A Layered Control Perspective on Legged Locomotion: Embedding Reduced Order Models via Hybrid Zero Dynamics

Aaron D. Ames; Adrian B. Ghansah; Max H. Cohen; Sergio A. Esteban

REVIEW 3 major objections 5 minor 37 references

A Layered Control Perspective on Legged Locomotion: Embedding Reduced Order Models via Hybrid Zero Dynamics

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read A stable periodic orbit of a reduced-order model of locomotion, embedded in the full-order dynamics as a hybrid zero dynamics manifold, yields an input-to-state stable periodic orbit of the full-order hybrid system.

desk verdict A useful conditional transfer theorem from ROM to FOM stability, but the key mismatch bound is unquantified and the simulation does not verify the theorem's assumptions. read the letter →

arxiv 2509.00294 v1 pith:DXBK65R3 submitted 2025-08-30 eess.SY cs.ROcs.SY

Sergio A. Esteban , Max H. Cohen , Adrian B. Ghansah , Aaron D. Ames This is my paper · ORCID

classification eess.SYcs.ROcs.SY MSC 93C3093D2570Q05

keywords reduced-ordermodelshybridzerodynamicsinput-to-statestabilitybipedallocomotionlinearinvertedpendulumsystemswithimpulseeffectsPoincarémapRES-CLF

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 tries to close the gap between two ways of designing walking gaits: fast, simple reduced-order models (ROMs) like the linear inverted pendulum, and full-order hybrid zero dynamics (HZD), which come with stability guarantees but require expensive offline trajectory optimization. The authors propose to embed a ROM's behavior into a zero dynamics manifold of the full-order robot, then prove that exponential stability of the ROM's periodic orbit implies a robust form of stability for the full-order hybrid system's periodic orbit. The robustness is input-to-state stability: bounded mismatches between the ROM and the true robot dynamics are rejected, so the full robot stays near a stable walking orbit as long as the mismatch is small enough. If true, this means a practitioner can synthesize a provably stable gait from a cheap ROM without running full-order trajectory optimization, and the same formal guarantees carry over to the actual robot.

What carries the argument

The central object is the zero dynamics manifold M = {(η, z) : η = ψ(z)}, with embedding ι(z) = (ψ(z), z); ψ lifts ROM behavior (e.g., step length commands from the HLIP controller) into desired actuated states of the full-order robot. Hybrid invariance of M is ensured by Lemma 1 (continuous invariance via a feedback law) and Lemma 2 (discrete invariance via a reset compatibility condition). The argument then transfers stability through two tools: the RES-CLF Vε drives the full-order dynamics to M, and the discrepancy d_k := Ω(z_k) − Ξ ∘ Q_cl ∘ Ξ^{-1}(z_k) between the HZD Poincaré map and the ROM acts as a bounded disturbance, making the zero dynamics E-ISS via a converse Lyapunov argument.

What would settle it

Simulate the 5-link biped with an increasingly inaccurate ROM (e.g., perturb the HLIP height z0 or mass by larger amounts) and compute the realized sequence d_k := Ω(z_k) − Ξ ∘ Q_cl ∘ Ξ^{-1}(z_k); if the empirical sup norm of d_k exceeds the Theorem 2 bound while the orbit still appears stable in the simulation, the stated sufficient condition is not the operative mechanism, whereas if stability is lost exactly when ∥d∥∞ crosses the predicted threshold, the theorem is confirmed.

Watch

Extended reading notes

Core claim

The main result, Theorem 2, states that if a ROM has an exponentially stable fixed point, and if the ROM's state space is diffeomorphic to the zero dynamics coordinates of the full-order hybrid system, then the full-order closed-loop hybrid system has a locally exponentially input-to-state stable (E-ISS) periodic orbit. The construction works by choosing a manifold M = {(η, z) : η = ψ(z)} that encodes the ROM behavior, rendering it hybrid invariant, and using an RES-CLF controller to drive the full-order dynamics to M. The discrepancy between the actual hybrid zero dynamics Poincaré map and the ROM, denoted d_k, is treated as a disturbance; Lemma 4 shows the zero dynamics are E-ISS, and the

Load-bearing premise

The model mismatch d_k between the true hybrid zero dynamics and the ROM must stay within the bound ρ from Theorem 2, but the paper never quantifies that bound or verifies it for the simulation; if the mismatch exceeds ρ, the ISS guarantee has no content.

Editorial extensions

If this is right

ROM-based gait synthesis, such as HLIP footstep planning, can inherit formal stability guarantees for the full-order robot without solving a full-order trajectory optimization problem.
The results justify using simple inverted-pendulum models as the core of layered control stacks: stable step-to-step dynamics in the ROM imply robust periodic walking in the full hybrid system, provided the model mismatch is small.
The E-ISS formulation quantifies robustness: disturbances that enter as bounded model mismatch are rejected with a linear-gain ISS bound, giving a design margin for terrains or perturbations.
The hybrid invariance conditions (Lemma 2) give concrete constraints on how the desired actuated behavior ψ must align with the impact map, which can guide the choice of virtual constraints in other bipedal systems.
The decomposition into actuated and unactuated coordinates means stability of the full state follows from stability of the zero dynamics plus the RES-CLF convergence to M, so modular controller design remains valid.

Reading between the lines

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

The paper leaves the bound on d_k unquantified; a natural next step would be to compute or estimate ρ for a given ROM-FOM pair, turning the existence result into a practical verification tool.
The framework is not limited to LIP-type ROMs: any ROM with an exponentially stable fixed point and a diffeomorphic state map could be embedded, so SLIP or angular-momentum-based templates may also yield certified full-order gaits.
Because the disturbance d_k is treated abstractly, the same proof structure could accommodate terrain variations or parameter changes as disturbances, potentially extending the guarantee to walking on mildly uneven ground.
The PD controller used in the simulation, rather than the theoretically required RES-CLF, suggests a gap between theory and implementation; closing this gap by formally treating re-planning as part of the disturbance would strengthen the practical claim.

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.

Desk Editor's Note

read the letter

Here's my take on arXiv:2509.00294. The core new idea is a transfer theorem: exponential stability of a ROM fixed point, combined with hybrid invariance of an embedding manifold and an RES-CLF, gives local E-ISS of the full-order hybrid orbit, provided the mismatch between the true HZD Poincaré map and the ROM map is uniformly bounded by some ρ. That's a genuinely useful framing for a community that routinely designs with LIP/SLIP and wants formal guarantees. Lemma 4 is a standard perturbation argument, but the whole package—manifold embedding, ISS disturbance model, and the specific theorem—is new and worth reading.

The paper does several things well. It constructs the manifold from a ROM via ψ and gives conditions for hybrid invariance. It is explicitly honest in Section IV-D that the implementation uses PD control and replanning rather than the RES-CLF controller and fixed manifold assumed by Theorem 2. Remark 2 admits that characterizing the discrepancy is hard. That honesty is appreciated.

The soft spot is the one the stress-test flags: the ρ in Theorem 2 is purely existential. The theorem says 'if the disturbance is small enough,' but the paper never quantifies how small is small enough, nor does it compute d_k in the simulation. The visual closeness of orbits in Fig. 3 is suggestive but doesn't establish the bound. As a result, the simulation does not demonstrate the theorem's conditions; it demonstrates that a heuristic controller works on the 5-link model. The abstract's phrasing—'a stable periodic orbit in the ROM implies an input-to-state stable periodic orbit of the FOM'—overstates the conditional result. Also, the RES-CLF existence and the discrete invariance condition (Lemma 2) are assumed, not verified for the example.

I want to be fair: the theorem itself looks correct under its assumptions. The gap is between the assumptions and the example. That is a standard issue in nonlinear control papers, but here the gap is load-bearing because the central promise is that ROM design transfers to the FOM. Without a bound on d_k, the promise is not fulfilled for the presented case study.

Who gets value from this? Theoreticians in legged locomotion who want a formal template for ROM-to-FOM stability transfer will find it useful. Practitioners should read it as a cautionary tale: the simulation shows what works, not a validation of the theorem.

I would send it to peer review. A good referee should demand a quantitative estimate or an a posteriori computation of d_k along the simulated trajectory, and a clearer statement of the theorem's conditions in the abstract. It's not desk-reject material, but it needs substantial revision.

Referee Report

3 major / 5 minor

Summary. The paper proposes a layered control framework that embeds a reduced-order model (ROM) inside the full-order hybrid dynamics of a legged robot via hybrid zero dynamics (HZD). The main theoretical contribution is Theorem 2, which states that if a ROM's discrete-time closed-loop map has an exponentially stable fixed point, and if the manifold encoding the ROM is hybrid invariant and possesses a rapidly exponentially stabilizing control Lyapunov function (RES-CLF), then, for sufficiently small controller gain parameter epsilon and sufficiently small model mismatch rho, the corresponding embedded periodic orbit is locally exponentially input-to-state stable (E-ISS) for the full-order closed-loop hybrid system. The proof proceeds through a perturbation argument (Lemma 4) treating the difference between the HZD Poincare map and the ROM map in z-coordinates as a disturbance. The result is illustrated in simulation on a 5-link planar biped with a hybrid linear inverted pendulum (HLIP) ROM. The paper also includes design conditions for manifold invariance and a concrete construction of the embedding for the biped example.

Significance. If the result holds in the stated conditional form, it is a valuable formal bridge between ROM-based gait synthesis and full-order stability guarantees, potentially reducing reliance on expensive offline trajectory optimization. The proof is self-contained, uses standard converse Lyapunov and ISS tools, and does not rely on fitting or numeric optimization for the theorem. The paper is honest in Remark 2 about the difficulty of characterizing the model mismatch and in Section IV-D about implementation deviations. However, the central applicability to the example is not established because the key small-mismatch bound on d_k is never quantified, and the simulation's controller violates the theorem's RES-CLF and fixed-manifold assumptions. The abstract and conclusion overstate the theorem by omitting these conditions. Thus the theoretical contribution is sound but its demonstrated scope is narrower than claimed.

major comments (3)

[Eq. (31), Lemma 4, Theorem 2, Remark 2] Theorem 2 is explicitly conditional on the existence of rho>0 such that the model mismatch d_k defined in Eq. (31) satisfies ||d||_infinity <= rho along the relevant trajectories. Lemma 4 only produces rho existentially via converse Lyapunov theorems; no quantitative estimate of rho or of the neighborhood on which the bound holds is provided. Section IV-E only demonstrates visual closeness between O_r and O_z; it does not compute d_k at any step. If the actual mismatch exceeds rho, the theorem's conclusion has no content. The abstract's unconditional phrasing, 'a stable periodic orbit in the ROM implies an input-to-state stable periodic orbit of the FOM', is therefore stronger than what is proven. This is load-bearing, and the paper should either quantify/verify the bound in the case study or consistently state the result as conditional on an unverified disturbance bound.
[Section IV-D, Implementation Details] The simulation uses PD control instead of an RES-CLF controller and intermittently re-plans z+ in Eq. (40), effectively changing the manifold M and the step-to-step map online. Both practices violate the explicit assumptions of Theorem 2, which requires a fixed hybrid-invariant manifold M driven by a controller satisfying the RES-CLF condition. The paper acknowledges these differences and says accounting for them is future work, but this means the simulation cannot be cited as an experimental verification of Theorem 2. The stability observed in Fig. 3 may be due to these extra elements, which are outside the theory. The authors should either run a simulation under the theorem's assumptions (e.g., with an actual RES-CLF and no replanning) or clearly label the numerical study as merely heuristic and separate from the formal result.
[Lemma 3 and Theorem 2] The theorem requires (via Lemma 3) that z* = Xi(r*) be a fixed point of the HZD Poincare map Omega, in addition to r* being an exponentially stable fixed point of the ROM map Qcl. This is a nontrivial matching condition: exponential stability of the ROM fixed point does not imply that the embedding maps the ROM fixed point to an HZD fixed point. The paper does not state this as a separate design requirement or show how it is satisfied in the biped example. If the condition fails, the periodic orbit Oz of the HZD may not exist at all, so this assumption is load-bearing and should be made explicit and verified.

minor comments (5)

[Abstract] The abstract's 'a stable periodic orbit in the ROM implies an input-to-state stable periodic orbit of the FOM' omits the necessary conditions (hybrid invariance of M, RES-CLF existence, small epsilon, bounded model mismatch). Please rephrase to match the conditional statement of Theorem 2.
[Lemma 1, Eq. (21)] In the proof of Lemma 1, the expression for \dot{h} uses \dot{\psi}_2(\eta,z) but the controller is written in terms of \dot{\psi}_2(\Phi(x)). It would help to clarify the coordinate dependence and to check that the controller indeed cancels all terms in the second block.
[Section IV-C, Eq. (39)] The map IKbase(p, z0) is asserted to be part of a diffeomorphism Xi, but the text only says it can be designed to be smooth and unique. This is plausible but should be stated as a design assumption, and the dependence on z0 should be explicit in the notation for Xi.
[Section IV-E, Fig. 3] The caption of Fig. 3(b) says 'Illustrates the ISS zero dynamics' but the figure appears to show trajectories, not an ISS certificate. Consider renaming to 'zero dynamics trajectories' to avoid implying a Lyapunov function is plotted.
[General notation] The paper uses \|d\|_infty for the sup norm over the discrete-time index k but \|d\| in the text sometimes refers to a pointwise norm (e.g., in the proof of Lemma 4). Please distinguish these consistently.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the stability-transfer result is conditional on an explicit model-mismatch bound and uses standard background theorems, not fitted or definitionally forced inputs.

full rationale

The derivation chain is not circular. The manifold M is constructed from a user-chosen ROM via ψ (Eqs. (20) and (41)), but the hybrid zero dynamics in (26) are the exact restriction of the full-order dynamics to M, not the ROM by definition. Lemma 4 treats the discrepancy d_k := Ω(z_k) − Ξ∘Q_cl∘Ξ^{-1}(z_k) (Eq. (31)) as an additive disturbance and invokes the converse Lyapunov theorem [36] to conclude E-ISS; the required bound ∥d∥∞ ≤ ρ is an explicit assumption in Theorem 2, not a fitted or renamed quantity. The final step is justified by [34, Thm. 1] and [37, Thm. 1] as external theorems; [37] is a prior published result and is not invoked to establish its own hypotheses. The paper itself flags the key limitation in Remark 2: 'completely characterizing the discrepancy between the HZD and a ROM is challenging', and in Sec. IV-D admits the implementation deviates from the theory (PD control instead of an RES-CLF, intermittent re-planning). These are unmet-assumption / rigor gaps, not circularity. The abstract's unconditional phrasing overstates the theorem's conditional content, but this is an overclaim, not a definitional or self-citation reduction.

Assumptions & free parameters 5 free parameters · 7 assumptions · 0 invented entities

The theorem itself is parameter-free; all listed parameters belong to the simulation example. The axioms are either standard control assumptions or conditions specific to the paper's construction that are not verified in the application.

free parameters (5)

HLIP height z0
Chosen for the simulation; sets the pendulum frequency in (37). Not derived.
Step period Tssp
Set to 0.3 s in the example; affects the S2S maps AR, BR.
Feedback gain K in control interface (40)
Tuned to stabilize the FOM; influences psi and the discrepancy d.
Feedforward step length l*
Chosen for the desired velocity c.
Desired CoM height p_d^z and torso angle theta_d
Parameters in the IK map psi (41), chosen by hand.

assumptions (7)

domain assumption The coordinate transformation Phi in (11)-(12) is a global diffeomorphism onto N x Z.
Used to decompose (2) into actuated/unactuated dynamics in (10); not proven for general robots.
domain assumption rank(B)=m and U=R^m (full actuation of actuated coordinates, unbounded inputs).
Required in Lemma 1 for the existence of the invariance controller (21).
ad hoc to paper psi_2(z) := dpsi_1/dz omega(eta,z) is independent of eta.
Assumed in Lemma 1 so that psi=(psi_1,psi_2) defines a manifold without coupling to eta. Not verified in the case study.
ad hoc to paper Discrete invariance condition (23): Delta_eta(psi(z),z)=psi(Delta_z(psi(z),z)).
Required for hybrid invariance in Lemma 2; not checked for the IK-based psi in Sec. IV.
domain assumption There exists an RES-CLF V_epsilon for (10) with respect to M.
Assumed in Theorem 2; in the example PD control is used instead, so this condition is not established.
domain assumption The model mismatch d in (31) satisfies ||d||_infinity <= rho on a neighborhood.
This is the key smallness assumption for the ISS transfer in Theorem 2; left unquantified and only visually motivated.
standard math Converse Lyapunov theorems from [36, Thm. 1,2] and the transfer result [37, Thm. 1] are valid.
Used in Lemma 4 and Theorem 2; externally published results.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Layered Control Perspective on Legged Locomotion: Embedding Reduced Order Models via Hybrid Zero Dynamics." pith.science (2026). https://pith.science/paper/DXBK65R3

@misc{pith2026250900294,
  author       = {Pith},
  title        = {Pith review of: A Layered Control Perspective on Legged Locomotion: Embedding Reduced Order Models via Hybrid Zero Dynamics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DXBK65R3}},
  note         = {Machine review of arXiv:2509.00294}
}

read the original abstract

Reduced-order models (ROMs) provide a powerful means of synthesizing dynamic walking gaits on legged robots. Yet this approach lacks the formal guarantees enjoyed by methods that utilize the full-order model (FOM) for gait synthesis, e.g., hybrid zero dynamics. This paper aims to unify these approaches through a layered control perspective. In particular, we establish conditions on when a ROM of locomotion yields stable walking on the full-order hybrid dynamics. To achieve this result, given an ROM we synthesize a zero dynamics manifold encoding the behavior of the ROM -- controllers can be synthesized that drive the FOM to this surface, yielding hybrid zero dynamics. We prove that a stable periodic orbit in the ROM implies an input-to-state stable periodic orbit of the FOM's hybrid zero dynamics, and hence the FOM dynamics. This result is demonstrated in simulation on a linear inverted pendulum ROM and a 5-link planar walking FOM.

Figures

Figures reproduced from arXiv: 2509.00294 by the authors.

**Figure 1.** We consider a layered control approach that leverages a reduced-order model (ROM) of locomotion to stabilize the dynamics of a full-order model (FOM) bipedal robot. nection between the FOM dynamics and their corresponding ROM abstractions is the approximate simulation approach [22] wherein a Lyapunov-like function, termed a simulation function, provides a quantitative metric characterizing the distance between tra… view at source ↗

**Figure 2.** (a) A linear inverted pendulum model is used as the ROM for (b) a 5-link planar walker as the FOM. In the FOM, counterclockwise rotation corresponds to positive angles. IV. APPLICATION TO BIPEDAL WALKING In this section, we elucidate the developed theory by demonstrating its application to bipedal locomotion, where we use a hybrid linear inverted pendulum (HLIP) controller to realize walking on a five-link planar bi… view at source ↗

**Figure 3.** The plots show steady-state phase portraits over 3 seconds for the robot’s zero dynamics and actuated coordinates while walking at 1.0 m/s with a 0.3 s step period. (a) Shows the nominal periodic orbit of the reduced-order model, Or. (b) Illustrates the ISS zero dynamics of the FOM alongside the transformed ROM dynamics, Ξ. (c) Shows the stability of the actuated coordinates. D. Implementation Details Our implementa… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 34 canonical work pages

[1]

M. H. Raibert, Legged robots that balance . MIT press, 1986

work page 1986
[2]

Template models for control,

P. Wensing and S. Revzen, “Template models for control,” Bioinspired Legged Locomotion. Elsevier , pp. 240–266, 2017

work page 2017
[3]

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,” in Proc. IEEE/RSJ Int. Conf. Int. Robot. Syst., vol. 1, pp. 239–246, 2001

work page 2001
[4]

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

work page 2022
[5]

Angular momentum about the contact point for control of bipedal locomotion: Validation in a lip-based controller,

Y . Gong and J. Grizzle, “Angular momentum about the contact point for control of bipedal locomotion: Validation in a lip-based controller,” arXiv preprint arXiv:2008.10763 , 2020

arXiv 2008
[6]

Multi-domain walking with reduced-order models of locomotion,

M. Dai, J. Lee, and A. D. Ames, “Multi-domain walking with reduced-order models of locomotion,” in American Control Confer- ence, pp. 2830–2837, 2024

work page 2024
[7]

The spring-mass model for running and hopping,

R. Blickhan, “The spring-mass model for running and hopping,” Journal of biomechanics , vol. 22, no. 11-12, pp. 1217–1227, 1989

work page 1989
[8]

High-speed humanoid running through control with a 3d-slip model,

P. M. Wensing and D. E. Orin, “High-speed humanoid running through control with a 3d-slip model,” in IEEE/RSJ Int. Conf. Int. Robot. Syst., pp. 5134–5140, 2013

work page 2013

Show all 37 references

[9]

Capture point: A step toward humanoid push recovery,

J. Pratt, J. Carff, S. Drakunov, and A. Goswami, “Capture point: A step toward humanoid push recovery,” in 2006 6th IEEE-RAS international conference on humanoid robots , pp. 200–207, Ieee, 2006

2006
[10]

Bipedal walking control based on capture point dynamics,

J. Englsberger, C. Ott, M. A. Roa, A. Albu-Sch ¨affer, and G. Hirzinger, “Bipedal walking control based on capture point dynamics,” in IEEE/RSJ Int. Conf. Int. Robot. Syst. , pp. 4420–4427, 2011

2011
[11]

Three-dimensional bipedal walking control based on divergent component of motion,

J. Englsberger, C. Ott, and A. Albu-Sch ¨affer, “Three-dimensional bipedal walking control based on divergent component of motion,” IEEE Transactions on Robotics , vol. 31, no. 2, pp. 355–368, 2015

2015
[12]

Dynamic walking on compliant and uneven terrain using dcm and passivity-based whole-body control,

G. Mesesan, J. Englsberger, G. Garofalo, C. Ott, and A. Albu-Sch ¨affer, “Dynamic walking on compliant and uneven terrain using dcm and passivity-based whole-body control,” in IEEE-RAS 19th International Conference on Humanoid Robots (Humanoids) , pp. 25–32, 2019

2019
[13]

Learning spring mass locomotion: Guiding policies with a reduced- order model,

K. Green, Y . Godse, J. Dao, R. L. Hatton, A. Fern, and J. Hurst, “Learning spring mass locomotion: Guiding policies with a reduced- order model,” IEEE Robot. Autom. Lett. , vol. 6, no. 2, 2021

2021
[14]

Rl-augmented mpc framework for agile and robust bipedal footstep locomotion planning and con- trol,

S. H. Bang, C. A. Jov ´e, and L. Sentis, “Rl-augmented mpc framework for agile and robust bipedal footstep locomotion planning and con- trol,” in 2024 IEEE-RAS 23rd International Conference on Humanoid Robots (Humanoids), pp. 607–614, IEEE, 2024

2024
[15]

Reduced- order model guided contact-implicit model predictive control for humanoid locomotion,

S. A. Esteban, V . Kurtz, A. B. Ghansah, and A. D. Ames, “Reduced- order model guided contact-implicit model predictive control for humanoid locomotion,” arXiv preprint arXiv:2502.15630 , 2025

2025 arXiv
[16]

Terrain- adaptive, alip-based bipedal locomotion controller via model predic- tive control and virtual constraints,

G. Gibson, O. Dosunmu-Ogunbi, Y . Gong, and J. Grizzle, “Terrain- adaptive, alip-based bipedal locomotion controller via model predic- tive control and virtual constraints,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems , pp. 6724–6731, 2022

2022
[17]

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) , pp. 12686–12693, 2024

2024
[18]

Decoupled control of bipedal locomotion based on hzd and h-lip,

Y . Ye, Y . Yue, W. Gao, and S. Zhang, “Decoupled control of bipedal locomotion based on hzd and h-lip,” in 2023 IEEE International Conference on Robotics and Biomimetics (ROBIO) , pp. 1–6, 2023

2023
[19]

Robust agility via learned zero dynamics policies,

N. Csomay-Shanklin, W. D. Compton, I. D. J. Rodriguez, E. R. Ambrose, Y . Yue, and A. D. Ames, “Robust agility via learned zero dynamics policies,” in 2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 9116–9123, IEEE, 2024

2024
[20]

Reinforcement learning for reduced-order models of legged robots,

Y .-M. Chen, H. Bui, and M. Posa, “Reinforcement learning for reduced-order models of legged robots,” in IEEE Int. Conf. Robot. Autom., pp. 5801–5807, 2024

2024
[21]

Combining trajectory optimization, supervised machine learning, and model structure for mitigating the curse of dimensionality in the control of bipedal robots,

X. Da and J. Grizzle, “Combining trajectory optimization, supervised machine learning, and model structure for mitigating the curse of dimensionality in the control of bipedal robots,” The International Journal of Robotics Research , vol. 38, no. 9, pp. 1063–1097, 2019

2019
[22]

Hierarchical control system design using approximate simulation,

A. Girard and G. J. Pappas, “Hierarchical control system design using approximate simulation,” Automatica, vol. 45, no. 2, 2009

2009
[23]

Formal connections between template and anchor models via approximate simulation,

V . Kurtz, R. R. da Silva, P. M. Wensing, and H. Lin, “Formal connections between template and anchor models via approximate simulation,” in IEEE-RAS Int. Conf. on Humanoid Robots , 2019

2019
[24]

Approximate simulation for template-based whole-body control,

V . Kurtz, P. M. Wensing, and H. Lin, “Approximate simulation for template-based whole-body control,” IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 558–565, 2020

2020
[25]

Hybrid zero dynamics of planar biped walkers,

E. R. Westervelt, J. W. Grizzle, and D. E. Koditschek, “Hybrid zero dynamics of planar biped walkers,” IEEE Transactions on Automatic Control, vol. 48, no. 1, pp. 42–56, 2003

2003
[26]

Dynamic multi-domain bipedal walking with atrias through slip based human-inspired control,

A. Hereid, S. Kolathaya, M. S. Jones, J. Van Why, J. W. Hurst, and A. D. Ames, “Dynamic multi-domain bipedal walking with atrias through slip based human-inspired control,” in Proc. Int. Conf. on Hybrid Systems: Computation and Control , pp. 263–272, 2014

2014
[27]

A. D. Ames and I. Poulakakis, Hybrid zero dynamics control of legged robots. Butterworth-Heinemann Oxford, UK, 2017

2017
[28]

Realizing dynamic and efficient bipedal locomotion on the humanoid robot durus,

J. Reher, E. A. Cousineau, A. Hereid, C. M. Hubicki, and A. D. Ames, “Realizing dynamic and efficient bipedal locomotion on the humanoid robot durus,” in 2016 IEEE International Conference on Robotics and Automation (ICRA), pp. 1794–1801, IEEE, 2016

2016
[29]

Dynamic walking with compliance on a cassie bipedal robot,

J. Reher, W.-L. Ma, and A. D. Ames, “Dynamic walking with compliance on a cassie bipedal robot,” in 2019 18th European Control Conference (ECC), pp. 2589–2595, IEEE, 2019

2019
[30]

Virtual constraints and hybrid zero dynamics for realizing underactuated bipedal locomotion,

J. W. Grizzle and C. Chevallereau, “Virtual constraints and hybrid zero dynamics for realizing underactuated bipedal locomotion,” arXiv preprint arXiv:1706.01127, 2017

2017 arXiv
[31]

Constructive nonlinear control of underactuated systems via zero dynamics policies,

W. D. Compton, I. D. J. Rodriguez, N. Csomay-Shanklin, Y . Yue, and A. D. Ames, “Constructive nonlinear control of underactuated systems via zero dynamics policies,” in 2024 IEEE 63rd Conference on Decision and Control (CDC) , pp. 8350–8357, IEEE, 2024

2024
[32]

Rapidly exponentially stabilizing control lyapunov functions and hybrid zero dynamics,

A. D. Ames, K. Galloway, K. Sreenath, and J. W. Grizzle, “Rapidly exponentially stabilizing control lyapunov functions and hybrid zero dynamics,” IEEE Trans. Autom. Ctrl. , vol. 59, no. 4, 2014

2014
[33]

E. R. Westervelt, J. W. Grizzle, C. Chevallereau, J. H. Choi, and B. Morris, Feedback control of dynamic bipedal robot locomotion . CRC press, 2018

2018
[34]

Input-to-state stability of periodic orbits of systems with impulse effects via poincar ´e analysis,

S. Veer, Rakesh, and I. Poulakakis, “Input-to-state stability of periodic orbits of systems with impulse effects via poincar ´e analysis,” IEEE Trans. Autom. Ctrl., vol. 64, no. 11, pp. 4583–4598, 2019

2019
[35]

Input-to-state stability for discrete-time nonlinear systems,

Z. P. Jiang and Y . Wang, “Input-to-state stability for discrete-time nonlinear systems,” Automatica, vol. 47, pp. 857–869, 2001

2001
[36]

A converse lyapunov theorem for discrete- time systems with disturbances,

Z. P. Jiang and Y . Wang, “A converse lyapunov theorem for discrete- time systems with disturbances,” Syst. Ctrl. Lett. , vol. 45, 2002

2002
[37]

Input to state stabilizing control lyapunov functions for robust bipedal locomotion,

S. Kolathaya, J. Reher, A. Hereid, and A. D. Ames, “Input to state stabilizing control lyapunov functions for robust bipedal locomotion,” in American Control Conference, pp. 2224–2230, 2018

2018

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

M. H. Raibert, Legged robots that balance . MIT press, 1986

work page 1986

[2] [2]

Template models for control,

P. Wensing and S. Revzen, “Template models for control,” Bioinspired Legged Locomotion. Elsevier , pp. 240–266, 2017

work page 2017

[3] [3]

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,” in Proc. IEEE/RSJ Int. Conf. Int. Robot. Syst., vol. 1, pp. 239–246, 2001

work page 2001

[4] [4]

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

work page 2022

[5] [5]

Angular momentum about the contact point for control of bipedal locomotion: Validation in a lip-based controller,

Y . Gong and J. Grizzle, “Angular momentum about the contact point for control of bipedal locomotion: Validation in a lip-based controller,” arXiv preprint arXiv:2008.10763 , 2020

arXiv 2008

[6] [6]

Multi-domain walking with reduced-order models of locomotion,

M. Dai, J. Lee, and A. D. Ames, “Multi-domain walking with reduced-order models of locomotion,” in American Control Confer- ence, pp. 2830–2837, 2024

work page 2024

[7] [7]

The spring-mass model for running and hopping,

R. Blickhan, “The spring-mass model for running and hopping,” Journal of biomechanics , vol. 22, no. 11-12, pp. 1217–1227, 1989

work page 1989

[8] [8]

High-speed humanoid running through control with a 3d-slip model,

P. M. Wensing and D. E. Orin, “High-speed humanoid running through control with a 3d-slip model,” in IEEE/RSJ Int. Conf. Int. Robot. Syst., pp. 5134–5140, 2013

work page 2013