Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Bracing for Impact: Robust Humanoid Push Recovery and Locomotion with Reduced Order Models

T0 review · 4 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 reduced-order controller lets a walking humanoid brace against walls and recover from torso pushes up to 100 N.

desk verdict The integration is real and the evaluation is systematic, but the paper never isolates the arm-wall bracing from the other controller changes, so its central claim is unverified. read the letter →

arxiv 2505.11495 v2 pith:UZZJ4CCW submitted 2025-05-16 cs.RO

Lizhi Yang , Blake Werner , Adrian B.Ghansah , Aaron D. Ames This is my paper · ORCID
classification cs.RO
keywords humanoidpushrecoveryreducedordermodelsmodelpredictivecontrolHybridLinearInvertedPendulumSingleRigidBodymulti-contactlocomotionwallbracingsimulation
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

This paper argues that a humanoid robot can keep walking through strong pushes by treating nearby walls as extra points of support, using its arms to brace while the legs adjust their steps. The authors combine two established reduced-order models—the Single Rigid Body (SRB) model, which plans contact forces on a torso with lumped mass, and the Hybrid Linear Inverted Pendulum (HLIP) model, which sets step lengths and timing—into one model-predictive controller that switches into a recovery mode when a push is detected. In simulation of a 24-degree-of-freedom, 20 kg humanoid, the combined controller survives 70.6% of 1,440 push scenarios (up to 100 N for 0.2 s at walking speeds up to 0.5 m/s), while the HLIP-only baseline survives 16.8%. The paper takes this as evidence that multi-contact, wall-assisted recovery can be planned in real time from reduced-order models rather than requiring whole-body optimization or learning.

What carries the argument

The load-bearing object is the SRB-MPC-HLIP controller, a model-predictive controller built on two reduced-order models. The SRB model (Equation (6) and its discretized form in Equation (11)) treats the robot as a single rigid body with mass and inertia concentrated at the torso, acted on by contact forces and moments at the hands, left foot, and right foot; the HLIP model (Equations (22)–(23)) supplies step-to-step dynamics that prescribe swing-foot placement. The MPC layer solves the quadratic program of Equation (18) with the contact constraints of Equation (17), including hand-on-wall friction cones, and maps optimal contact wrenches to joint torques through Jacobians. The mechanism that carries the argument is the joint activation of hand contact (via inverse kinematics to the wall) and step-pattern adjustment (higher step frequency, wall-aware step limits) whenever the predicted CoM states deviate beyond set thresholds.

What would settle it

Run the same 1,440-push battery on hardware (or in a full-order simulation that retains arm and leg inertia) with the wall within reach. If the fraction of survived pushes falls to near the 16.8% HLIP-only level, or if measured hand contact forces deviate systematically from the MPC-commanded values, the central claim that SRB-based bracing causes the recovery improvement would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that combining SRB-MPC with HLIP dynamics lets a walking humanoid use its arms to brace against walls and still recover from pushes that would defeat a stepping-only controller. The SRB-MPC layer solves, at 250 Hz or faster, a quadratic program for hand and foot contact forces that satisfies friction-cone and force-bound constraints; the HLIP layer raises step frequency from 3 Hz to 5 Hz during recovery and lengthens steps in the push direction to enlarge the support polygon. The two layers are coupled through a push detector that compares predicted center-of-mass velocity and angular velocity against reference values (0.4 m/s and 0.2 rad/s thresholds) and through forward-kinematics knowledge of wall distance. On the 1,440-case push battery, the combined controller lifts the safe-set fraction from 16.8% to 70.6%, a 420% increase, and it also keeps torso angular velocities closer to zero during nominally commanded walking. The authors further report recovery from multi-directional pushes against walls slanted inward at 5 degrees.

Load-bearing premise

The plan assumes the robot's mass and inertia are concentrated in the torso, so arms and legs act as massless force generators; if limb dynamics contribute meaningfully to balance during large pushes, the planned contact forces will not match the real robot's behavior.

Editorial extensions

If this is right

  • A 20 kg humanoid with 24 degrees of freedom can survive a 100 N, 0.2 s torso push while walking at 0.5 m/s, provided a reachable wall exists.
  • The same controller, without retraining, handles walls slanted inward by 5 degrees and pushes arriving from either side.
  • Worst-case omnidirectional pushes without any wall are rejected up to 30 N, against 20 N for HLIP alone, so some robustness gain comes from stepping even before arm contact.
  • Recovery is fast enough for real-time control: the MPC QP runs at 250 Hz and the step period shortens from 3 Hz to 5 Hz during recovery.
  • The safe-set increase from 16.8% to 70.6% indicates that environmental bracing, not just stepping, is what closes most of the gap.

Reading between the lines

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

  • If the same architecture transfers to hardware, walls, railings, and furniture could be treated as a single unified contact resource during locomotion, which would change how humanoid deployment sites are prepared.
  • A natural next experiment would be pushes from behind or pushes that occur when no wall is within arm reach, since the current recovery mode is gated on arm workspace reaching a wall; the paper does not report those cases.
  • One could test the SRB assumption directly by adding mass to the arms or legs and rerunning the push battery; if the safe-set fraction drops toward the HLIP baseline, the torso-lumped-mass idealization is the binding constraint.
  • The 0.4 m/s and 0.2 rad/s detection thresholds are tuned empirically; a learned or adaptive threshold could trade false alarms against missed pushes, an extension the paper lists as future work.
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

4 major / 5 minor

Summary. This paper proposes a unified push-recovery and locomotion controller for a 24-DOF, 20 kg humanoid, combining a Single Rigid Body model predictive controller (SRB-MPC) with Hybrid Linear Inverted Pendulum (HLIP) stepping control. The robot uses its arms to brace against walls detected via LiDAR when a push is detected, and the MPC re-plans hand and foot contact forces while HLIP adjusts step timing and placement. The authors evaluate the approach in MuJoCo over 1,440 push scenarios (forces 30-100 N, heights 0.1-0.5 m, commanded speeds 0-0.5 m/s, two push sides, three push times), reporting that the SRB-MPC-HLIP controller survives 70.6% of scenarios versus 16.8% for HLIP alone, plus a qualitative angled-wall and multi-directional push demonstration. The paper also includes tracking comparisons and a no-wall push test while stepping in place.

Significance. If the results are supported, the paper would provide a useful modular architecture: a convex MPC formulation with multi-contact wall bracing, a principled HLIP-based footstep adaptation, and an unusually systematic perturbation sweep. Strengths include the explicit derivation of the SRB-MPC dynamics and constraints, the 1,440-case evaluation protocol, the use of a high-fidelity simulator, and the fact that no parameter was fitted to the reported safe-set metric. However, the central causal claim about arm/wall bracing needs a direct ablation, and several controller parameters and per-cell results are not reported; these gaps currently prevent the quantitative claims from being fully assessed.

major comments (4)
  1. [Section IV-C compared with Sections III-A and III-B] The paper's headline contribution is that the arms brace against walls, but the only quantitative comparison is the full SRB-MPC-HLIP controller versus pure HLIP. That comparison cannot separate the contribution of the hand-contact force from the simultaneous changes in foot-force optimization, push-detection thresholds, step-frequency increase (3 Hz to 5 Hz), and wall detection. Please add an ablation that disables the hand-contact force (uh = 0 in Eq. (5)) while keeping every other module and tuning fixed, and report the same 1,440-case safe-set metric. Without this, the improvement cannot be attributed to wall bracing.
  2. [Section IV-C] The phrase 'recover from pushes up to 100 N for 0.2 s while walking at commanded speeds up to 0.5 m/s' is ambiguous because the aggregate 70.6% safe set (Fig. 4) does not show whether the extreme cells (100 N at 0.5 m/s, including each height, side, and phase combination) individually succeed. Please provide per-condition success counts or a heatmap over the swept grid, and state whether the extreme values describe a single successful trajectory or a robust region.
  3. [Sections III-A and III-D] The push-detection thresholds (0.4 m/s and 0.2 rad/s) are described as based on 'emperical testing,' and the MPC cost weights Q and R, the HLIP gain K, and the contact-force bounds Ffmax, Fhmin, Fhmax are not reported. Because these parameters determine when recovery is activated and how forces are allocated, the 70.6% result is not reproducible and its sensitivity to these tunings is unknown. Please report the numeric values or release the code/configuration, and include a sensitivity study for the detection thresholds.
  4. [Sections II-B and V] Eq. (6) is derived under the assumption that the robot's mass and inertia are concentrated in the torso and that limb dynamics are negligible, and the paper itself lists 'internal disturbances caused by the movements of the limbs' as future work. For the largest pushes in the sweep, the discrepancy between the SRB-predicted contact forces and the full-order MuJoCo response is unquantified. Please compare planned versus actual hand and foot forces in simulation, or otherwise bound the model error, so that the wall-bracing forces are credible at the extremes of the tested range. This is a correctness-risk concern, not a claim of circularity.
minor comments (5)
  1. [Sections III-A and III-G] There are typos: 'emperical' should be 'empirical' in Section III-A, and 'feed-foward' should be 'feed-forward' in Section III-G.
  2. [Figure 4 and Section IV-C] The reported increase is numerically inconsistent: 16.8% to 70.6% is a factor of 4.20, which is a 320% increase, not a 420% increase; the caption of Fig. 4(a) says a '225% larger safeset' while the forces shown are 20 N and 30 N. Please harmonize these numbers.
  3. [Section IV-A] Please define the friction coefficient μ and the default MuJoCo friction settings, since the friction cones in Eq. (17) depend on them.
  4. [Section III-F] The HLIP feedback gain K in Eq. (23) is said to stabilize A - BK, but the dimension of the HLIP state and the construction of K are not specified; a brief definition would improve reproducibility.
  5. [Section IV] The paper does not report error bars or repeated stochastic trials; if the simulator has any nondeterminism, please state the number of seeds used for each condition.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the safe-set gain is measured, not fitted; the controller's inputs are independent of the reported evaluation metric.

full rationale

The paper's central quantitative claim, the 70.6% safe-set coverage from 1,440 simulated push cases, is the output of an evaluation, not an input to the controller design. No parameter is fitted to the safe-set metric or to the 16.8%-versus-70.6% comparison. The SRB and HLIP models are defined by their own dynamics (Eqs. (1), (2), (11)), and the MPC cost weights, friction-cone constraints, and push-detection thresholds are fixed controller design choices; the detection thresholds are described as based on empirical testing, but they are not the quantity being predicted. The HLIP baseline is prior work used for comparison, and the cited same-lab HLIP references are components of the implementation rather than the sole justification for the claimed improvement. The closest concern is that no arm-disabled ablation isolates the causal role of wall bracing, but that is an attribution/completeness issue, not a circular reduction: the reported result does not reduce by construction to any fitted input or self-citation. Thus no specific circular step can be identified under the required standard.

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

The central contribution is an integration of existing reduced-order models; no new physical entity is introduced. The main uncharged items are controller tuning parameters and domain assumptions about model fidelity, contact, and simulation transfer.

free parameters (6)
  • Push detection linear velocity threshold = 0.4 m/s
    Section III-A: recovery mode activates when predicted CoM velocity deviates more than 0.4 m/s from reference, chosen 'based on empirical testing'.
  • Push detection angular velocity threshold = 0.2 rad/s
    Section III-A: recovery mode activates when predicted angular velocity deviates more than 0.2 rad/s from reference, chosen 'based on empirical testing'.
  • MPC cost weights Q and R = not reported
    Sections II-D and III-D define the QP with weighting matrices Q and R but no values or tuning procedure are given; they strongly affect the optimized contact forces.
  • HLIP feedback gain K = not reported
    Section III-F, Eq. (23): the stepping law uses a gain K stabilizing A-BK, but K is not specified.
  • Stepping frequency in recovery mode = 5 Hz (recovery) versus 3 Hz (normal)
    Section III-A: the robot increases stepping frequency from 3 Hz to 5 Hz during recovery; this choice is not justified or swept.
  • Contact force bounds Ffmax, Fhmin, Fhmax = not reported
    Section III-C, Eq. (17): maximum foot force and minimum/maximum hand forces are 'dynamically adjusted' but the values and adjustment law are not provided.
assumptions (6)
  • domain assumption The robot's mass and inertia are concentrated in the torso, with arms and legs effectively massless force generators.
    Section II-B states the SRB model is reasonable because mass and inertia are concentrated around the torso and limbs are relatively light; the entire MPC formulation in Section III-B depends on this.
  • domain assumption Gyroscopic torque is negligible and roll/pitch angles are small enough for linearization.
    Section III-B: the paper neglects omega cross (I omega) and uses small-angle assumptions in Eq. (9) through Eq. (11); these become questionable for large pushes that produce large torso rotations.
  • domain assumption The discretized dynamics computed from the current state remain valid despite large deviations from the reference trajectory.
    Section III-C asserts Bd[i] is calculated from current robot states and is always correct, but no formal guarantee is given for the linearized model under large pushes.
  • domain assumption HLIP planar dynamics and step-to-step feedback describe the CoM behavior during perturbed locomotion.
    Section II-C: HLIP is used to generate swing foot trajectories and support polygons; the paper assumes this reduced model remains valid under the pushes tested.
  • domain assumption LiDAR-based wall detection and rigid wall contact are reliable in the tested scenarios.
    Section III-A uses line detection from LiDAR to determine dwall and Section IV uses MuJoCo default friction and rigid walls; the controller assumes the hand can actually make and sustain contact.
  • domain assumption MuJoCo simulation with Stage sensor simulation faithfully predicts hardware behavior.
    Section IV validates only in simulation; no hardware experiments are reported, so transfer to a physical robot with sensor noise, compliance, and actuator limits is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bracing for Impact: Robust Humanoid Push Recovery and Locomotion with Reduced Order Models." pith.science (2026). https://pith.science/paper/UZZJ4CCW

@misc{pith2026250511495,
  author       = {Pith},
  title        = {Pith review of: Bracing for Impact: Robust Humanoid Push Recovery and Locomotion with Reduced Order Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UZZJ4CCW}},
  note         = {Machine review of arXiv:2505.11495}
}
read the original abstract

Push recovery during locomotion will facilitate the deployment of humanoid robots in human-centered environments. In this paper, we present a unified framework for walking control and push recovery for humanoid robots, leveraging the arms for push recovery while dynamically walking. The key innovation is to use the environment, such as walls, to facilitate push recovery by combining Single Rigid Body model predictive control (SRB-MPC) with Hybrid Linear Inverted Pendulum (HLIP) dynamics to enable robust locomotion, push detection, and recovery by utilizing the robot's arms to brace against such walls and dynamically adjusting the desired contact forces and stepping patterns. Extensive simulation results on a humanoid robot demonstrate improved perturbation rejection and tracking performance compared to HLIP alone, with the robot able to recover from pushes up to 100N for 0.2s while walking at commanded speeds up to 0.5m/s. Robustness is further validated in scenarios with angled walls and multi-directional pushes.

Figures

Figures reproduced from arXiv: 2505.11495 by the authors.

Figure 1
Figure 1. The proposed SRB-MPC-HLIP controller successfully uti [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Framework for the SRB-MPC-HLIP controller. The robot runs state estimation and wall detection based on the sensor data from [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Tracking performance of the pure HLIP controller and the SRB-MPC-HLIP controller. The proposed controller can maintain [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: (a) Worst-case all-direction safeset of both controllers when pushed while stepping in place with no environmental assistance. The [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Snapshots from simulated experiments,, the robot is pushed [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. A Change of Frame Makes Balance Observable: Distillation-Free Humanoid Single-Leg Stance

    cs.RO 2026-08 conditional novelty 6.0 of 10

    A support-relative dynamic capture-point observation, reconstructible without base linear velocity, lets a humanoid policy hold clean single-leg balance at 86/90 in simulation and deploy on a Unitree G1 without distillation.

Reference graph

Works this paper leans on

33 extracted references · 24 canonical work pages · cited by 1 Pith paper

  1. [1]

    Leg stiffness and stride frequency in human running,

    C. T. Farley and O. Gonzalez, “Leg stiffness and stride frequency in human running,” Journal of biomechanics, vol. 29, no. 2, pp. 181–186, 1996

    work page 1996

  2. [2]

    Human-inspired control of bipedal walking robots,

    A. D. Ames, “Human-inspired control of bipedal walking robots,” IEEE Transactions on Automatic Control , vol. 59, no. 5, pp. 1115– 1130, 2014

    work page 2014

  3. [3]

    No falls, no resets: Reliable humanoid behavior in the darpa robotics challenge,

    C. G. Atkeson, B. P. W. Babu, N. Banerjee, D. Berenson, C. P. Bove, X. Cui, M. DeDonato, R. Du, S. Feng, P. Franklin, et al., “No falls, no resets: Reliable humanoid behavior in the darpa robotics challenge,” in 2015 IEEE-RAS 15th International Conference on Humanoid Robots (Humanoids). IEEE, 2015, pp. 623–630

    work page 2015

  4. [4]

    A compre- hensive survey on humanoid robot development,

    S. Saeedvand, M. Jafari, H. S. Aghdasi, and J. Baltes, “A compre- hensive survey on humanoid robot development,” The Knowledge Engineering Review, vol. 34, p. e20, 2019

    work page 2019

  5. [5]

    Robust feedback control of zmp-based gait for the humanoid robot nao,

    J. J. Alcaraz-Jim ´enez, D. Herrero-P ´erez, and H. Mart ´ınez-Barber´a, “Robust feedback control of zmp-based gait for the humanoid robot nao,” The International Journal of Robotics Research , vol. 32, no. 9-10, pp. 1074–1088, 2013

    work page 2013

  6. [6]

    Animated cassie: A dynamic relatable robotic character,

    Z. Li, C. Cummings, and K. Sreenath, “Animated cassie: A dynamic relatable robotic character,” in 2020 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) . IEEE, 2020, pp. 3739–3746

    work page 2020

  7. [7]

    Force-and-moment-based model predictive control for achieving highly dynamic locomotion on bipedal robots,

    J. Li and Q. Nguyen, “Force-and-moment-based model predictive control for achieving highly dynamic locomotion on bipedal robots,” in 2021 60th IEEE Conference on Decision and Control (CDC) . IEEE, 2021, pp. 1024–1030

    work page 2021

  8. [8]

    Robust disturbance rejection for robotic bipedal walking: System-level-synthesis with step-to-step dynamics approximation,

    X. Xiong, Y . Chen, and A. D. Ames, “Robust disturbance rejection for robotic bipedal walking: System-level-synthesis with step-to-step dynamics approximation,” in 2021 60th IEEE Conference on Decision and Control (CDC) . IEEE, 2021, pp. 697–704

    work page 2021

Show all 33 references
  1. [9]

    3-d underactuated bipedal walking via h-lip based gait synthesis and stepping stabilization,

    X. Xiong and A. Ames, “3-d underactuated bipedal walking via h-lip based gait synthesis and stepping stabilization,” IEEE Transactions on Robotics, vol. 38, no. 4, pp. 2405–2425, 2022

    2022

  2. [10]

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

    Z. Li, X. B. Peng, P. Abbeel, S. Levine, G. Berseth, and K. Sreenath, “Reinforcement learning for versatile, dynamic, and robust bipedal locomotion control,” The International Journal of Robotics Research , p. 02783649241285161, 2024

    2024

  3. [11]

    Tailoring solution accuracy for fast whole-body model predictive control of legged robots,

    C. Khazoom, S. Hong, M. Chignoli, E. Stanger-Jones, and S. Kim, “Tailoring solution accuracy for fast whole-body model predictive control of legged robots,” IEEE Robotics and Automation Letters , 2024

    2024

  4. [12]

    Gait-net-augmented implicit kino-dynamic mpc for dynamic variable-frequency humanoid locomo- tion over discrete terrains,

    J. Li, Z. Duan, J. Ma, and Q. Nguyen, “Gait-net-augmented implicit kino-dynamic mpc for dynamic variable-frequency humanoid locomo- tion over discrete terrains,” arXiv preprint arXiv:2502.02934 , 2025

    2025 arXiv

  5. [13]

    Falling motion control for humanoid robots while walking,

    K. Ogata, K. Terada, and Y . Kuniyoshi, “Falling motion control for humanoid robots while walking,” in 2007 7th IEEE-RAS International Conference on Humanoid Robots . IEEE, 2007, pp. 306–311

    2007

  6. [14]

    Unified multi-contact fall mitigation planning for humanoids via contact transition tree optimization,

    S. Wang and K. Hauser, “Unified multi-contact fall mitigation planning for humanoids via contact transition tree optimization,” in 2018 IEEE-RAS 18th International Conference on Humanoid Robots (Hu- manoids). IEEE, 2018, pp. 1–9

    2018

  7. [15]

    Push recovery by stepping for humanoid robots with force controlled joints,

    B. J. Stephens and C. G. Atkeson, “Push recovery by stepping for humanoid robots with force controlled joints,” in 2010 10th IEEE- RAS International conference on humanoid robots . IEEE, 2010, pp. 52–59

    2010

  8. [16]

    Learning push recovery for a bipedal humanoid robot with dynamical movement primitives,

    D. Luo, X. Han, Y . Ding, Y . Ma, Z. Liu, and X. Wu, “Learning push recovery for a bipedal humanoid robot with dynamical movement primitives,” in 2015 IEEE-RAS 15th International Conference on Humanoid Robots (Humanoids) . IEEE, 2015, pp. 1013–1019

    2015

  9. [17]

    On the emergence of whole- body strategies from humanoid robot push-recovery learning,

    D. Ferigo, R. Camoriano, P. M. Viceconte, D. Calandriello, S. Traver- saro, L. Rosasco, and D. Pucci, “On the emergence of whole- body strategies from humanoid robot push-recovery learning,” IEEE Robotics and Automation Letters , vol. 6, no. 4, pp. 8561–8568, 2021

    2021

  10. [18]

    Seamless reaction strategy for bipedal locomotion exploiting real-time nonlinear model predictive control,

    J. Choe, J.-H. Kim, S. Hong, J. Lee, and H.-W. Park, “Seamless reaction strategy for bipedal locomotion exploiting real-time nonlinear model predictive control,” IEEE Robotics and Automation Letters , pp. 1–8, 2023

    2023

  11. [19]

    Approximate hybrid model predictive control for multi-contact push recovery in complex environments,

    T. Marcucci, R. Deits, M. Gabiccini, A. Bicchi, and R. Tedrake, “Approximate hybrid model predictive control for multi-contact push recovery in complex environments,” in 2017 IEEE-RAS 17th interna- tional conference on humanoid robotics (Humanoids) . IEEE, 2017, pp. 31–38

    2017

  12. [20]

    Robust humanoid contact planning with learned zero-and one-step capturability prediction,

    Y .-C. Lin, L. Righetti, and D. Berenson, “Robust humanoid contact planning with learned zero-and one-step capturability prediction,” IEEE Robotics and Automation Letters , vol. 5, no. 2, pp. 2451–2458, 2020

    2020

  13. [21]

    Humanoid arm motion planning for im- proved disturbance recovery using model hierarchy predictive control,

    C. Khazoom and S. Kim, “Humanoid arm motion planning for im- proved disturbance recovery using model hierarchy predictive control,” in 2022 International Conference on Robotics and Automation (ICRA). IEEE, 2022, pp. 6607–6613

    2022

  14. [22]

    Dynamic locomotion in the mit cheetah 3 through convex model-predictive control,

    J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim, “Dynamic locomotion in the mit cheetah 3 through convex model-predictive control,” in 2018 IEEE/RSJ international conference on intelligent robots and systems (IROS) . IEEE, 2018, pp. 1–9

    2018

  15. [23]

    Mujoco: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,” in 2012 IEEE/RSJ international conference on intelligent robots and systems . IEEE, 2012, pp. 5026–5033

    2012

  16. [24]

    Dynamic walking: Toward agile and efficient bipedal robots,

    J. Reher and A. D. Ames, “Dynamic walking: Toward agile and efficient bipedal robots,” Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, no. 1, pp. 535–572, 2021

    2021

  17. [25]

    Orientation- aware model predictive control with footstep adaptation for dynamic humanoid walking,

    Y . Ding, C. Khazoom, M. Chignoli, and S. Kim, “Orientation- aware model predictive control with footstep adaptation for dynamic humanoid walking,” in 2022 IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids) . IEEE, 2022, pp. 299–305

    2022

  18. [26]

    A model predictive capture point control framework for robust humanoid balancing via ankle, hip, and stepping strategies,

    M.-J. Kim, D. Lim, G. Park, and J. Park, “A model predictive capture point control framework for robust humanoid balancing via ankle, hip, and stepping strategies,” arXiv preprint arXiv:2307.13243 , 2023

    2023 arXiv

  19. [27]

    Robust bipedal walking with closed-loop mpc: Adios stabilizers,

    A. Dallard, M. Benallegue, N. Scianca, F. Kanehiro, and A. Kheddar, “Robust bipedal walking with closed-loop mpc: Adios stabilizers,” 2024

    2024

  20. [28]

    Fall prediction for bipedal robots: The standing phase,

    M. E. Mungai, G. Prabhakaran, and J. W. Grizzle, “Fall prediction for bipedal robots: The standing phase,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 13 135–13 141

    2024

  21. [29]

    Weighted line fitting algorithms for mobile robot map building and efficient data representation,

    S. T. Pfister, S. I. Roumeliotis, and J. W. Burdick, “Weighted line fitting algorithms for mobile robot map building and efficient data representation,” in 2003 IEEE International Conference on Robotics and Automation (Cat. No. 03CH37422) , vol. 1. IEEE, 2003, pp. 1304–1311

    2003

  22. [30]

    Featherstone, Rigid body dynamics algorithms

    R. Featherstone, Rigid body dynamics algorithms . Springer, 2014

    2014

  23. [31]

    qpoases: A parametric active-set algorithm for quadratic program- ming,

    H. J. Ferreau, C. Kirches, A. Potschka, H. G. Bock, and M. Diehl, “qpoases: A parametric active-set algorithm for quadratic program- ming,” Mathematical Programming Computation, vol. 6, pp. 327–363, 2014

    2014

  24. [32]

    Dynamic walking on highly underactuated point foot humanoids: Closing the loop between hzd and hlip,

    A. B. Ghansah, J. Kim, K. Li, and A. D. Ames, “Dynamic walking on highly underactuated point foot humanoids: Closing the loop between hzd and hlip,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2024, pp. 12 686– 12 693

    2024

  25. [33]

    Massively multi-robot simulation in stage,

    R. Vaughan, “Massively multi-robot simulation in stage,” Swarm intelligence, vol. 2, pp. 189–208, 2008

    2008

Pith tools

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