network are conditioned on those estimates, and an on-policy fine-tuning step corrects the estimator for the distribution shift caused by switching between agile and recovery policies. If the claim holds, the robot should navigate slippery floors, added payloads, and slope changes at higher speed and with fewer collisions than non-adaptive or latent-adaptive baselines. The reported real-world result is a 19.8% speed increase and a 2.36 times lower collision rate than ABS.

What carries the argument

The load-bearing mechanism is the parameter-conditioned reach-avoid (RA) value network, a learned approximation of the time-discounted reach-avoid value function $V^\pi_\gamma(s,e)$ governed by the Discounted Reach-Avoid Bellman Equation (DRABE). The RA network outputs a value whose sign decides whether the agile policy keeps control or hands over to a recovery policy; conditioning it on estimated physics parameters makes the switch adapt to the current situation. The second component is the proprioceptive physics-parameter estimator, trained jointly with the agile policy via a fusion interpolation between ground-truth and estimated parameters, and fine-tuned on-policy to reduce the history distribution shift caused by policy switching. A Lipschitz-continuity argument for $V^\pi_\gamma$ with respect to environment parameters $e$ is given to justify that small estimation errors do not cause the safety value to jump discontinuously.

What would settle it

Take the deployed BAS stack and artificially bias the estimator's mass output by a fixed 2 kg offset while leaving friction and center-of-mass estimates correct; if the collision rate or the false-trigger rate changes materially compared with the unbiased run, then the safety guarantee does not hold under exactly the kind of estimation error the estimator is supposed to tolerate.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that explicit, policy-invariant physical parameters, estimated online from proprioception, are a better conditioning signal for a reach-avoid safety guard than the latent dynamics encodings used by adaptive locomotion policies. Conditioning both the agile goal-reaching policy and the learned reach-avoid value function on these estimated parameters makes the safety trigger fire earlier when payload is heavy and later when it is light, matching the intuition that heavier payloads are more dangerous. The paper further claims that the same safety improvement transfers to the real world, where BAS outperforms ABS and an RMA-plus-Lagrangian baseline on both agility and safety metrics across payload, friction, and slope trials.

Load-bearing premise

The load-bearing premise is that the learned reach-avoid value network, trained in simulation with ground-truth parameters, continues to act as a correct safety trigger when it is fed the estimator's real-world outputs; if estimation error pushes the value across the switching threshold at the wrong moment, the recovery policy cannot prevent the collision.

Editorial extensions

If this is right

  • If BAS is correct, adaptive safe locomotion no longer requires knowing the physics in advance: the robot infers payload, friction, and center-of-mass shift from its own motion history, so deployment to an unseen floor or load can reuse the same trained stack.
  • Because the safety guard is parameter-conditioned, the robot can trade off agility and caution online; a heavy payload makes the recovery policy intervene earlier, and removing the payload relaxes the guard without retraining.
  • The on-policy fine-tuning phase implies that estimator accuracy is a bottleneck: after fine-tuning, estimation loss drops from 0.570 to 0.323 and collision rate from 3.10% to 1.11%, suggesting that fidelity of the estimated parameters directly controls safety performance.
  • In simulation, BAS reaches the goal in 93.84% of dynamic-environment episodes versus 63.83% for ABS, so the claimed improvement is not only about avoiding collisions but about completing the task.

Reading between the lines

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

  • A testable extension would be to feed the same estimated parameters into a model-based controller or a Hamilton-Jacobi reachability solver, not just a learned value network, to see whether the empirical safety improvement comes from the conditioning signal or from the learned value representation itself.
  • The paper's own BAS-random ablation, where estimated parameters are replaced with random numbers, collapses success to zero; this suggests the framework is deliberately non-conservative, so a deployment that wants robustness to estimator failure would need a separate monitor that detects abnormally large estimation error.
  • Because the reach-avoid value function is Lipschitz in the estimated parameters, one could in principle compute a worst-case safety margin from the estimator's error bound; the paper does not do this, but the theory presented makes such a bound a natural next step.
Share X Bluesky LinkedIn Reddit HN

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 presents BAS, an extension of the Agile But Safe (ABS) framework for quadruped locomotion. BAS trains an agile obstacle-avoidance policy together with a recovery policy, a proprioception-based estimator of physical parameters (payload mass, CoM shift, friction), and a reach-avoid (RA) value network that governs switching to the recovery policy; both the agile policy and the RA network are conditioned on the estimated parameters. An on-policy fine-tuning stage is introduced to reduce distribution shift in the estimator's input history. The authors report simulation comparisons against ABS, RMA-based, Lagrangian, and ablation baselines, plus real-world tests on a Unitree Go1 under payload, friction, and slope variations, claiming a 19.8% increase in speed and a 2.36 times lower collision rate than ABS.

Significance. If the empirical comparisons are reliable, BAS is a practically useful recipe: it combines an interpretable, policy-invariant physics estimator with a learned RA guard and demonstrates real-world transfer across several physical perturbations. The strengths of the paper are the real-robot experiments, the ablation of the fusion and fine-tuning choices, and the explicit conditioning of both policy and guard on estimated parameters. However, the safety claim is not backed by a formal certificate, and the statistical basis for the headline numbers is thin (simulation results are reported without seeds or error bars, and real-world safety uses only 8 trials per condition); the theoretical appendix does not close this gap.

major comments (3)
  1. [3.2 and Appendix A] The RA guard is the load-bearing safety mechanism: the system switches to the recovery policy when V_gamma(s, e_hat) > 0. The theoretical support in Appendix A does not establish that this sign decision is correct under estimation error. Theorem 1 proves, modulo the proof issues below, Lipschitz continuity of a discounted min-over-tau value function with respect to e, but the manuscript does not show that this function coincides with the fixed point of the DRABE operator used for training in Eq. (2), nor does it bound the approximation error of the learned network or the effect of replacing e by e_hat. Lipschitz continuity alone gives no sign margin: with the estimation loss of 0.323 reported in Table 2 and the acknowledged inaccuracy of real-world friction estimates in Section 4.4, a perturbation could flip the trigger. I recommend adding an explicit conservatism margin or an empirical sensitivity analysis (e.g., the distribution of RA values near the threshold against estimator error), or softening the safety language.
  2. [4.1, 4.4, Tables 1-4] The quantitative claims rest on experiments without variance or significance reporting. Tables 1 and 2 give point estimates with no indication of how many seeds or randomizations were used, and the text treats small differences as meaningful. In the real world, Table 4 uses 8 trials per condition for safety and 3 trials for agility, without confidence intervals or statistical tests. Under a binomial model, the Slope condition (BAS 5/8 vs ABS 3/8) is not statistically distinguishable, and the aggregate 81.25% vs 34.38% averages over heterogeneous conditions. The headline '19.8% speed increase' and '2.36x lower collision rate' should be reported with intervals or explicitly as descriptive statistics. I also note that BAS-random in Table 1(c) collapses to 100% collision, which underscores that the safety margin under partial estimator failure needs explicit quantification.
  3. [Appendix A, Eq. (5)] The proof of Theorem 1 needs correction. The displayed inequality after Eq. (3) has the second term comparing xi^{pi,e2}_s(kappa) with itself, which cannot yield the claimed bound; the trajectory difference Delta xi(t) is then bounded by an expression involving an undefined quantity p. In Eq. (5), the factor L_fpi appears in the upper bound without justification: solving the displayed maximization gives a bound proportional to gamma^{t*} log(1+L_fpi) / -log(gamma(1+L_fpi)), with no extra L_fpi. More substantively, the condition gamma(1+L_fpi)<1 is assumed but never verified for the trained policies; for a legged system with a high-gain policy this condition is not automatic. These issues do not necessarily invalidate the empirical RA training, but the 'theoretical insights' claimed as contribution 4 should either be fixed or be reframed as heuristic motivation.
minor comments (5)
  1. [Tables 1 and 2] Table 1 reports BAS v_peak as 2.70 m/s while Table 2 reports 2.68 m/s for the same configuration; please reconcile the numbers and state the evaluation settings used in each table.
  2. [Abstract and Table 4] The claimed '19.8% increase in speed' should be defined precisely; the agility times in Table 4 (1.74 s vs 1.41 s) imply roughly 23% speed increase if the distance is fixed, so the metric used for the abstract claim should be stated.
  3. [3.3] Please clarify whether the on-policy fine-tuning is performed in simulation only, since real-world deployment has no ground-truth physics available for supervision; if fine-tuning is simulation-only, state this explicitly.
  4. [3.1-3.3 and Appendix] For reproducibility, provide the network architectures, estimator history length, alpha schedule, reward weights, PPO hyperparameters, and the DR parameter sampling procedure used for the reported results.
  5. [Appendix A] There are several typos and undefined symbols: 'collsion' appears in the Table 1 caption, 'Funtion' appears in Appendix A, and the proof uses p1 and p2 without defining them; please correct these.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the adaptive-safety gains are empirically evaluated against external baselines, and the estimator is supervised by ground-truth physics.

full rationale

The central derivation chain is not circular. BAS's claimed improvements are measured by collision rates, reach rates, and real-world trial outcomes (Tables 1 and 4), not by quantities that were fitted into those safety metrics. The physical-parameter estimator is trained with supervised regression against ground-truth simulation parameters (Section 3.1), and the RA value network is trained via standard DRABE value iteration (Equation 2) on rollouts of the same policies, which is policy evaluation rather than a self-fulfilling prediction. The reuse of the authors' prior ABS framework is explicit and declared ('builds upon the pipeline of prior work Agile But Safe (ABS)' and 'we also retain πrecovery designs from ABS'), and ABS serves as an empirical baseline rather than as an unverified premise that forces the conclusion. The theoretical Lipschitz argument in Appendix A is under-specified—it analyzes the undiscounted min-max form of Equation (1) rather than the DRABE fixed point and assumes γ(1+L_fπ)<1 without verifying it for trained policies—but this is an evidentiary gap, not circularity. The paper's disclosed limitations (estimated friction values differ from real-world ground truth, 8 trials per real-world condition, only static obstacles tested) reduce the strength of the safety claims but do not make the derivation self-referential. No load-bearing step reduces, by construction, to its own input.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The core empirical system rests on standard RL and reach-avoid machinery borrowed from cited work, plus unvalidated domain assumptions: DRABE convergence, Lipschitz continuity, static parameters in theory, and sim-to-real validity of the learned value function. The chosen domain randomization ranges and fusion schedule are hand-set and shape all reported results. No new physical entities are postulated; the estimator and RA network are learned modules, not theoretical constructs with independent evidence.

free parameters (4)
  • Domain randomization ranges for payload, friction, CoM shifts, and external forces = Table 3: payload -2 to 12 kg, friction 0.25 to 1.5, CoM shifts -0.05 to 0.15 m, forces -15 to 15 N
    Chosen by hand to define the training distribution; the paper's generalization claim is only evaluated inside these ranges.
  • Fusion interpolation coefficient alpha = alpha = min(2 * training rate, 1)
    Schedule chosen manually to blend ground-truth and estimated physical parameters during joint training; the claimed estimation accuracy depends on it.
  • History length for the estimator = 50 steps (o_t to o_t-49)
    Chosen without an ablation; estimator accuracy and real-time tracking depend on this window.
  • Reward weights for agile and recovery policies = Not reported, inherited from ABS
    The paper says it follows ABS reward settings without listing them; the behavior and safety comparisons depend on these weights.
assumptions (6)
  • standard math The DRABE operator B_gamma is a contraction mapping (Hsu* et al. 2021)
    Used to train the RA network with value iteration (Section 3.2, Equation 2); the proof is cited, not reproduced.
  • standard math V_gamma is Lipschitz continuous in s (Li et al. 2024a)
    Starting point for Theorem 1 in Appendix A; treated as established background.
  • domain assumption The dynamics f_pi is Lipschitz in (s,e) with constant L_f_pi satisfying gamma(1 + L_f_pi) < 1
    Required by Theorem 1; the paper attempts to enforce robustness via L2 regularization and weight clipping but gives no verification that the learned policy satisfies the bound.
  • domain assumption Physical parameters are static within a trajectory for the theoretical analysis
    Stated in Section 2 (footnote 1); real-world experiments allow time-variant parameters, so the theory only covers a special case of deployment.
  • domain assumption Sim-to-real transfer of the RA value function and estimator
    The central safety claim assumes IsaacGym training with domain randomization produces a value function valid on the real robot; supported only by 8-trial statistics in Table 4, not by formal certification.
  • domain assumption Sensor mapping o = h(s) provides enough information to recover physical parameters
    The estimator regresses payload, CoM shift, and friction from proprioceptive history; identifiability is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bridging Adaptivity and Safety: Learning Agile Collision-Free Locomotion Across Varied Physics." pith.science (2026). https://pith.science/paper/FOMRKKS6

@misc{pith2026250104276,
  author       = {Pith},
  title        = {Pith review of: Bridging Adaptivity and Safety: Learning Agile Collision-Free Locomotion Across Varied Physics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FOMRKKS6}},
  note         = {Machine review of arXiv:2501.04276}
}
read the original abstract

Real-world legged locomotion systems often need to reconcile agility and safety for different scenarios. Moreover, the underlying dynamics are often unknown and time-variant (e.g., payload, friction). In this paper, we introduce BAS (Bridging Adaptivity and Safety), which builds upon the pipeline of prior work Agile But Safe (ABS)(He et al.) and is designed to provide adaptive safety even in dynamic environments with uncertainties. BAS involves an agile policy to avoid obstacles rapidly and a recovery policy to prevent collisions, a physical parameter estimator that is concurrently trained with agile policy, and a learned control-theoretic RA (reach-avoid) value network that governs the policy switch. Also, the agile policy and RA network are both conditioned on physical parameters to make them adaptive. To mitigate the distribution shift issue, we further introduce an on-policy fine-tuning phase for the estimator to enhance its robustness and accuracy. The simulation results show that BAS achieves 50% better safety than baselines in dynamic environments while maintaining a higher speed on average. In real-world experiments, BAS shows its capability in complex environments with unknown physics (e.g., slippery floors with unknown frictions, unknown payloads up to 8kg), while baselines lack adaptivity, leading to collisions or. degraded agility. As a result, BAS achieves a 19.8% increase in speed and gets a 2.36 times lower collision rate than ABS in the real world. Videos: https://adaptive-safe-locomotion.github.io.

Figures

Figures reproduced from arXiv: 2501.04276 by the authors.

Figure 1
Figure 1. 1) The robot can handle collision-free locomotion in even super slippery terrain [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. BAS Pipeline Overview. challenges and then propose fusion interpolation in the joint-train pipeline to boost the accuracy further. Additionally, note that physical parameters are policy-invariant variables. So, compared to pre￾dicting dynamics that tangle with policies, predicting physical parameters is more suitable for cases where multiple policies are used together like He et al. (2024b); Hoeller et al. (2023). W… view at source ↗
Figure 3
Figure 3. Mass estimation tracking of BAS, BAS w/o fusion and BAS w/o joint-train pipeline. [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Heatmaps of RA values under the different mass of payloads at the state of 3.0m/s base [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Real Experiment for Adaptive Safety test Settings, where [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Adaptation analysis in real with online changes in the environment. 1) BAS [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 12 canonical work pages

  1. [1]

    Constrained policy optimization, 2017

    Joshua Achiam, David Held, Aviv Tamar, and Pieter Abbeel. Constrained policy optimization, 2017. URL https://arxiv.org/abs/1705.10528

    arXiv 2017

  2. [2]

    Development of a prototype of a search and rescue robot equipped with multiple cameras

    Mukhriddin Arabboev, Shohruh Begmatov, Khabibullo Nosirov, Alisher Shakhobiddinov, Jean Chamberlain Chedjou, and Kyandoghere Kyamakya. Development of a prototype of a search and rescue robot equipped with multiple cameras. In 2021 International Conference on Information Science and Communications Technologies (ICISCT), pages 1--5, 2021. doi:10.1109/ICISCT...

    arXiv 2021

  3. [3]

    Somil Bansal, Mo Chen, Sylvia Herbert, and Claire J. Tomlin. Hamilton-jacobi reachability: A brief overview and recent advances, 2017. URL https://arxiv.org/abs/1709.07523

    work page Pith review arXiv 2017

  4. [4]

    Parameter-conditioned reachable sets for updating safety assurances online

    Javier Borquez, Kensuke Nakamura, and Somil Bansal. Parameter-conditioned reachable sets for updating safety assurances online. In 2023 IEEE International Conference on Robotics and Automation (ICRA), page 10553–10559. IEEE, May 2023. doi:10.1109/icra48891.2023.10160554. URL http://dx.doi.org/10.1109/ICRA48891.2023.10160554

    arXiv 2023

  5. [5]

    Hall, Zhaocong Yuan, Siqi Zhou, Jacopo Panerati, and Angela P

    Lukas Brunke, Melissa Greeff, Adam W. Hall, Zhaocong Yuan, Siqi Zhou, Jacopo Panerati, and Angela P. Schoellig. Safe learning in robotics: From learning-based control to safe reinforcement learning. Annual Review of Control, Robotics, and Autonomous Systems, 5 0 (Volume 5, 2022): 0 411--444, 2022. ISSN 2573-5144. doi:https://doi.org/10.1146/annurev-contro...

    doi:10.1146/annurev-control-042920-020211 2022

  6. [6]

    Perceptive whole-body planning for multilegged robots in confined spaces

    Russell Buchanan, Lorenz Wellhausen, Marko Bjelonic, Tirthankar Bandyopadhyay, Navinda Kottege, and Marco Hutter. Perceptive whole-body planning for multilegged robots in confined spaces. Journal of Field Robotics, 38 0 (1): 0 68--84, 2021. doi:https://doi.org/10.1002/rob.21974. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/rob.21974

    doi:10.1002/rob.21974 2021

  7. [7]

    A Collision-Free MPC for Whole-Body Dynamic Locomotion and Manipulation

    Jia-Ruei Chiu, Jean-Pierre Sleiman, Mayank Mittal, Farbod Farshidian, and Marco Hutter. A collision-free mpc for whole-body dynamic locomotion and manipulation, 2022. URL https://arxiv.org/abs/2202.12385

    work page Pith review arXiv 2022

  8. [8]

    Coohoi: Learning cooperative human-object interaction with manipulated object dynamics

    Jiawei Gao, Ziqin Wang, Zeqi Xiao, Jingbo Wang, Tai Wang, Jinkun Cao, Xiaolin Hu, Si Liu, Jifeng Dai, and Jiangmiao Pang. Coohoi: Learning cooperative human-object interaction with manipulated object dynamics. arXiv preprint arXiv:2406.14558, 2024

    arXiv 2024

Show all 45 references
  1. [9]

    Henry Gouk, Eibe Frank, Bernhard Pfahringer, and Michael J. Cree. Regularisation of neural networks by enforcing lipschitz continuity, 2020. URL https://arxiv.org/abs/1804.04368

    2020 arXiv

  2. [10]

    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 a

    2024 arXiv

  3. [11]

    Agile but safe: Learning collision-free high-speed legged locomotion, 2024 b

    Tairan He, Chong Zhang, Wenli Xiao, Guanqi He, Changliu Liu, and Guanya Shi. Agile but safe: Learning collision-free high-speed legged locomotion, 2024 b . URL https://arxiv.org/abs/2401.17583

    2024 arXiv

  4. [12]

    Anymal parkour: Learning agile navigation for quadrupedal robots, 2023

    David Hoeller, Nikita Rudin, Dhionis Sako, and Marco Hutter. Anymal parkour: Learning agile navigation for quadrupedal robots, 2023. URL https://arxiv.org/abs/2306.14874

    2023 arXiv

  5. [13]

    Safety and liveness guarantees through reach-avoid reinforcement learning

    Kai-Chieh Hsu*, Vicenç Rubies-Royo*, Claire Tomlin, and Jaime Fisac. Safety and liveness guarantees through reach-avoid reinforcement learning. In Robotics: Science and Systems XVII, RSS2021. Robotics: Science and Systems Foundation, July 2021. doi:10.15607/rss.2021.xvii.077. ...

    2021 doi

  6. [14]

    Ren, Duy P

    Kai-Chieh Hsu, Allen Z. Ren, Duy P. Nguyen, Anirudha Majumdar, and Jaime F. Fisac. Sim-to-lab-to-real: Safe reinforcement learning with shielding and generalization guarantees. Artificial Intelligence, 314: 0 103811, January 2023. ISSN 0004-3702. doi:10.1016/j.artint.2022.1038...

    2023

  7. [15]

    Learning agile and dynamic motor skills for legged robots

    Jemin Hwangbo, Joonho Lee, Alexey Dosovitskiy, Dario Bellicoso, Vassilios Tsounis, Vladlen Koltun, and Marco Hutter. Learning agile and dynamic motor skills for legged robots. Science Robotics, 4 0 (26), January 2019. ISSN 2470-9476. doi:10.1126/scirobotics.aau5872. URL http:/...

    2019 doi

  8. [16]

    Concurrent training of a control policy and a state estimator for dynamic and robust legged locomotion

    Gwanghyeon Ji, Juhyeok Mun, Hyeongjun Kim, and Jemin Hwangbo. Concurrent training of a control policy and a state estimator for dynamic and robust legged locomotion. IEEE Robotics and Automation Letters, 7 0 (2): 0 4630–4637, April 2022. ISSN 2377-3774. doi:10.1109/lra.2022.31...

    2022

  9. [17]

    D. Kim, D. Carballo, J. Di Carlo, B. Katz, G. Bledt, B. Lim, and S. Kim. Vision aided dynamic exploration of unstructured terrain with a small-scale quadruped robot. In 2020 IEEE International Conference on Robotics and Automation (ICRA), pages 2464--2470, 2020. doi:10.1109/IC...

    2020

  10. [18]

    Openvla: An open-source vision-language-action model, 2024

    Moo Jin Kim, Karl Pertsch, Siddharth Karamcheti, Ted Xiao, Ashwin Balakrishna, Suraj Nair, Rafael Rafailov, Ethan Foster, Grace Lam, Pannag Sanketi, Quan Vuong, Thomas Kollar, Benjamin Burchfiel, Russ Tedrake, Dorsa Sadigh, Sergey Levine, Percy Liang, and Chelsea Finn. Openvla...

    2024 arXiv

  11. [19]

    Rma: Rapid motor adaptation for legged robots, 2021

    Ashish Kumar, Zipeng Fu, Deepak Pathak, and Jitendra Malik. Rma: Rapid motor adaptation for legged robots, 2021. URL https://arxiv.org/abs/2107.04034

    2021 arXiv

  12. [20]

    Learning quadrupedal locomotion over challenging terrain

    Joonho Lee, Jemin Hwangbo, Lorenz Wellhausen, Vladlen Koltun, and Marco Hutter. Learning quadrupedal locomotion over challenging terrain. Science Robotics, 5 0 (47), October 2020. ISSN 2470-9476. doi:10.1126/scirobotics.abc5986. URL http://dx.doi.org/10.1126/scirobotics.abc5986

    2020 doi

  13. [21]

    Offline reinforcement learning: Tutorial, review, and perspectives on open problems, 2020

    Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. Offline reinforcement learning: Tutorial, review, and perspectives on open problems, 2020. URL https://arxiv.org/abs/2005.01643

    2020 arXiv

  14. [22]

    Jingqi Li, Donggun Lee, Somayeh Sojoudi, and Claire J. Tomlin. Infinite-horizon reach-avoid zero-sum games via deep reinforcement learning, 2024 a . URL https://arxiv.org/abs/2203.10142

    2024 arXiv

  15. [23]

    Reinforcement learning for versatile, dynamic, and robust bipedal locomotion control, 2024 b

    Zhongyu Li, Xue Bin Peng, Pieter Abbeel, Sergey Levine, Glen Berseth, and Koushil Sreenath. Reinforcement learning for versatile, dynamic, and robust bipedal locomotion control, 2024 b . URL https://arxiv.org/abs/2401.16889

    2024 arXiv

  16. [24]

    Robust regression for safe exploration in control

    Anqi Liu, Guanya Shi, Soon-Jo Chung, Anima Anandkumar, and Yisong Yue. Robust regression for safe exploration in control. In Learning for Dynamics and Control, pages 608--619. PMLR, 2020

    2020

  17. [25]

    Regularization matters in policy optimization, 2021

    Zhuang Liu, Xuanlin Li, Bingyi Kang, and Trevor Darrell. Regularization matters in policy optimization, 2021. URL https://arxiv.org/abs/1910.09191

    2021 arXiv

  18. [26]

    Constrained variational policy optimization for safe reinforcement learning, 2022

    Zuxin Liu, Zhepeng Cen, Vladislav Isenbaev, Wei Liu, Zhiwei Steven Wu, Bo Li, and Ding Zhao. Constrained variational policy optimization for safe reinforcement learning, 2022. URL https://arxiv.org/abs/2201.11927

    2022 arXiv

  19. [27]

    Hybrid internal model: Learning agile legged locomotion with simulated robot response

    Junfeng Long, ZiRui Wang, Quanyi Li, Liu Cao, Jiawei Gao, and Jiangmiao Pang. Hybrid internal model: Learning agile legged locomotion with simulated robot response. In The Twelfth International Conference on Learning Representations, 2024

    2024

  20. [28]

    Pie: Parkour with implicit-explicit learning framework for legged robots, 2024

    Shixin Luo, Songbo Li, Ruiqi Yu, Zhicheng Wang, Jun Wu, and Qiuguo Zhu. Pie: Parkour with implicit-explicit learning framework for legged robots, 2024. URL https://arxiv.org/abs/2408.13740

    2024 arXiv

  21. [29]

    Isaac gym: High performance gpu-based physics simulation for robot learning, 2021

    Viktor Makoviychuk, Lukasz Wawrzyniak, Yunrong Guo, Michelle Lu, Kier Storey, Miles Macklin, David Hoeller, Nikita Rudin, Arthur Allshire, Ankur Handa, and Gavriel State. Isaac gym: High performance gpu-based physics simulation for robot learning, 2021. URL https://arxiv.org/a...

    2021 arXiv

  22. [30]

    Hamilton–jacobi formulation for reach–avoid differential games

    Kostas Margellos and John Lygeros. Hamilton–jacobi formulation for reach–avoid differential games. IEEE Transactions on Automatic Control, 56 0 (8): 0 1849--1861, 2011. doi:10.1109/TAC.2011.2105730

    2011

  23. [31]

    Panahi, Fereidoun H

    Farzad H. Panahi, Fereidoun H. Panahi, and Tomoaki Ohtsuki. An intelligent path planning mechanism for firefighting in wireless sensor and actor networks. IEEE Internet of Things Journal, 10 0 (11): 0 9646--9661, 2023. doi:10.1109/JIOT.2023.3235998

    2023

  24. [32]

    Benchmarking safe exploration in deep reinforcement learning

    Alex Ray, Joshua Achiam, and Dario Amodei. Benchmarking safe exploration in deep reinforcement learning. Technical report, OpenAI, 2019. URL https://openai.com/research/benchmarking-safe-exploration-in-deep-reinforcement-learning

    2019

  25. [33]

    Learning to walk in minutes using massively parallel deep reinforcement learning, 2022

    Nikita Rudin, David Hoeller, Philipp Reist, and Marco Hutter. Learning to walk in minutes using massively parallel deep reinforcement learning, 2022. URL https://arxiv.org/abs/2109.11978

    2022 arXiv

  26. [34]

    Mastering the game of Go without human knowledge

    David Silver, Julian Schrittwieser, Karen Simonyan, Ioannis Antonoglou, Aja Huang, Arthur Guez, Thomas Hubert, Lucas Baker, Matthew Lai, Adrian Bolton, Yutian Chen, Timothy Lillicrap, Fan Hui, Laurent Sifre, George van den Driessche, Thore Graepel, and Demis Hassabis. Masterin...

    2017 doi

  27. [35]

    Path planning of forest fire-fighting robots based on deep learning

    Shang-jie Sun, Shu-hai Jiang, Song-he Cui, Yue Kang, and Yu-tang Chen. Path planning of forest fire-fighting robots based on deep learning. 36: 0 51–57, 2020. ISSN 1006-8023

    2020

  28. [36]

    Toward understanding key estimation in learning robust humanoid locomotion, 2024

    Zhicheng Wang, Wandi Wei, Ruiqi Yu, Jun Wu, and Qiuguo Zhu. Toward understanding key estimation in learning robust humanoid locomotion, 2024. URL https://arxiv.org/abs/2403.05868

    2024 arXiv

  29. [37]

    Safe deep policy adaptation, 2024 a

    Wenli Xiao, Tairan He, John Dolan, and Guanya Shi. Safe deep policy adaptation, 2024 a . URL https://arxiv.org/abs/2310.08602

    2024 arXiv

  30. [38]

    Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility

    Wenli Xiao, Haoru Xue, Tony Tao, Dvij Kalaria, John M Dolan, and Guanya Shi. Anycar to anywhere: Learning universal dynamics model for agile and adaptive mobility. arXiv preprint arXiv:2409.15783, 2024 b

    2024 arXiv

  31. [39]

    Crpo: A new approach for safe reinforcement learning with convergence guarantee, 2021

    Tengyu Xu, Yingbin Liang, and Guanghui Lan. Crpo: A new approach for safe reinforcement learning with convergence guarantee, 2021. URL https://arxiv.org/abs/2011.05869

    2021 arXiv

  32. [40]

    Full-order sampling-based mpc for torque-level locomotion control via diffusion-style annealing

    Haoru Xue, Chaoyi Pan, Zeji Yi, Guannan Qu, and Guanya Shi. Full-order sampling-based mpc for torque-level locomotion control via diffusion-style annealing. arXiv preprint arXiv:2409.15610, 2024

    2024 arXiv

  33. [41]

    Cajun: Continuous adaptive jumping using a learned centroidal controller, 2023

    Yuxiang Yang, Guanya Shi, Xiangyun Meng, Wenhao Yu, Tingnan Zhang, Jie Tan, and Byron Boots. Cajun: Continuous adaptive jumping using a learned centroidal controller, 2023. URL https://arxiv.org/abs/2306.09557

    2023 arXiv

  34. [42]

    Yun, Rui Chen, Chase Dunaway, John M

    Kai S. Yun, Rui Chen, Chase Dunaway, John M. Dolan, and Changliu Liu. Safe control of quadruped in varying dynamics via safety index adaptation, 2024. URL https://arxiv.org/abs/2409.09882

    2024

  35. [43]

    Wococo: Learning whole-body humanoid control with sequential contacts

    Chong Zhang, Wenli Xiao, Tairan He, and Guanya Shi. Wococo: Learning whole-body humanoid control with sequential contacts. arXiv preprint arXiv:2406.06005, 2024 a

    2024 arXiv

  36. [44]

    Catch it! learning to catch in flight with mobile dexterous hands

    Yuanhang Zhang, Tianhai Liang, Zhenyang Chen, Yanjie Ze, and Huazhe Xu. Catch it! learning to catch in flight with mobile dexterous hands. arXiv preprint arXiv:2409.10319, 2024 b

    2024 arXiv

  37. [45]

    3d-vla: A 3d vision-language-action generative world model, 2024

    Haoyu Zhen, Xiaowen Qiu, Peihao Chen, Jincheng Yang, Xin Yan, Yilun Du, Yining Hong, and Chuang Gan. 3d-vla: A 3d vision-language-action generative world model, 2024. URL https://arxiv.org/abs/2403.09631

    2024 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.