Pith. sign in

REVIEW 4 major objections 6 minor 92 references

Diffeomorphic Obstacle Avoidance for Contractive Dynamical Systems via Implicit Representations

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a contractive dynamical system can be reshaped around obstacles by a flow generated from a signed distance field, preserving the contraction guarantee through diffeomorphic coordinate changes.

desk verdict The central avoidance claim is not supported by the equations: SDC/SDDC are defined as pullbacks, so they are trajectory-equivalent to the original dynamics and cannot avoid obstacles. read the letter →

arxiv 2504.18860 v1 pith:KTO2OHJM submitted 2025-04-26 cs.RO cs.LGcs.SYeess.SY

Ken-Joel Simmoteit , Philipp Schillinger , Leonel Rozo This is my paper · ORCID
classification cs.ROcs.LGcs.SYeess.SY
keywords contractiontheorysigneddistancefieldsdiffeomorphictransformsobstacleavoidancelearningfromdemonstrationneuralcontractivedynamicalsystemsrobotmanipulationstabilitypreservation
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 claims that a contractive dynamical system, a robot skill whose trajectories converge exponentially to a learned motion, can be reshaped around obstacles without losing that convergence guarantee. The device is a flow generated by a barrier-weighted gradient of a signed distance field: the flow defines a diffeomorphism, and because contraction is invariant under diffeomorphic coordinate changes, the transformed system stays contractive. The paper introduces two versions, a differential coordinate change (SDDC) and a pullback coordinate change (SDC), and evaluates them on synthetic 2D skills and real kitchen manipulation tasks with whole-body obstacle avoidance. If the claim holds, learned skills can be given reactive safety without redesigning their stability certificates.

What carries the argument

The load-bearing object is the flow-based diffeomorphism $\psi$: the time-$t$ map of the vector field $V(x,q)=-b(x,q)\nabla_q\Gamma_{\mathrm{SDF}}(x,q)$, where $\Gamma_{\mathrm{SDF}}$ is a signed distance field, a robot distance field (RDF), or a configuration-space distance field (CDF), and $b$ is an inverse barrier that diverges at the obstacle surface. The flow turns scalar distance information into a coordinate transformation, and its Jacobian $J_\psi$ supplies the contraction metric $G_\psi=J_\psi^\top J_\psi$; contraction invariance under such coordinate changes is what lets the modulated system inherit the original skill's exponential convergence. A swept barrier term optionally masks the avoidance field where the motion moves away from the obstacle, and a friction term rescales the modulated velocity to better match the original skill.

What would settle it

Take two nearby initial conditions of the modulated system near a concave obstacle, integrate them with a high-accuracy solver, and measure their separation over time; if it does not decay exponentially at the predicted contraction rate (or grows) before the obstacle is cleared, the preservation claim fails. Equivalently, evaluate $\det J_\psi$ along the flow on a grid approaching the obstacle surface: if it approaches zero or changes sign anywhere in the working domain, the flow is not a global diffeomorphism and the contraction-invariance theorem cannot be applied.

Watch

Extended reading notes

Core claim

The central claim is that obstacle avoidance for contractive robot skills can be achieved by composing the skill dynamics with a diffeomorphism built from the environment's implicit signed distance field. Specifically, the gradient $\nabla_q\Gamma_{\mathrm{SDF}}$ scaled by an inverse barrier $b_{\mathrm{inv}}=s_{\mathrm{grad}}/(\Gamma_{\mathrm{SDF}}-t_{\mathrm{save}})$ defines an infinitesimal generator whose flow $\psi(y)=y-\int_0^t b\,\nabla_q\Gamma_{\mathrm{SDF}}\,du$ is a candidate diffeomorphism. Transforming the contractive dynamics $f_c$ either as a differential coordinate change $\delta y=J_\psi^{-1}\delta q$ (SDDC) or as a pullback $\dot y=J_\psi^{-1}f_c(\psi(y))$ with Riemannian metric $G_\psi=J_\psi^\top J_\psi$ (SDC) preserves contraction by Theorem 1, so the robot avoids obstacles while retaining exponential convergence to the intended motion. The paper also claims that this reshaping stays close to the learned vector field, as measured by its new curvature and misalignment metrics, and demonstrates the approach on learned whole-body kitchen skills.

Load-bearing premise

The argument rests on assuming the obstacle-avoidance flow is a smooth, invertible coordinate change over the entire working area, with a well-conditioned derivative everywhere and a uniformly well-behaved induced metric; the paper asserts these properties rather than proving them, even though the barrier that generates the flow becomes singular exactly at the obstacle surface.

Editorial extensions

If this is right

  • A contractive skill learned in joint space can be given whole-body obstacle avoidance without retraining the skill or changing its learned vector field.
  • The approach applies to any contractive dynamical system, not only the neural contractive systems used in the experiments, because the preservation argument is purely geometric.
  • Obstacle avoidance can be quantified beyond minimum distance: the proposed flow-curvature and vector-field-misalignment metrics allow direct comparison of how much a modulation distorts the original skill.
  • Real-time execution is feasible for cluttered scenes when the implicit distance is queried at the nearest point of the point cloud, as shown in the kitchen experiments.
  • The guarantees and reliable behavior are demonstrated for convex obstacles; concave obstacle shapes can trap the flow and stall the skill.

Reading between the lines

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

  • The construction implies that the contraction-preservation guarantee is inherited only on the region where the flow is a true diffeomorphism; a practical implementation could monitor the condition of the flow's derivative online and treat a near-singular Jacobian as a signal to stop modulating.
  • Because the method works with any signed-distance representation, a natural extension is to learn the scene as a neural signed distance field and use the same flow for mobile manipulation; the paper mentions this direction but does not test it.
  • The relative-motion barrier sketched in the future-work section suggests a concrete extension to moving obstacles: the same flow construction with an obstacle-velocity-dependent barrier would give reactive avoidance in dynamic scenes, with the contraction guarantee still riding on the flow being a diffeomorphism.
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 / 6 minor

Summary. The paper proposes a framework, Signed Distance Field Diffeomorphic Transform (SDT), for adding obstacle avoidance to contractive dynamical systems while preserving contraction. It constructs an infinitesimal generator from the gradient of an SDF weighted by inverse or swept barrier functions, integrates this generator into a flow psi (Eq. 24), and uses psi to modulate a neural contractive dynamical system either through a differential coordinate change (SDDC, Eq. 26) or a coordinate change (SDC, Eq. 27). The authors claim that Theorem 1 entails contraction preservation and therefore safe obstacle avoidance. They introduce two metrics (RFC, VM), compare against MM, DT, ARPF on LASA synthetic tasks, and demonstrate two real-world kitchen tasks with learned RDF/CDF representations.

Significance. If the central claim were correct, the paper would fill a real gap: combining implicit robot/scene representations with contraction-preserving whole-body obstacle avoidance. The experimental apparatus is substantial: learned RDF/CDF representations, swept barrier features, comparisons over SDF architectures, real-robot demonstrations, and inference-time decompositions. The paper also makes a good-faith attempt to quantify trajectory curvature and vector-field alignment. However, the core mathematical object analyzed for stability is not the object that performs avoidance; the defined SDC is a pullback of the original system, and SDDC is not defined as a dynamical system. Because the main contribution is precisely the claim that these transforms yield both avoidance and contraction, the paper's central conclusion is unsupported.

major comments (4)
  1. [IV-B4, Eq. (27)] Equation (27) defines fSDC(y)=Jpsi(y)^{-1} fc(psi(y)), which is the pullback of fc by psi, not an obstacle-avoiding pushforward. With q=psi(y), one immediately obtains q_dot=Jpsi(y) y_dot=fc(psi(y))=fc(q), so the configuration-space trajectory is exactly the unmodulated NCDS trajectory. Since psi is a diffeomorphism, a trajectory avoids the obstacle set O if and only if its preimage avoids psi^{-1}(O); a coordinate change cannot turn a colliding trajectory into a non-colliding one. Therefore the claim in Section IV-B3 that 'the robot can successfully avoid obstacles while preserving the stability of the underlying skill' does not follow from Eq. (27), and the avoidance visible in Figures 6 and 9 must be produced by some other, undisclosed controller. The authors need to define and analyze the pushforward system q_dot=Jpsi(psi^{-1}(q)) fc(psi^{-1}(q)) (or an equivalent) and prove obstacle avoidance for that system.
  2. [IV-B3, Eq. (26)] Equation (26) is a virtual-displacement relation delta y=Jpsi^{-1} delta q; no dynamical system fSDDC is ever defined. The text concludes 'the transformed system fSDDC is contractive', but there is no ODE for y or q to which Theorem 1 could be applied. If the intended definition is y_dot=Jpsi^{-1} fc(psi(y)), it is again a pullback and inherits the avoidance problem of the previous comment; if something else is intended, it needs to be written down and analyzed.
  3. [Appendix B4, Eq. (24)] The contraction proof assumes that psi is a global smooth diffeomorphism with pointwise nonsingular Jpsi and a uniformly positive-definite induced metric Gpsi=Jpsi^T Jpsi. The generator (22) contains the inverse barrier 1/(Gamma_SDF - tsave), which is singular on the obstacle boundary; global existence of the flow (24) is not established, Jpsi may become unbounded or singular in a neighborhood of the boundary, and uniform positive-definiteness of Gpsi is not checked. Consequently the claimed global contraction guarantee is not proven even for the pullback system considered in Appendix B4.
  4. [IV-B5, Eq. (28)] The statement 'According to Theorem 1, contraction is preserved under affine transformations, consequently the friction (28) does not compromise the contraction guarantees' is unsupported: the factor eta_f ||fc(y)||/||fm(y)|| is a state-dependent scalar scaling of the vector field, not an affine feedback transformation as in Theorem 1(1), and positive scalar scaling does not in general preserve contraction. Since the friction variant is presented as part of the method (Figure 6c), the contraction guarantee for that variant needs a separate proof.
minor comments (6)
  1. [Eq. (14)] The noise-injected loss in Eq. (14) is defined with a leading minus sign, so minimizing Lnoise maximizes the squared error; this is likely a sign error and should be corrected.
  2. [IV-B2, Eqs. (19) and (37)] The symbol for the safety threshold appears as tsave in Eq. (19) and tsafe in Eq. (37); please unify.
  3. [Algorithm 2, line 5] The notation psi(x, qt) is ambiguous because Eq. (24) takes (x, y) and returns a configuration q; clarify which variables are inputs and which are outputs in Algorithm 2.
  4. [V-A, Eq. (31)] The definition of RFC takes maxima over points x in tau_base and tau_m, but curvature is a function of time; the reparametrization and alignment of the two trajectories should be specified.
  5. [VII-B, Eq. (39)] In the dynamic-obstacle barrier formula, the denominator |qSDT| should presumably be |q_dot_SDT|; the expression is dimensionally inconsistent as written.
  6. [Table II] Several entries contain misplaced commas (e.g., '18 , 63' and '5 , 13') that should be decimal points.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the contraction-preservation argument is a direct application of external theorems and the new contribution is not a renamed fit or prediction.

full rationale

The paper's derivation chain is self-contained with respect to circularity. The learned NCDS (Eqs. 9-11) is taken from prior work as a contractive input; the paper's contribution is a flow-based coordinate transformation (Eq. 24) and the claim that contraction is inherited via Theorem 1 of Manchester and Slotine [41]. The proof in Appendix B4 directly verifies the contraction inequality for the transformed system using the induced metric Gψ = Jψ^T Jψ, so the stability conclusion is derived from stated assumptions rather than assumed. The barrier function and infinitesimal generator (Eqs. 18-22) are constructed from the SDF, and no fitted parameter is later relabeled as a prediction. The self-citations to NCDS [4,5] provide the base contractive dynamics but are not the load-bearing step for the new transform's contraction guarantee. The skeptic's observation that Eq. (27) is a pullback and therefore trajectory-equivalent to fc in ψ-coordinates is a correctness concern about whether the claimed avoidance follows from the presented equations; it is not a circular reduction of a derived result to its inputs, and it does not warrant a circularity score under the stated rubric. The paper itself acknowledges limitations (concave obstacles, discretization, inference time) in Section VII-A, which further indicates the claims are not being protected by definition.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

No new physical entities are introduced. The obstacle-avoiding manifold Y and the SDC/SDDC transforms are mathematical constructions built from existing concepts. The central claim rests on prior contraction theory, on the flow/diffeomorphism assumption for the SDF barrier field, and on the accuracy of learned SDFs. The friction-term extension additionally uses an unproven assumption about state-dependent scaling preserving contraction.

free parameters (4)
  • sgrad = 0.05 for SDDC, 0.1 for SDC and DT in the experiments
    Scaling factor in the barrier function (Eq. 19) controls how strongly the SDF gradient repels; the paper tunes it differently per method without a sensitivity analysis across the comparison.
  • tsafe / tsave = 0.1
    Safety threshold subtracted in the barrier function (Eq. 19); set to 0.1 for all methods in the experiments.
  • beta_f = not reported
    Hyperparameter in the friction term (Eq. 29) that controls velocity slowdown near obstacles; no value is given in the paper.
  • epsilon (NCDS eigenvalue bound) = learned via regularizer (Eq. 13)
    Small positive eigenvalue bound in the NCDS Jacobian (Eq. 9) is optimized as part of training, following Beik-Mohammadi et al. [4]; this is part of the base skill model rather than the avoidance method.
assumptions (5)
  • standard math Contraction is invariant under smooth coordinate changes (Manchester and Slotine [41]).
    The whole method rests on this theorem, invoked in Section IV-B3/B4 and used in Appendix B4 to conclude that SDC and SDDC preserve contraction.
  • domain assumption The flow in Eq. (24) is a diffeomorphism on the working domain for the integration horizon used.
    Needed for Jψ to be nonsingular and for the pullback to be well-defined. Not proven, and the barrier 1/Γ_SDF is singular on the SDF zero level set.
  • domain assumption The learned SDF is smooth and accurate enough that ∇Γ and the barrier-infinitesimal generator V define a well-behaved flow.
    The method relies on the Eikonal regularizer to produce unit gradients, but the paper acknowledges in Section VII-A that local inaccuracies in learned distance functions could theoretically cause local instability.
  • domain assumption The NCDS base model is contractive with a uniformly negative definite symmetric Jacobian.
    Used in Appendix B4 to conclude A + A^T + 2αI ≺ 0. This is taken from prior work by the same group and is not re-derived here.
  • ad hoc to paper State-dependent positive scaling of a contractive vector field preserves contraction.
    The friction term (Eq. 28) multiplies the modulated field by a positive state-dependent scalar and is claimed, via Theorem 1, to preserve contraction. The cited affine-feedback theorem does not cover this operation and no proof specific to this case is supplied.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffeomorphic Obstacle Avoidance for Contractive Dynamical Systems via Implicit Representations." pith.science (2026). https://pith.science/paper/KTO2OHJM

@misc{pith2026250418860,
  author       = {Pith},
  title        = {Pith review of: Diffeomorphic Obstacle Avoidance for Contractive Dynamical Systems via Implicit Representations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KTO2OHJM}},
  note         = {Machine review of arXiv:2504.18860}
}
read the original abstract

Ensuring safety and robustness of robot skills is becoming crucial as robots are required to perform increasingly complex and dynamic tasks. The former is essential when performing tasks in cluttered environments, while the latter is relevant to overcome unseen task situations. This paper addresses the challenge of ensuring both safety and robustness in dynamic robot skills learned from demonstrations. Specifically, we build on neural contractive dynamical systems to provide robust extrapolation of the learned skills, while designing a full-body obstacle avoidance strategy that preserves contraction stability via diffeomorphic transforms. This is particularly crucial in complex environments where implicit scene representations, such as Signed Distance Fields (SDFs), are necessary. To this end, our framework called Signed Distance Field Diffeomorphic Transform, leverages SDFs and flow-based diffeomorphisms to achieve contraction-preserving obstacle avoidance. We thoroughly evaluate our framework on synthetic datasets and several real-world robotic tasks in a kitchen environment. Our results show that our approach locally adapts the learned contractive vector field while staying close to the learned dynamics and without introducing highly-curved motion paths, thus outperforming several state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2504.18860 by the authors.

Figure 1
Figure 1. Overview of the proposed Signed Distance Field Diffeo￾morphic Transform: (1) A neural contractive dynamical system; (2) A learned implicit distance function; (3) A barrier function, and (4) a contraction-preserving diffeomorphic transform. antees exponential convergence to a trajectory [38, 67]. This concept has been recently leveraged in LfD frameworks, where a robot skill, represented by a first-order dynamical sy… view at source ↗
Figure 2
Figure 2. Illustration of a 2D SDF: Contour lines of an star-shaped SDF with the distance ΓSDF to the obstacle surface, depicted as a solid red line. IV. CONTRACTION PRESERVING OBSTACLE AVOIDANCE Here we introduce our contraction-preserving obstacle avoidance method. First, we explain the contractive dynamical system used in this paper. Later, we describe the contraction￾preserving obstacle avoidance method. A. Learned Contra… view at source ↗
Figure 6
Figure 6. Comparison of contraction-preserving obstacle avoidance using SDDC and SDC on an example of the LASA dataset. The vector [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figures from the paper (11 more)
Figure 7
Figure 7. Figure 7: Illustration of the obstacles SDF used in the 2D LASA dataset experiments. The contour lines represent the distance ΓSDF to the obstacle surface, depicted as a solid red line. Four simple obstacles are considered: Circle, Box, Triangle, Arc [PITH_FULL_IMAGE:figures/fu…
Figure 7
Figure 7. Figure 7: Using this representation, we sample N = 160K points {xi} N i=0 from a grid around the obstacle surface and we then assign each point its corresponding SDF di = ΓSDF(xi) and gradient values gi = ∇xΓSDF(xi). Given this training set D = {(xi , di , gi)} N i=0, a two-dime…
Figure 10
Figure 10. Figure 10: Real-world robot setup in a kitchen environment. It [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]
Figure 9
Figure 9. Figure 9: Obstacle avoidance with SDC and SDDC using a inverse barrier on a NCDS trained on the Sine trajectories of the LASA dataset. The circle-shaped obstacle is depicted by the red line. The velocity profile shows the absolute velocity value along the trajectory. The vector …
Figure 11
Figure 11. Figure 11: EMPTYING THE DISHWASHER task: A series of overlapped snapshots shows the PLACE PLATE skill performed by the Franka￾Emika Panda robot using an RDF with grasped plate (left and middle panel). The right plot shows a comparison of the resulting trajectories using differen…
Figure 13
Figure 13. Figure 13: Comparison of the inference times for the SDC method [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]
Figure 12
Figure 12. Figure 12: Adaptation robustness test for the PLACE PLATE IN RACK skill performed on the Frank-Emika Panda robot using CDF and under perturbations at 10s and 21s. STACKED CUPS and PLACE PLATE. Specifically, we test the SDT pipeline with an RDF as distance function and the SDC me…
Figure 14
Figure 14. Figure 14: Obstacle avoidance with SDC using an inverse barrier with swept features on an NCDS model trained on a multi-modal LASA dataset skill. Two circular obstacles are depicted by red lines. Table V: Overview of various ODE solvers ODE Solver Description Pros Cons Complexit…
Figure 15
Figure 15. Figure 15: Move Sine skill performed on the robot with external [PITH_FULL_IMAGE:figures/full_fig_p020_15.png]
Figure 16
Figure 16. Figure 16: Overview of all 6 motion sequences in the dishwasher dataset. Demonstrations are displayed as black solid trajectories, with start and end points represented as red crosses and gray points, respectively. (a) Motion sequence without obstacle avoidance (b) Obstacle avoi…
Figure 17
Figure 17. Figure 17: EMPTYING THE DISHWASHER task: A series of overlapped snapshots shows the PLACE PLATE IN RACK skill performed by the Franka-Emika Panda robot using an CDF with grasped plate in the right panel and without the grasped plate in the middle panel. The left plot shows the e…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

92 extracted references · 43 canonical work pages

  1. [1]

    Contractive dynamical imitation policies for efficient out-of-sample recovery

    Amin Abyaneh, Mahrokh Ghoddousi Boroujeni, Hsiu- Chin Lin, and Giancarlo Ferrari-Trecate. Contractive dynamical imitation policies for efficient out-of-sample recovery. In International Conference on Learning Rep- resentations (ICLR), 2025. URL https://openreview.net/ forum?id=lILEtkWOXD. 2

    2025

  2. [2]

    Ames, Samuel Coogan, Magnus Egerst- edt, Gennaro Notomista, Koushil Sreenath, and Paulo Tabuada

    Aaron D. Ames, Samuel Coogan, Magnus Egerst- edt, Gennaro Notomista, Koushil Sreenath, and Paulo Tabuada. Control barrier functions: Theory and appli- cations. In European Control Conference (ECC) , pages 3420–3431, 2019. URL https://doi.org/10.23919/ECC. 2019.8796030. 14

    arXiv 2019

  3. [3]

    Argall, Sonia Chernova, Manuela Veloso, and Brett Browning

    Brenna D. Argall, Sonia Chernova, Manuela Veloso, and Brett Browning. A survey of robot learning from demonstration. Robotics and Autonomous Systems , 57 (5):469–483, 2009. ISSN 0921-8890. URL https://doi. org/10.1016/j.robot.2008.10.024. 1

    doi:10.1016/j.robot.2008.10.024 2009

  4. [4]

    Neural contractive dynamical systems

    Hadi Beik-Mohammadi, Søren Hauberg, Georgios Ar- vanitidis, Nadia Figueroa, Gerhard Neumann, and Leonel Rozo. Neural contractive dynamical systems. In International Conference on Learning Representations (ICLR), 2024. URL https://openreview.net/forum?id= iAYIRHOYy8. 1, 2, 4, 8, 9

    2024

  5. [5]

    Extended Neural Contractive Dynamical Systems: On Multiple Tasks and Riemannian Safety Regions

    Hadi Beik-Mohammadi, Søren Hauberg, Georgios Ar- vanitidis, Gerhard Neumann, and Leonel Rozo. Extended neural contractive dynamical systems: On multiple tasks and Riemannian safety regions, 2024. URL https://arxiv. org/abs/2411.11405. 2, 5

    work page Pith review arXiv 2024

  6. [6]

    Berndt and James Clifford

    Donald J. Berndt and James Clifford. Using dy- namic time warping to find patterns in time series. In International Conference on Knowledge Discov- ery and Data Mining (KDD) , page 359–370. AAAI Press, 1994. URL https://cdn.aaai.org/Workshops/1994/ WS-94-03/WS94-03-031.pdf. 9, 20

    1994

  7. [7]

    Robot Programming by Demonstration , pages 1371–1394

    Aude Billard, Sylvain Calinon, R ¨udiger Dillmann, and Stefan Schaal. Robot Programming by Demonstration , pages 1371–1394. Springer Berlin Heidelberg, Berlin, Heidelberg, 2008. ISBN 978-3-540-30301-5. URL https: //doi.org/10.1007/978-3-540-30301-5 60. 1

    doi:10.1007/978-3-540-30301-5 2008

  8. [8]

    Billard, Sylvain Calinon, and R ¨udiger Dillmann

    Aude G. Billard, Sylvain Calinon, and R ¨udiger Dillmann. Learning from Humans , pages 1995–

    1995

Show all 92 references
  1. [9]

    Learning stable dynamical systems using contraction the- ory

    Caroline Blocher, Matteo Saveriano, and Dongheui Lee. Learning stable dynamical systems using contraction the- ory. In International Conference on Ubiquitous Robots and Ambient Intelligence (URAI) . IEEE, 2017. doi: 10.1109/urai.2017.7992901. URL http://dx.doi.org/10. 1109/URA...

    2017

  2. [10]

    Ricky T. Q. Chen, Yulia Rubanova, Jesse Bettencourt, and David K Duvenaud. Neural ordinary differential equations. In Advances in Neural Information Processing Systems (NeurIPS) , 2018. URL https://proceedings.neurips.cc/paper files/paper/2018/ file/69386f6bb1dfed68692a24c8686...

    2018

  3. [11]

    Differen- tiable robot neural distance function for adaptive grasp synthesis on a unified robotic arm-hand system, 2023

    Yiting Chen, Xiao Gao, Kunpeng Yao, Lo ¨ıc Nieder- hauser, Yasemin Bekiroglu, and Aude Billard. Differen- tiable robot neural distance function for adaptive grasp synthesis on a unified robotic arm-hand system, 2023. URL https://arxiv.org/abs/2309.16085. 2, 3, 5

    2023 arXiv

  4. [12]

    Safe control with learned certificates: A survey of neural Lyapunov, barrier, and contraction methods for robotics and control

    Charles Dawson, Sicun Gao, and Chuchu Fan. Safe control with learned certificates: A survey of neural Lyapunov, barrier, and contraction methods for robotics and control. IEEE Transactions on Robotics (T-RO), 39: 1749–1767, 2022. URL http://dx.doi.org/10.1109/TRO. 2022.3232542. 1

    2022

  5. [13]

    Fourie, Nadia Figueroa, and Julie A

    Christopher K. Fourie, Nadia Figueroa, and Julie A. Shah. On-manifold strategies for reactive dynamical system modulation with nonconvex obstacles. IEEE Transactions on Robotics (T-RO) , 40:2390–2409, 2024. URL https://doi.org/10.1109/TRO.2024.3378179. 2, 14

    2024

  6. [14]

    Distance and collision probability estimation from gaussian surface models,

    Kshitij Goel and Wennie Tabib. Distance and collision probability estimation from gaussian surface models,

  7. [15]

    Implicit geometric regularization for learning shapes

    Amos Gropp, Lior Yariv, Niv Haim, Matan Atzmon, and Yaron Lipman. Implicit geometric regularization for learning shapes. In International Conference on Machine Learning (ICML), pages 3789–3799. PMLR, 2020. URL https://proceedings.mlr.press/v119/gropp20a.html. 2, 5, 9, 12, 19

    2020

  8. [16]

    Carmichael, and Jochen Deuse

    Victor Hernandez Moreno, Steffen Jansing, Mikhail Po- likarpov, Marc G. Carmichael, and Jochen Deuse. Ob- stacles and opportunities for learning from demonstration in practical industrial assembly: A systematic literature review. Robotics and Computer-Integrated Manufactur- in...

    2024

  9. [17]

    Kernelized movement primitives

    Yanlong Huang, Leonel Rozo, Jo ˜ao Silv ´erio, and Dar- win G Caldwell. Kernelized movement primitives. International Journal of Robotics Research (IJRR) , 38(7):833–852, 2019. URL https://doi.org/10.1177/ 0278364919846363. 1

    2019

  10. [18]

    Avoidance of convex and concave obstacles with conver- gence ensured through contraction

    Lukas Huber, Aude Billard, and Jean-Jacques Slotine. Avoidance of convex and concave obstacles with conver- gence ensured through contraction. IEEE Robotics and Automation Letters (RA-L), 4(2):1462–1469, April 2019. ISSN 2377-3774. URL http://dx.doi.org/10.1109/LRA. 2019.2893676. 1, 2

    2019

  11. [19]

    Avoiding dense and dynamic obstacles in enclosed spaces: Application to moving in crowds

    Lukas Huber, Jean-Jacques Slotine, and Aude Billard. Avoiding dense and dynamic obstacles in enclosed spaces: Application to moving in crowds. IEEE Transac- tions on Robotics (T-RO) , 38(5):3113–3132, 2022. doi: 10.1109/TRO.2022.3164789. 1, 2, 7, 8, 9, 10

    2022

  12. [20]

    Humphreys

    James E. Humphreys. Introduction to Lie Algebras and Representation Theory. Springer New York, 1972. ISBN 9781461263982. doi: 10.1007/978-1-4612-6398-2. URL http://dx.doi.org/10.1007/978-1-4612-6398-2. 4

    1972 doi

  13. [21]

    Dynamical movement primitives: Learning attractor models for motor behav- iors

    Auke Jan Ijspeert, Jun Nakanishi, Heiko Hoffmann, Peter Pastor, and Stefan Schaal. Dynamical movement primitives: Learning attractor models for motor behav- iors. Neural Computation , 25:328–373, 2013. URL https://doi.org/10.1162/NECO a 00393. 1

    2013 doi

  14. [22]

    Neural fields in robotics: A survey, 2024

    Muhammad Zubair Irshad, Mauro Comi, Yen-Chen Lin, Nick Heppert, Abhinav Valada, Rares Ambrus, Zsolt Kira, and Jonathan Tremblay. Neural fields in robotics: A survey, 2024. URL https://arxiv.org/abs/2410.20220. 2

    2024 arXiv

  15. [23]

    Learning neural contracting dynamics: Extended linearization and global guarantees

    Sean Jaffe, Alexander Davydov, Deniz Lapsekili, Ambuj Singh, and Francesco Bullo. Learning neural contracting dynamics: Extended linearization and global guarantees. In Conference on Neural Information Processing Systems (NeurIPS), 2024. URL https://openreview.net/forum?id= YY...

    2024

  16. [24]

    Least-squares fitting of algebraic spline surfaces

    Bert J ¨uttler and Alf Felis. Least-squares fitting of algebraic spline surfaces. Advances in Computational Mathematics, 17(1/2):135–152, 2002. ISSN 1019-7168. doi: 10.1023/a:1015200504295. URL http://dx.doi.org/ 10.1023/A:1015200504295. 9, 19

    2002 doi

  17. [25]

    Learning stable nonlinear dynamical systems with Gaus- sian mixture models

    Seyed Mohammad Khansari-Zadeh and Aude Billard. Learning stable nonlinear dynamical systems with Gaus- sian mixture models. IEEE Transactions on Robotics (T-RO), 27(5):943–957, 2011. URL http://dx.doi.org/10. 1109/TRO.2011.2159412. 1, 2

    2011

  18. [26]

    A dynamical system approach to realtime obstacle avoidance

    Seyed Mohammad Khansari-Zadeh and Aude Billard. A dynamical system approach to realtime obstacle avoidance. Autonomous Robots , 32(4):433–454, March

  19. [27]

    O. Khatib. Real-time obstacle avoidance for manipulators and mobile robots. In IEEE International Conference on Robotics and Automation (ICRA) , volume 2, pages 500– 505, 1985. URL http://dx.doi.org/10.1109/ROBOT.1985. 1087247. 1, 9, 10

    1985 doi

  20. [28]

    Jin-Oh Kim and Pradeep K. Khosla. Real-time obstacle avoidance using harmonic potential functions. IEEE Transactions on Robotics and Automation, 8(3):338–349,

  21. [29]

    On the design of region-avoiding met- rics for collision-safe motion generation on Riemannian manifolds

    Holger Klein, No ´emie Jaquier, Andre Meixner, and Tamim Asfour. On the design of region-avoiding met- rics for collision-safe motion generation on Riemannian manifolds. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 2346–2353,

  22. [30]

    Neural joint space implicit signed distance functions for reactive robot manipulator control

    Mikhail Koptev, Nadia Figueroa, and Aude Billard. Neural joint space implicit signed distance functions for reactive robot manipulator control. IEEE Robotics and Automation Letters (RA-L) , 8(2):480–487, February

  23. [31]

    Learning from demonstration facilitates human-robot collaborative task execution

    Maria Koskinopoulou, Stylianos Piperakis, and Panos Trahanias. Learning from demonstration facilitates human-robot collaborative task execution. In ACM/IEEE International Conference on Human-Robot Interaction (HRI), pages 59–66, 2016. URL https://doi.org/10.1109/ HRI.2016.7451734. 1

    2016

  24. [32]

    John M. Lee. Riemannian Manifolds . Springer New York, 1997. ISBN 9780387227269. doi: 10.1007/ b98852. URL http://dx.doi.org/10.1007/b98852. 3, 7

    1997 doi

  25. [33]

    John M. Lee. Introduction to Smooth Manifolds . Springer New York, 2012. ISBN 9781441999825. doi: 10.1007/978-1-4419-9982-5. URL http://dx.doi.org/10. 1007/978-1-4419-9982-5. 4

    2012 doi

  26. [34]

    Lemme, Y

    A. Lemme, Y . Meirovitch, M. Khansari-Zadeh, T. Flash, A. Billard, and J. J. Steil. Open-source benchmarking for learned reaching motion generation in robotics. Paladyn, Journal of Behavioral Robotics , 6(1), 2015. URL http: //doi.org/10.1515/pjbr-2015-0002. 8, 10

    2015 doi

  27. [35]

    Configuration space distance fields for manipu- lation planning

    Yiming Li, Xuemin Chi, Amirreza Razmjoo, and Sylvain Calinon. Configuration space distance fields for manipu- lation planning. In Proc. Robotics: Science and Systems (R:SS), 2024. URL https://arxiv.org/abs/2406.01137. 3, 5, 9

    2024 arXiv

  28. [36]

    doi: 10.1109/lra.2022.3227860

    ISSN 2377-3774. doi: 10.1109/lra.2022.3227860. URL http://dx.doi.org/10.1109/LRA.2022.3227860. 2, 3, 5

    2022

  29. [37]

    Regularized deep signed distance fields for reactive motion generation

    Puze Liu, Kuo Zhang, Davide Tateo, Snehal Jauhri, Jan Peters, and Georgia Chalvatzaki. Regularized deep signed distance fields for reactive motion generation. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages 6673–6680, 2022. URL https://doi...

    2022

  30. [38]

    Winfried Lohmiller and Jean-Jacques E. Slotine. On contraction analysis for non-linear systems. Automatica, 34(6):683–696, 1998. ISSN 0005-1098. URL https: //doi.org/10.1016/S0005-1098(98)00019-3. 1, 3

    1998 doi

  31. [39]

    JacNet: Learning functions with structured Jacobians

    Jonathan Lorraine and Safwan Hossain. JacNet: Learning functions with structured Jacobians. INNF Workshop at the International Conference on Machine Learning (ICML) , 2019. URL https://invertibleworkshop.github.io/INNF 2019/ accepted papers/pdfs/INNF 2019 paper 10.pdf. 4

    2019

  32. [40]

    Robot programming

    Tomas Lozano-Perez. Robot programming. Proceedings of the IEEE , 71(7):821–841, 1983. doi: 10.1109/PROC. 1983.12681. 1

    1983

  33. [41]

    Manchester and Jean-Jacques E

    Ian R. Manchester and Jean-Jacques E. Slotine. Con- trol contraction metrics: Convex and intrinsic criteria for nonlinear feedback design. IEEE Transactions on Automatic Control, 62(6):3046–3053, 2017. URL http: //dx.doi.org/10.1109/TAC.2017.2668380. 3, 19

    2017

  34. [42]

    Representing robot geometry as distance fields: Applications to whole-body manipulation

    Yiming Li, Yan Zhang, Amirreza Razmjoo, and Sylvain Calinon. Representing robot geometry as distance fields: Applications to whole-body manipulation. In IEEE International Conference on Robotics and Automation (ICRA), pages 15351–15357, 2024. URL http://dx.doi. org/10.1109/ICR...

    2024

  35. [43]

    Rmmi: Enhanced obstacle avoidance for reactive mobile manipulation using an implicit neural map, 2024

    Nicolas Marticorena, Tobias Fischer, Jesse Haviland, and Niko Suenderhauf. Rmmi: Enhanced obstacle avoidance for reactive mobile manipulation using an implicit neural map, 2024. URL https://arxiv.org/abs/2408.16206. 3

    2024 arXiv

  36. [44]

    Klaus Neumann and Jochen J. Steil. Learning robot motions with stable dynamical systems under diffeo- morphic transformations. Robotics and Autonomous Systems (RAS) , 70:1–15, 2015. ISSN 0921-8890. URL https://doi.org/10.1016/j.robot.2015.04.006. 2

    2015 doi

  37. [45]

    iSDF: Real-time neural signed distance fields for robot perception

    Joseph Ortiz, Alexander Clegg, Jing Dong, Edgar Su- car, David Novotny, Michael Zollhoefer, and Mustafa Mukadam. iSDF: Real-time neural signed distance fields for robot perception. In Robotics: Science and Systems (R:SS), 2022. URL https://www.roboticsproceedings.org/ rss18/p0...

    2022

  38. [46]

    On the passivity-based impedance control of flexible joint robots

    Christian Ott, Alin Albu-Schaffer, Andreas Kugi, and Gerd Hirzinger. On the passivity-based impedance control of flexible joint robots. IEEE Transactions on Robotics (T-RO), 24(2):416–429, 2008. URL https://doi. org/10.1109/TRO.2008.915438. 12

    2008

  39. [47]

    DeepSDF: Learning continuous signed distance functions for shape represen- tation

    Jeong Joon Park, Peter Florence, Julian Straub, Richard Newcombe, and Steven Lovegrove. DeepSDF: Learning continuous signed distance functions for shape represen- tation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 165–174, June 2019. URL ht...

    2019

  40. [48]

    Online learning of piecewise polynomial signed distance fields for manipulation tasks, 2024

    Ante Mari ´c, Yiming Li, and Sylvain Calinon. Online learning of piecewise polynomial signed distance fields for manipulation tasks, 2024. URL https://arxiv.org/abs/ 2401.07698. 4

    2024 arXiv

  41. [49]

    Adaptive approxima- tion of signed distance fields through piecewise contin- uous interpolation

    Eduard Pujol and Antonio Chica. Adaptive approxima- tion of signed distance fields through piecewise contin- uous interpolation. Computers and Graphics , 114:337– 346, 2023. ISSN 0097-8493. doi: https://doi.org/10.1016/ j.cag.2023.06.020. URL https://www.sciencedirect.com/ sci...

    2023

  42. [50]

    Hilbert maps: Scal- able continuous occupancy mapping with stochas- tic gradient descent

    Fabio Ramos and Lionel Ott. Hilbert maps: Scal- able continuous occupancy mapping with stochas- tic gradient descent. The International Journal of Robotics Research (IJRR), 35(14):1717–1730, December

  43. [51]

    Euclideaniz- ing flows: Diffeomorphic reduction for learning stable dynamical systems

    Muhammad Asif Rana, Anqi Li, Dieter Fox, Byron Boots, Fabio Ramos, and Nathan Ratliff. Euclideaniz- ing flows: Diffeomorphic reduction for learning stable dynamical systems. In Conference on Learning for Dynamics and Control (L4DC) , pages 630–639, 2020. URL https://proceeding...

    2020

  44. [52]

    Learning con- tracting nonlinear dynamics from human demonstration for robot motion planning

    Harish Ravichandar and Ashwin Dani. Learning con- tracting nonlinear dynamics from human demonstration for robot motion planning. In ASME Dynamic Sys- tems, and Control Conference (DSCC) . American So- ciety of Mechanical Engineers, October 2015. doi: 10.1115/dscc2015-9870. UR...

    2015 doi

  45. [53]

    Polydoros, Sonia Chernova, and Aude Billard

    Harish Ravichandar, Athanasios S. Polydoros, Sonia Chernova, and Aude Billard. Recent advances in robot learning from demonstration. Annual Review of Control, Robotics, and Autonomous Systems , 3(1): 297–330, May 2020. ISSN 2573-5144. doi: 10.1146/ annurev-control-100819-06320...

    2020 doi

  46. [54]

    Elementary Differential Geometry

    Andrew Pressley. Elementary Differential Geometry . Springer London, 2010. ISBN 9781848828919. doi: 10.1007/978-1-84882-891-9. URL http://dx.doi.org/10. 1007/978-1-84882-891-9. 8

    2010 doi

  47. [55]

    Kia, and Negar Mehr

    Navid Rezazadeh, Maxwell Kolarich, Solmaz S. Kia, and Negar Mehr. Learning contraction policies from offline data. IEEE Robotics and Automation Letters (RA-L) , 7 (2):2905–2912, 2022. URL https://doi.org/10.1109/LRA. 2022.3145100. 2

    2022

  48. [56]

    Ivan Dario Jimenez Rodriguez, Noel Csomay-Shanklin, Yisong Yue, and Aaron D. Ames. Neural gaits: Learning bipedal locomotion via control barrier functions and zero dynamics policies. In Learning for Dynamics and Control Conference (L4DC) , volume 168, pages 1060– 1072, 2022. U...

    2022

  49. [57]

    Caldwell, Pablo Jim´enez, and Carme Torras

    Leonel Rozo, Sylvain Calinon, Darwin G. Caldwell, Pablo Jim´enez, and Carme Torras. Learning physical col- laborative robot behaviors from human demonstrations. IEEE Transactions on Robotics (T-RO) , 32(3):513–527,

  50. [58]

    Special issue on learning for human–robot collaboration

    Leonel Rozo, Heni Ben Amor, Sylvain Calinon, Anca Dragan, and Dongheui Lee. Special issue on learning for human–robot collaboration. Autonomous Robots , 42(5):953–956, 2018. URL https://doi.org/10.1007/ s10514-018-9756-z. 1

    2018

  51. [59]

    Kupcsik, Philipp Schillinger, Meng Guo, Robert Krug, Niels van Duijkeren, Markus Spies, Patrick Kesper, Sabrina Hoppe, Hanna Ziesche, Mathias B ¨urger, and Kai O

    Leonel Rozo, Andras G. Kupcsik, Philipp Schillinger, Meng Guo, Robert Krug, Niels van Duijkeren, Markus Spies, Patrick Kesper, Sabrina Hoppe, Hanna Ziesche, Mathias B ¨urger, and Kai O. Arras. The e-bike mo- tor assembly: Towards advanced robotic manipulation for flexible manu...

    2024 doi

  52. [60]

    Learning from demonstration

    Stefan Schaal. Learning from demonstration. In Neural Information Processing Systems (NeurIPS) , 1996. URL https://proceedings.neurips.cc/paper files/paper/1996/ file/68d13cf26c4b4f4f932e3eff990093ba-Paper.pdf. 1

    1996

  53. [61]

    Learn- ing position and orientation dynamics from demonstra- tions via contraction analysis

    Harish chaandar Ravichandar and Ashwin Dani. Learn- ing position and orientation dynamics from demonstra- tions via contraction analysis. Autonomous Robots , 43(4):897–912, May 2018. ISSN 1573-7527. doi: 10.1007/s10514-018-9758-x. URL http://dx.doi.org/10. 1007/s10514-018-9758-x. 1, 2

    2018 doi

  54. [62]

    Learning stabilizable nonlinear dynamics with contraction-based regularization

    Sumeet Singh, Spencer M Richards, Vikas Sindhwani, Jean-Jacques E Slotine, and Marco Pavone. Learning stabilizable nonlinear dynamics with contraction-based regularization. The International Journal of Robotics Research (IJRR) , 40(10-11):1123–1150, 2021. URL https://doi.org/1...

    2021 doi

  55. [63]

    M. W. Spong. Modeling and control of elastic joint robots. Journal of Dynamic Systems, Measurement, and Control, 109(4):310–318, December 1987. ISSN 1528-

    1987

  56. [64]

    Toward teaching by demonstration for robot-assisted minimally invasive surgery

    Hang Su, Andrea Mariani, Salih Ertug Ovur, Arianna Menciassi, Giancarlo Ferrigno, and Elena De Momi. Toward teaching by demonstration for robot-assisted minimally invasive surgery. IEEE Transactions on Au- tomation Science and Engineering, 18(2):484–494, 2021. URL https://doi....

    2021

  57. [65]

    URL https://doi.org/10.1109/TRO.2016.2540623. 1

    2016

  58. [66]

    Neural con- traction metrics for robust estimation and control: A convex optimization approach

    Hiroyasu Tsukamoto and Soon-Jo Chung. Neural con- traction metrics for robust estimation and control: A convex optimization approach. IEEE Control Systems Letters, 5(1):211–216, 2021. URL https://doi.org/10. 1109/LCSYS.2020.3001646. 2

    2021

  59. [67]

    Hiroyasu Tsukamoto, Soon-Jo Chung, and Jean- Jacques E. Slotine. Contraction theory for nonlinear stability analysis and learning-based control: A tutorial overview. Annual Reviews in Control, 52:135–169, 2021. ISSN 1367-5788. 1, 3

    2021

  60. [68]

    Urain, M

    J. Urain, M. Ginesi, D. Tateo, and J. Peters. Imitation- flow: Learning deep stable stochastic dynamic systems by normalizing flows. In International Conference on Intelligent Robots and Systems (IROS), pages 5231–5237. IEEE, 2020. URL https://doi.org/10.1109/IROS45743. 2020.9...

    2020

  61. [69]

    Simpson-Porco and Francesco Bullo

    John W. Simpson-Porco and Francesco Bullo. Con- traction theory on riemannian manifolds. Systems & Control Letters, 65:74–80, 2014. ISSN 0167-6911. URL https://doi.org/10.1016/j.sysconle.2013.12.016. 3

    2014 doi

  62. [70]

    Learning Riemannian stable dynamical systems via diffeomorphisms

    Jiechao Zhang, Hadi Beik Mohammadi, and Leonel Rozo. Learning Riemannian stable dynamical systems via diffeomorphisms. In Conference on Robot Learning (CoRL), 2022. URL https://openreview.net/forum?id= o8dLx8OVcNk. 2

    2022

  63. [71]

    Diffeomorphic transforms for generalised imitation learn- ing

    Weiming Zhi, Tin Lai, Lionel Ott, and Fabio Ramos. Diffeomorphic transforms for generalised imitation learn- ing. In Learning for Dynamics and Control Conference (L4DC), pages 508–519, 2022. URL https://proceedings. mlr.press/v168/zhi22a.html. 1, 2, 8, 9, 10, 11 APPENDIX A. Si...

    2022

  64. [74]

    Learning certified control using contraction metric

    Dawei Sun, Susmit Jha, and Chuchu Fan. Learning certified control using contraction metric. In Conference on Robot Learning (CoRL) , pages 1519–1539, 2020. URL https://proceedings.mlr.press/v155/sun21b.html. 2

    2020

  65. [78]

    Learning manipulation actions from human demonstrations

    Tim Welschehold, Christian Dornhege, and Wolfram Burgard. Learning manipulation actions from human demonstrations. In International Conference on Intel- ligent Robots and Systems (IROS) , pages 3772–3777,

  66. [79]

    URL https://doi.org/10.1109/IROS.2016.7759555. 1

    2016

  67. [82]

    Loss Functions: According to Park et al. [47], learned SDF methods with parameters θ can be interpreted as a decision boundary that assigns points to either the inside Hi or outside Hf of an object, while additionally providing the distance to the surface. To achieve these pro...

  68. [83]

    Under a high contraction rate, the green trajectory remains in its initial modal state

    Multi-modal demonstrations: Figure 14 displays the evaluation of the SDC on a multi-modal LASA dataset skill. Under a high contraction rate, the green trajectory remains in its initial modal state. Conversely, a reduced contraction rate, as evidenced by the purple trajectory, ...

  69. [84]

    Flow Solver: Table V shows a overview of different types of solvers that can be used to solve the flow in Eq. (24). Demonstrations NCDS SDC SDC 0.02 0.06 0.25 1.00 4.00 16.00 64.00 ‖V(x)‖ Figure 14: Obstacle avoidance with SDC using an inverse barrier with swept features on an...

  70. [85]

    (23) can violate contraction stability

    Reactive Obstacle Avoidance: The following is a proof which shows that reactive obstacle avoidance as in Eq. (23) can violate contraction stability. Proof: Consider the obstacle-avoidance term as de- scribed in Equation (23), we need to show the contraction conditions as descr...

  71. [86]

    The proof is inspired by Manchester and Slotine [41]

    Signed Distance Field Diffeomorphic Transform: In the following, we prove that the SDC preserves the contraction property and that the contraction metric is equivalent to the Riemannian metric of the obstacle-avoiding manifold Y. The proof is inspired by Manchester and Slotine...

  72. [87]

    In our context, τbase denotes the base trajectory of length l(τbase), and it is compared to a collision-free modulated trajectory τm of length l(τm)

    Evaluation Metrics: We use the DTWD [6] defined as, DTWD τbase(x),τm(x) =X j∈l(τbase(x)) min i∈l(τm(x)) d τi,base(x),τj,m(x) + X i∈l(τm(x)) min j∈l(τbase(x)) d τi,base(x),τj,m(x) , (58) measuring the similarity between two trajectories, which may vary in length and/or velocity...

  73. [88]

    Dishwasher Dataset: Figure 16 shows all the demon- strated skills used for the interaction with the dishwasher. Note that the last state in the demonstration is repeated N = 20 times to ensure that the contractive model learns a dynamic that reaches a target state at the end o...

  74. [89]

    The result is shown in Fig

    Robustness Test: In the following, the contraction be- havior and robustness of the Move Sine skill is tested by introducing a manual perturbation during skill execution. The result is shown in Fig. 15. The disturbances are observed with a velocity spike (Fig. 15b), a position...

  75. [90]

    17a shows the robot successfully executing the Place Plate in Rack skill

    Real World Skill Execution: Fig. 17a shows the robot successfully executing the Place Plate in Rack skill

  76. [91]

    17b and Fig

    Real World Obstacle Avoidance: Further experiments in Fig. 17b and Fig. 17c demonstrate the need to extend the robot’s SDF with the SDF of the grasped object

  77. [92]

    Clearly, the choice of the SDF is most critical

    Real World Inference time: Table VI shows the inference times of ODE solvers during real-world experiments, which approximately match the complexities O listed in Table V of the paper’s appendix. Clearly, the choice of the SDF is most critical. Solvers RDF CDF SDC Convex 20.9 ...

  78. [1992]

    doi: 10.1109/70.143352. 2

    doi

  79. [2012]

    URL http://dx.doi.org/10.1007/ s10514-012-9287-y

    ISSN 1573-7527. URL http://dx.doi.org/10.1007/ s10514-012-9287-y. 2

  80. [2014]

    ISBN 978-3-319-32552-1

    Springer International Publishing, Cham, 2016. ISBN 978-3-319-32552-1. URL https://doi.org/10.1007/ 978-3-319-32552-1 74. 1

    2016

  81. [2016]

    URL http://dx.doi.org/10.1177/ 0278364916684382

    ISSN 1741-3176. URL http://dx.doi.org/10.1177/ 0278364916684382. 9, 11, 12

  82. [2023]

    10341266

    URL http://dx.doi.org/10.1109/IROS55552.2023. 10341266. 6

    2023

  83. [2024]

    URL https://arxiv.org/abs/2402.00186. 14

    arXiv

  84. [9028]

    URL http://dx.doi.org/10.1115/1.3143860. 12

    doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.