Pith. sign in

REVIEW 3 major objections 5 minor 38 references

Dynamic Bipedal MPC with Foot-level Obstacle Avoidance and Adjustable Step Timing

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 This paper presents a real-time MPC framework that lets a bipedal robot dodge fast-moving obstacles and, in the same optimization, choose whether to step over or step around foot-level obstacles.

desk verdict A well-engineered MPC extension for bipedal body and foot avoidance with a clever adaptive step-time heuristic, but the foot-safety guarantee leans on an unverified tracking margin and the validation is mostly qualitative. read the letter →

arxiv 2505.13715 v1 pith:LYJNEEWQ submitted 2025-05-19 cs.RO cs.SYeess.SY

Tianze Wang , Christian Hubicki This is my paper · ORCID
classification cs.ROcs.SYeess.SY
keywords ModelPredictiveControlbipedalwalkingobstacleavoidanceMixed-IntegerQuadraticProgrammingadaptivesteptimingfootplacementLinearInvertedPendulum
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 one real-time MPC can handle both body-level and foot-level collision avoidance for dynamic bipedal walking, and that the foot-avoidance layer is the first model-based approach to implicitly choose between stepping over and stepping around a ground obstacle based on the state of the center-of-mass dynamics. The practical payoff is that a walking robot no longer needs a hand-tuned step frequency or a separate footstep planner: the optimizer adjusts step timing, selects a foothold region, and shapes the swing-foot height profile together. The authors back the claim with multibody simulations on Cassie and Digit and with Digit hardware experiments showing a step-over maneuver. A sympathetic reader would take the central contribution to be the integration: lifting non-convex foot-avoidance into a convex MIQP within a LIPM-based MPC while keeping real-time rates.

What carries the argument

The load-bearing object is the LIPM-based MPC with binaries: state variables are pelvis and foot trajectories; costs penalize velocity tracking at step transitions, nominal foot-to-pelvis distance, foot effort, and slack variables for avoidance; constraints include the LIPM dynamics, half-space body-avoidance constraints, and, for feet, a big-M formulation that selects exactly one of four convex ground quadrants around each obstacle. A soft minimum-travel-distance constraint, encoded with auxiliary positive/negative parts and binaries, forces the prediction to keep moving so the optimizer cannot stall behind an obstacle. Step timing is handled not as a continuous variable but as discrete choices at 0.1 s intervals, switched by a heuristic based on predicted foothold tracking error.

What would settle it

Instrument Digit's swing foot during a step-over of a 0.2 m obstacle with the controller's 0.4 m padded region, and log the minimum vertical and lateral clearance over the obstacle across many trials. If measured clearance ever falls below roughly 0.1 m, the padding on each side, while the MPC still outputs a step-over plan, then the simplified straight-line swing model does not guarantee the advertised 3D avoidance.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that a Linear Inverted Pendulum Model (LIPM) based MPC can be extended to plan 3D swing-foot avoidance by decomposing the ground around each obstacle into four convex quadrants, selecting one quadrant with binary variables in a Mixed-Integer Quadratic Program, and imposing height constraints along the straight line connecting the current and selected foothold. The same MPC also adjusts step timing: a tracking-error heuristic shortens the next stance phase when the foothold error predicts that staying on that foot would block a dodge, and the collision region is modeled as a velocity-stretched ellipse so the avoidance direction accounts for where the obstacle is going. The authors state that, to their knowledge, this is the first documented model-based approach for 3D dynamic bipedal walking that implicitly chooses to step over or navigate around foot-level obstacles depending on the state of its CoM dynamics.

Load-bearing premise

The 3D foot-avoidance guarantee relies on the lower-level controller tracking the planned straight-line swing path closely enough that the padded collision region absorbs all tracking error, since the MPC itself never models the swing foot's dynamics or kinematics.

Editorial extensions

If this is right

  • A robot using this controller can dodge a 5 m/s obstacle by shortening its next stance phase to 0.2 s instead of 0.4 s, as shown in Cassie simulation.
  • The same MPC will step around at low commanded speed and step over at higher speed, because the MIQP region choice trades travel distance against avoidance cost.
  • Foot-level obstacle negotiation runs in real time: MPC at 50 Hz with MIQP and 200 Hz otherwise, on Digit hardware.
  • The controller requires no a priori switch between 'step over' and 'step around' behaviors; the behavior emerges from the optimization.
  • The three-dimensional avoidance guarantee is tied to the straight-line swing reference, so the method's clearance can be assessed directly from the padded obstacle region.

Reading between the lines

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

  • If OSC tracking error stays within the padded margin, the straight-line swing reference makes the 3D clearance guarantee largely a matter of respecting height constraints; a natural extension is replacing the straight line with a spline or adding swing-foot kinematics directly into the MPC, which the paper lists as future work.
  • The velocity-stretched ellipse could be reused for moving obstacles with uncertain trajectories by inflating the ellipse with the obstacle's speed and heading, though the paper only demonstrates it with a known linear motion.
  • The tracking-error heuristic for step timing might transfer to other bipedal tasks that need fast weight transfer, such as push recovery, if the same error signal can be computed there.
  • The convex-quadrant decomposition could be generalized to multiple ground obstacles with a more advanced decomposition method, as the paper notes, which would extend the same MIQP mechanism to cluttered terrains.
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 presents a real-time MPC framework for dynamic bipedal robots that addresses both body-level and foot-level obstacle avoidance. For body avoidance, the paper extends a prior half-space relaxation formulation with an ellipse-based collision region that incorporates obstacle velocity, and it introduces a heuristic that shortens the upcoming stance duration when the MPC's predicted foothold tracking error exceeds a threshold. For foot avoidance, the ground around a rectangular obstacle is decomposed into four convex quadrants, a Mixed-Integer Quadratic Program selects which quadrant the landing foothold must lie in, and step-over height constraints are imposed along the straight line connecting the swing foot's start and end positions. A soft minimum-travel-distance constraint is added to prevent stalling in local minima. The framework is evaluated in multibody simulations on Cassie and Digit, and in a single hardware experiment on Digit where the robot steps over a 0.2 m obstacle.

Significance. If the claims hold, the framework would be a useful contribution to reactive bipedal locomotion: it is, to the authors' knowledge, the first model-based 3D dynamic bipedal walking approach that implicitly chooses to step over or navigate around foot-level obstacles based on CoM dynamics, and the adaptive step-timing heuristic is a plausible mechanism for faster dodging. The formulation is mostly convex with binary variables, the use of an existing Operational Space Controller for tracking is appropriate, and the paper demonstrates the complete pipeline in simulation and on Digit hardware. However, the validation is largely qualitative, and the central foot-avoidance safety claim rests on an unverified assumption about low-level tracking accuracy. The paper would be significantly strengthened by quantitative tracking-error data, repeated trials, and a clearer statement of the conditions under which the foot-avoidance guarantee holds.

major comments (3)
  1. [Sec. II-E.1, Eqs. (5)-(7)] The x-y plane foot-avoidance constraints only enforce that the swing foot's start and end footholds lie in the selected convex quadrant; the swing trajectory is then taken as the straight line between these endpoints, and the height constraints of Eq. (7) are imposed only along that line. Consequently, the 3D collision-free guarantee depends on the low-level OSC tracking this straight-line reference with an error smaller than the padded collision region diameter (0.4 m for a 0.2 m obstacle, Sec. IV.A.4). This is a load-bearing assumption, but the paper provides no tracking-error data, no perturbation or Monte Carlo study, and only a single qualitative hardware step-over. Since the authors list 'introducing swing foot kinematics/dynamics via nonlinear MPC' as future work in Sec. V, the current formulation itself omits swing-foot kinematics. I recommend providing quantitative tracking-error measurements during swing (e.g., lateral deviation under the OSC), or a formal bound on the required padding, or clearly softening the central claim to 'demonstrated on the tested configuration.'
  2. [Sec. III-C and Sec. IV.A.4] The experimental validation is thin relative to the strength of the claims. Cassie results in Fig. 6 are reported only as success or failure, with no quantitative metrics such as minimum distance to the obstacle, CoM tracking error, or the actual step-timing adjustments. The Digit hardware result in Fig. 7B is a single successful step-over, with no reporting of commanded versus actual foot/CoM trajectories, obstacle clearance, or repeated trials. The paper claims that the framework 'implicitly chooses to step over or navigate around' obstacles and that adaptive step timing improves avoidance; without systematic experiments with multiple obstacle sizes, speeds, placements, and a baseline without adaptive timing and without MIQP, these claims cannot be distinguished from a tuned demonstration. I suggest adding repeated trials and quantitative comparison metrics.
  3. [Sec. II-D.2] The adaptive step-time heuristic is derived from the authors' own MPC behavior: the switching threshold is based on the MPC's predicted foothold tracking error, and the decision rule (shorten the next stance phase by 0.1 s when the error exceeds a threshold) is not accompanied by any analysis of when this is guaranteed to improve avoidance or how the threshold was selected. The paper reports no sensitivity analysis for the threshold, minimum stance duration, or discrete step-time interval. This is a design choice rather than a logical error, but it weakens the claim that the framework reduces reliance on human operators, since the heuristic introduces new hand-tuned parameters.
minor comments (5)
  1. [Sec. II-B] In the general MPC formulation, the symbols q, u, rho, and b are used before being defined; please define them in that section or refer explicitly to the later subsection where they are introduced.
  2. [Sec. II-E.2, Eq. (11)] The minimum-travel-distance constraint depends on gains K_x and K_y, but the paper does not state how these are chosen or whether they are tuned separately for Cassie and Digit. Please report the values used in the experiments.
  3. [Sec. IV.A.2] The 'small offset term on the order of 0.01' added to the foot position is not specified with units, and it is unclear whether it is a fixed offset or state-dependent. Please clarify.
  4. [Sec. IV.A.1] The paper states that the MPC is updated at 50 Hz for MIQP and 200 Hz otherwise, but it does not report solver solve times or whether the real-time deadlines were consistently met. Please provide this information.
  5. [Sec. I-A] The related-work discussion mentions [28] as using spline-based swing trajectories, but it does not explicitly state how the present formulation differs with respect to swing-foot optimization. A sentence clarifying that distinction would help position the contribution.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the MPC extensions are self-contained and validated against baseline and hardware.

full rationale

The paper does not fit parameters to target outcomes and then repackage them as predictions. The base MPC formulation is inherited from the authors' prior work [18], but the paper explicitly extends it (Sec. II-C: 'we no longer regularize all nodes to track a target velocity...'), and the claimed contributions are the new ellipse-based half-space, the adaptive step-time heuristic, and the MIQP 3D foot-avoidance formulation. The adaptive step-time rule is a heuristic derived from the MPC's own predicted foothold-tracking error (Sec. II-D.2, Fig. 4B); this is a control-law design choice, not a fitted parameter relabeled as a prediction, because the success of the controller is evaluated by separate simulation and hardware avoidance outcomes against the original formulation. The MIQP foot avoidance (Eqs. 5-7) selects footholds via binary variables and imposes height constraints along the straight-line swing reference, so the claim that the controller 'implicitly chooses to step over or navigate around' obstacles is a description of the optimization's behavior rather than a conclusion equivalent to its inputs. The reliance on [18] for the baseline controller and the OSC is a normal self-citation and is not load-bearing for the new contributions, which are tested against a baseline and on Digit hardware. The admitted omission of swing-foot kinematics from the optimization (Sec. V: 'introducing swing foot kinematics/dynamics via nonlinear MPC') is a robustness and validation limitation of the foot-avoidance guarantee, not a circular derivation. No step in the derivation chain reduces by construction to its own input or to a fitted quantity renamed as a prediction.

Assumptions & free parameters 11 free parameters · 5 assumptions · 0 invented entities

The central claim rests on the LIPM reduced-order model, the ellipse model of obstacle motion, and a set of hand-tuned MPC costs and thresholds. The most consequential simplifications are the straight-line swing-foot trajectory assumption and the use of predicted foot-tracking error as a step-timing signal; both are acknowledged to be heuristics rather than derived guarantees.

free parameters (11)
  • Velocity tracking weight Wstate = not reported
    Hand-tuned weight penalizing deviation of step-transition velocity from reference; central to the MPC cost in Eq. (2).
  • Foothold tracking weight Wfoot = not reported
    Hand-tuned weight penalizing deviation of foothold from nominal foot-to-pelvis distance in Eq. (2).
  • Control effort weight Weffort = not reported
    Hand-tuned weight in Eq. (3) that ties foothold to the control input to account for leg motor authority.
  • Slack weight Wslack = not reported
    Hand-tuned weight on slack variables in the general MPC of Sec. II-B.
  • Foot avoidance weight Wfoot,avoid = not reported
    Hand-tuned weight penalizing foot avoidance slack in Eq. (8).
  • Minimum travel distance weight Wdis = not reported
    Hand-tuned weight on the travel-distance slack in Eq. (10), used to prevent stalling.
  • Directional gains K_x, K_y = not reported
    Tunable gains in Eq. (11) that prioritize movement in specific directions.
  • Safe distance r = not reported (set based on obstacle size)
    Sets the collision region diameter; affects both body and foot avoidance constraints.
  • Step-time adjustment threshold = not reported
    Threshold on predicted foot-tracking error in Sec. II-D.2 that triggers a 0.1 s reduction of the next stance phase; hand-chosen.
  • Minimum stance duration = not reported
    Lower bound on step time in the adaptive-step-time heuristic; hand-chosen.
  • Obstacle collision region diameter = 0.4 m for a 0.2 m obstacle in hardware
    Padded obstacle size to account for Digit foot length; chosen by hand.
assumptions (5)
  • domain assumption Linear Inverted Pendulum Model accurately captures CoM dynamics for foot placement and avoidance planning.
    Sec. II-C uses the LIPM analytical solution (Eq. 1); the controller relies on this reduced-order model for all MPC predictions.
  • domain assumption The obstacle's future swept path can be represented by an ellipse whose axis lengths depend on current position and velocity.
    Sec. II-D.1 models the collision region as an ellipse stretched by obstacle velocity; no obstacle state prediction or uncertainty model is included.
  • domain assumption A straight-line swing-foot trajectory with height constraints at each MPC node guarantees collision-free foot motion.
    Sec. II-E.1 connects endpoints with a straight line and applies height constraints; the lower-level controller is assumed to track this reference accurately.
  • ad hoc to paper Predicted foot-tracking error from the MPC is a reliable signal for when to shorten the next stance phase.
    Sec. II-D.2 uses this heuristic, supported only by offline MPC runs, not by a formal derivation or hardware validation.
  • ad hoc to paper The ground safe region around an obstacle is adequately represented by four convex quadrants for the single-obstacle case.
    Sec. II-E.1 decomposes the ground into four quadrants; extension to multiple obstacles is left to future work.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Dynamic Bipedal MPC with Foot-level Obstacle Avoidance and Adjustable Step Timing." pith.science (2026). https://pith.science/paper/LYJNEEWQ

@misc{pith2026250513715,
  author       = {Pith},
  title        = {Pith review of: Dynamic Bipedal MPC with Foot-level Obstacle Avoidance and Adjustable Step Timing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LYJNEEWQ}},
  note         = {Machine review of arXiv:2505.13715}
}
read the original abstract

Collision-free planning is essential for bipedal robots operating within unstructured environments. This paper presents a real-time Model Predictive Control (MPC) framework that addresses both body and foot avoidance for dynamic bipedal robots. Our contribution is two-fold: we introduce (1) a novel formulation for adjusting step timing to facilitate faster body avoidance and (2) a novel 3D foot-avoidance formulation that implicitly selects swing trajectories and footholds that either steps over or navigate around obstacles with awareness of Center of Mass (COM) dynamics. We achieve body avoidance by applying a half-space relaxation of the safe region but introduce a switching heuristic based on tracking error to detect a need to change foot-timing schedules. To enable foot avoidance and viable landing footholds on all sides of foot-level obstacles, we decompose the non-convex safe region on the ground into several convex polygons and use Mixed-Integer Quadratic Programming to determine the optimal candidate. We found that introducing a soft minimum-travel-distance constraint is effective in preventing the MPC from being trapped in local minima that can stall half-space relaxation methods behind obstacles. We demonstrated the proposed algorithms on multibody simulations on the bipedal robot platforms, Cassie and Digit, as well as hardware experiments on Digit.

Figures

Figures reproduced from arXiv: 2505.13715 by the authors.

Figure 1
Figure 1. (A) Bipedal robots must avoid dynamic and foot￾level obstacles in their environment. (B) A Digit humanoid using model-predictive control (MPC) to step over foot-level obstacles. A. Related Work Collision-free planning is an extensively researched area across a wide range of robotic platforms including mobile robots [2], [3], drones [4], [5], and legged robots [6]–[11]. For legged robots, safe navigation is particula… view at source ↗
Figure 2
Figure 2. Block diagram for the presented controller. [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. (A) A LIPM-based MPC is implemented to generate collision-free trajectories. (B) The approximation is im￾proved by linearizing constraints around a trajectory from the previous MPC solution. (C) We used a half-space relaxation with slack variables to model the collision-free regions [18]. MPC can select an avoidance path that is optimal not only for the obstacle’s current position but also for its future trajectory.… view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Illustration of our swing foot 3D obstacle avoidance: [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]
Figure 7
Figure 7. Figure 7: (A) A simulation of Digit stepping around and over an obstacle respectively depending on its commanded speed. (B) Digit achieves 3D swing foot avoidance by stepping over the obstacle on hardware. 4) Hardware Result: Digit Foot Avoidance: Here we show the Digit hardware…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 27 canonical work pages

  1. [28]

    Bipedal walking on constrained footholds with mpc footstep control,

    B. Acosta and M. Posa, “Bipedal walking on constrained footholds with mpc footstep control,” in2023 IEEE-RAS 22nd International Conference on Humanoid Robots (Humanoids). IEEE, 2023, pp. 1– 8

    work page 2023

  2. [1]

    Optimization-based control for dynamic legged robots,

    P. M. Wensing, M. Posa, Y . Hu, A. Escande, N. Mansard, and A. Del Prete, “Optimization-based control for dynamic legged robots,” IEEE Transactions on Robotics, 2023

    2023

  3. [2]

    Control of mobile robots using barrier functions under temporal logic specifications,

    M. Srinivasan and S. Coogan, “Control of mobile robots using barrier functions under temporal logic specifications,”IEEE Transactions on Robotics, vol. 37, no. 2, pp. 363–374, 2020

    work page 2020

  4. [3]

    Safe navigation in human occupied environments using sampling and control barrier functions,

    K. Majd, S. Yaghoubi, T. Yamaguchi, B. Hoxha, D. Prokhorov, and G. Fainekos, “Safe navigation in human occupied environments using sampling and control barrier functions,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 5794–5800

    work page 2021

  5. [4]

    Avoiding dynamic obstacles with real-time motion planning using quadratic programming for varied locomotion modes,

    J. White, D. Jay, T. Wang, and C. Hubicki, “Avoiding dynamic obstacles with real-time motion planning using quadratic programming for varied locomotion modes,” in2022 IEEE/RSJ International Con- ference on Intelligent Robots and Systems (IROS), 2022, pp. 13 626– 13 633

    work page 2022

  6. [5]

    Motion planning around obstacles with convex optimization,

    T. Marcucci, M. E. Petersen, D. von Wrangel, and R. Tedrake, “Motion planning around obstacles with convex optimization,”ArXiv, vol. abs/2205.04422, 2022

    arXiv 2022

  7. [6]

    Collision- free mpc for legged robots in static and dynamic scenes,

    M. Gaertner, M. Bjelonic, F. Farshidian, and M. Hutter, “Collision- free mpc for legged robots in static and dynamic scenes,” in2021 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2021, pp. 8266–8272

    work page 2021

  8. [7]

    Autonomous navigation for quadrupedal robots with optimized jumping through constrained obstacles,

    S. Gilroy, D. Lau, L. Yang, E. Izaguirre, K. Biermayer, A. Xiao, M. Sun, A. Agrawal, J. Zeng, Z. Li,et al., “Autonomous navigation for quadrupedal robots with optimized jumping through constrained obstacles,” in2021 IEEE 17th International Conference on Automation Science and Engineering (CASE). IEEE, 2021, pp. 2132–2139

    work page 2021

Show all 38 references
  1. [8]

    Multi-layered safety for legged robots via control barrier functions and model predictive control,

    R. Grandia, A. Taylor, A. Ames, and M. Hutter, “Multi-layered safety for legged robots via control barrier functions and model predictive control,” 05 2021, pp. 8352–8358

    2021

  2. [9]

    Efficient anytime clf reactive planning system for a bipedal robot on undulating terrain,

    J.-K. Huang and J. W. Grizzle, “Efficient anytime clf reactive planning system for a bipedal robot on undulating terrain,”IEEE Transactions on Robotics, pp. 1–18, 2023

    2023

  3. [10]

    Socially acceptable bipedal robot navigation via social zonotope network model predictive control,

    A. Shamsah, K. Agarwal, N. Katta, A. Raju, S. Kousik, and Y . Zhao, “Socially acceptable bipedal robot navigation via social zonotope network model predictive control,”arXiv preprint arXiv:2406.17151, 2024

    2024 arXiv

  4. [11]

    Real-time path planning in unknown environments for bipedal robots,

    A.-C. Hildebrandt, M. Klischat, D. Wahrmann, R. Wittmann, F. Sygulla, P. Seiwald, D. Rixen, and T. Buschmann, “Real-time path planning in unknown environments for bipedal robots,”IEEE Robotics and Automation Letters, vol. 2, no. 4, pp. 1856–1863, 2017

    2017

  5. [12]

    The 3d linear inverted pendulum mode: a simple modeling for a biped walking pattern generation,

    S. Kajita, F. Kanehiro, K. Kaneko, K. Yokoi, and H. Hirukawa, “The 3d linear inverted pendulum mode: a simple modeling for a biped walking pattern generation,” inProceedings 2001 IEEE/RSJ International Conference on Intelligent Robots and Systems, vol. 1, 2001, pp. 239–246

    2001

  6. [13]

    One-step ahead prediction of angular mo- mentum about the contact point for control of bipedal locomotion: Validation in a lip-inspired controller,

    Y . Gong and J. Grizzle, “One-step ahead prediction of angular mo- mentum about the contact point for control of bipedal locomotion: Validation in a lip-inspired controller,” in2021 IEEE International Conference on Robotics and Automation (ICRA), 2021, pp. 2832– 2838

    2021

  7. [14]

    The spring-mass model for running and hopping,

    R. Blickhan, “The spring-mass model for running and hopping,” Journal of Biomechanics, 1989

    1989

  8. [15]

    Vision-aided autonomous navigation of bipedal robots in height-constrained environments,

    Z. Li, J. Zeng, S. Chen, and K. Sreenath, “Vision-aided autonomous navigation of bipedal robots in height-constrained environments,” ArXiv, vol. abs/2109.05714, 2021

    2021 arXiv

  9. [16]

    Risk-averse control via cvar barrier functions: Application to bipedal robot locomotion,

    M. Ahmadi, X. Xiong, and A. D. Ames, “Risk-averse control via cvar barrier functions: Application to bipedal robot locomotion,”IEEE Control Systems Letters, vol. 6, pp. 878–883, 2022

    2022

  10. [17]

    Mpc- based humanoid pursuit-evasion in the presence of obstacles,

    D. De Simone, N. Scianca, P. Ferrari, L. Lanari, and G. Oriolo, “Mpc- based humanoid pursuit-evasion in the presence of obstacles,” in2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017, pp. 5245–5250

    2017

  11. [18]

    Real-time dynamic bipedal avoidance,

    T. Wang, J. White, and C. Hubicki, “Real-time dynamic bipedal avoidance,” in2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2023, pp. 8566–8573

    2023

  12. [19]

    A sequential mpc approach to reactive planning for bipedal robots using safe corridors in highly cluttered environments,

    K. S. Narkhede, A. M. Kulkarni, D. A. Thanki, and I. Poulakakis, “A sequential mpc approach to reactive planning for bipedal robots using safe corridors in highly cluttered environments,”IEEE Robotics and Automation Letters, vol. 7, no. 4, pp. 11 831–11 838, 2022

    2022

  13. [20]

    Walking control based on step timing adaptation,

    M. Khadiv, A. Herzog, S. A. A. Moosavian, and L. Righetti, “Walking control based on step timing adaptation,”IEEE Transactions on Robotics, vol. 36, no. 3, pp. 629–643, 2020

    2020

  14. [21]

    Adapting gait frequency for posture-regulating humanoid push-recovery via hierarchical model predictive control,

    J. Li, Z. Le, J. Ma, and Q. Nguyen, “Adapting gait frequency for posture-regulating humanoid push-recovery via hierarchical model predictive control,”arXiv preprint arXiv:2409.14342, 2024

    2024 arXiv

  15. [22]

    Kinematic optimization for bipedal robots: a framework for real-time collision avoidance,

    A.-C. Hildebrandt, S. Schwerd, R. Wittmann, D. Wahrmann, F. Sygulla, P. Seiwald, D. Rixen, and T. Buschmann, “Kinematic optimization for bipedal robots: a framework for real-time collision avoidance,”Autonomous Robots, vol. 43, pp. 1187–1205, 2019

    2019

  16. [23]

    A generic optimization-based framework for reactive collision avoidance in bipedal locomotion,

    C. Zhou, C. Fang, X. Wang, Z. Li, and N. Tsagarakis, “A generic optimization-based framework for reactive collision avoidance in bipedal locomotion,” in2016 IEEE International Conference on Au- tomation Science and Engineering (CASE), 2016, pp. 1026–1033

    2016

  17. [24]

    Discrete control barrier functions for safety-critical control of discrete systems with application to bipedal robot navigation,

    A. Agrawal and K. Sreenath, “Discrete control barrier functions for safety-critical control of discrete systems with application to bipedal robot navigation,” inRobotics: Science and Systems, 2017. [Online]. Available: https://api.semanticscholar.org/CorpusID:1780280

    2017

  18. [25]

    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

  19. [26]

    Footstep planning for autonomous walking over rough ter- rain,

    R. J. Griffin, G. Wiedebach, S. McCrory, S. Bertrand, I. Lee, and J. Pratt, “Footstep planning for autonomous walking over rough ter- rain,” in2019 IEEE-RAS 19th International Conference on Humanoid Robots (Humanoids), 2019, pp. 9–16

    2019

  20. [27]

    Handling non-convex constraints in mpc-based humanoid gait gen- eration,

    A. S. Habib, F. M. Smaldone, N. Scianca, L. Lanari, and G. Oriolo, “Handling non-convex constraints in mpc-based humanoid gait gen- eration,” in2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2022, pp. 13 167–13 173

    2022

  21. [29]

    Footstep planning on uneven terrain with mixed-integer convex optimization,

    R. Deits and R. Tedrake, “Footstep planning on uneven terrain with mixed-integer convex optimization,” in2014 IEEE-RAS International Conference on Humanoid Robots, 2014, pp. 279–286

    2014

  22. [30]

    Solving footstep planning as a feasibility problem using l1-norm minimization,

    D. Song, P. Fernbach, T. Flayols, A. Del Prete, N. Mansard, S. Ton- neau, and Y . J. Kim, “Solving footstep planning as a feasibility problem using l1-norm minimization,”IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 5961–5968, 2021

    2021

  23. [31]

    Fast online trajectory optimization for the bipedal robot cassie,

    T. Apgar, P. Clary, K. R. Green, A. Fern, and J. W. Hurst, “Fast online trajectory optimization for the bipedal robot cassie,”Robotics: Science and Systems XIV, 2018

    2018

  24. [32]

    Gurobi Optimizer Reference Manual,

    Gurobi Optimization, LLC, “Gurobi Optimizer Reference Manual,”

  25. [33]

    Mujoco: A physics engine for model-based control,

    E. Todorov, T. Erez, and Y . Tassa, “Mujoco: A physics engine for model-based control,” in2012 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2012, pp. 5026–5033

    2012

  26. [34]

    Contact- aided invariant extended kalman filtering for robot state estimation,

    R. Hartley, M. Ghaffari, R. M. Eustice, and J. W. Grizzle, “Contact- aided invariant extended kalman filtering for robot state estimation,” The International Journal of Robotics Research, vol. 39, no. 4, pp. 402–430, 2020

    2020

  27. [35]

    OSQP: an operator splitting solver for quadratic programs,

    B. Stellato, G. Banjac, P. Goulart, A. Bemporad, and S. Boyd, “OSQP: an operator splitting solver for quadratic programs,”Mathematical Programming Computation, vol. 12, no. 4, pp. 637–672, 2020. [Online]. Available: https://doi.org/10.1007/s12532-020-00179-2

    2020 doi

  28. [36]

    CasADi – A software framework for nonlinear optimization and optimal control,

    J. A. E. Andersson, J. Gillis, G. Horn, J. B. Rawlings, and M. Diehl, “CasADi – A software framework for nonlinear optimization and optimal control,”Mathematical Programming Computation, vol. 11, no. 1, pp. 1–36, 2019

    2019

  29. [37]

    Robotic operating system

    Stanford Artificial Intelligence Laboratory et al., “Robotic operating system.” [Online]. Available: https://www.ros.org

  30. [2024]

    Available: https://www.gurobi.com

    [Online]. Available: https://www.gurobi.com

Pith tools

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