CR-Solver: GPU-Accelerated Kinematics Solver for Tendon-driven Continuum Robots

Heqing Yang; LinJiang Huang; Linqing Zhong; Si Liu; Yang Yi

REVIEW 2 major objections 6 minor 34 references

CR-Solver: GPU-Accelerated Kinematics Solver for Tendon-driven Continuum Robots

T0 review · 2 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read CR-Solver claims GPU-accelerated, two-stage optimization can unify inverse kinematics, trajectory planning, and path following for tendon-driven continuum robots, reaching near-100% success and millimeter accuracy in under a second.

desk verdict Useful GPU-accelerated IK/planning tool for PCC continuum robots; the constraint-aware claims overstate what the penalty and sampled-collision implementation can guarantee. read the letter →

arxiv 2607.11340 v2 pith:ZCS3RIRU submitted 2026-07-13 cs.RO

Heqing Yang , Yang Yi , Linqing Zhong , Linjiang Huang , Si Liu This is my paper

classification cs.RO

keywords continuumrobotsinversekinematicstrajectoryplanningpathfollowingGPUaccelerationpiecewiseconstantcurvaturetendon-drivenactuationconstrainedoptimization

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 aims to close a practical gap: the motion-planning libraries that dominate rigid-body robotics cannot represent continuum robots, whose motion is continuous deformation, so continuum-robot users have been left with specialized, often slow solvers. CR-Solver is proposed as a single constrained nonlinear optimization framework that unifies inverse kinematics, trajectory planning, and path following for tendon-driven continuum robots, built on the piecewise constant curvature (PCC) model. The key claim is that a two-stage GPU-accelerated optimization strategy — massively parallel exploration of candidate solutions followed by gradient-based refinement of the best candidates — makes this unified solver fast, reliable, and accurate, with reported success rates near 100% and millimeter-level tracking error in sub-second runtimes. A sympathetic reader would care because if the claim holds, continuum robots gain a practical, accessible planning tool comparable in convenience to rigid-body planners, at a fraction of the development cost.

What carries the argument

The mechanism that carries the argument is a two-stage, GPU-batched optimization pipeline over the piecewise constant curvature (PCC) backbone model. PCC represents a tendon-driven continuum robot as a chain of constant-curvature arcs, each described by bending angle, azimuth, and arc length, so a configuration is a vector of these parameters; a tendon-length mapping translates those parameters into cable lengths, which are the physical controls. Stage one (exploration) generates and coarsely optimizes a large batch of candidate configurations in parallel — for IK this is random seed sampling, for trajectory planning it is a probabilistic roadmap with Dijkstra search, and for path following

What would settle it

Run CR-Solver on the cluttered 27-sphere IK benchmark, then recompute each returned configuration's true minimum distance to the obstacles and its joint-limit residuals from scratch (outside the solver). If any configuration counted as successful either violates a joint bound by more than a solver tolerance or has a minimum distance below the safety margin δ, then the penalty reformulation fails to enforce the constraints it claims, and the success-rate number loses its meaning. The same check can be applied mid-trajectory for the planning tasks.

Watch

Extended reading notes

Core claim

CR-Solver's central discovery is that the nonconvexity and initialization sensitivity of continuum-robot motion generation can be tamed by GPU parallelism: instead of running one optimizer on one guess, the solver runs hundreds of coarse optimizations in parallel on random seeds, selects the best candidates by cost, and then refines those with a batched trust-region Levenberg-Marquardt solver. Under the PCC kinematic model, this beam-search design reports 100% success on 1000-pose inverse kinematics benchmarks (both obstacle-free and cluttered), completes the collision-free IK benchmark in 0.30–0.46 s versus 2.35–269.20 s for CPU baselines, and unifies trajectory planning and path following

Load-bearing premise

The load-bearing premise is that the large penalty weights in the unconstrained reformulation make the bound and collision constraints effectively hard, so a solution that the optimizer reports as 'successful' genuinely respects joint limits and avoids obstacles — a premise the authors themselves qualify by noting the reformulation does not guarantee adherence to all hard constraints.

Editorial extensions

If this is right

If the reported results transfer to practice, continuum-robot users gain a real-time tool: a 1000-pose IK batch completes in 0.46 s on one GPU and each path-following solve takes under 125 ms.
A single solver handles IK, trajectory planning, and path following, so a roboticist no longer needs to assemble different algorithms for different motion-generation subproblems.
The ablation, where removing beam search drops IK success from 100% to 44%, implies that the parallel exploration stage is load-bearing for reliability, not just speed.
Because the framework is model-based and configurable in segment count and extensibility, it applies to any tendon-driven robot that fits the constant-curvature assumption, with no per-robot reimplementation.
The roadmap for trajectory planning is built offline once and reused for online queries, which suggests a practical path to interactive replanning, though the paper does not test dynamic environments.

Reading between the lines

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

A likely extension is replacing the PCC backbone with higher-fidelity models such as Cosserat rods or discrete elastic rods; the two-stage GPU pattern would still apply if those models provide differentiable forward kinematics, but the paper explicitly leaves this open.
The accuracy numbers are relative to the simulated PCC model; on a physical robot, tendon stretch, friction, and backbone stiffness would introduce errors that the solver does not model, so the millimeter claim should be read as simulation-level until hardware validation exists.
Because constraints are penalties rather than hard floors, a safety-critical deployment would need an independent verification pass that checks joint limits and obstacle distance on the returned solution.
The pure-Python implementation plus batched inference makes the framework a plausible fit for perception-in-the-loop control on embedded GPUs, but the paper does not measure end-to-end latency or robustness to sensor noise.

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

First thing you should know: this is a useful engineering paper, not a scientific breakthrough. The authors take standard PCC kinematics, standard PRM and LM optimization, and batch everything in JAX on a GPU, producing a single solver for IK, trajectory planning, and path following of tendon-driven continuum robots. That integration is genuinely new for this subfield, and the speedups over CPU baselines look real on their benchmarks. The beam search ablation is a nice touch and the formulation is clearly written.

Now the soft spots, in proportion. The biggest one is that the constraint-aware claims are stated more strongly than the implementation supports. The paper solves Eq. 17, a penalty reformulation with unpublished weights ρ_ineq and ρ_eq and an unpublished safety margin δ, and the authors admit in Section VI that this 'does not guarantee adherence to all hard constraints.' Success rate and collision-freedom are then judged using the same discretized distance function, d_min(q), computed from 10 points per 1-meter segment. At max bending this gives ~0.1 m spacing, so a 0.2 m-radius obstacle can sit between sample points and go undetected. The continuous robot could penetrate an obstacle while every sampled point is collision-free. That means the reported 100% success rates do not establish collision-free solutions for the actual continuous robot. This is not a niche worry; it lands directly on the abstract's 'constraint-aware' and 'collision-free' language.

Other issues are more minor. There are no error bars or repeated runs, and the baselines are partly reimplementations, which is acceptable for a tool paper but should be stated more carefully. The code availability is inconsistent: the abstract says it is publicly available on GitHub, the conclusion says 'we will release' it. That needs to be resolved. The paper would also be stronger with a sensitivity analysis over the penalty weights and the collision discretization resolution.

Credit where it's due: the authors are candid about their limitations, the problem formulations are clean, and the empirical comparisons, while self-benchmarked, are at least designed to be meaningful. The GPU speedups are plausible and the two-stage exploration/refinement idea is a reasonable recipe for the niche.

My take: this paper deserves a serious referee. A skeptical but constructive review could ask for the missing hyperparameters, a conservative continuous collision re-check, repeated runs, and a cleanup of the code-release statement. That is a well-scoped revision, not a fundamental rework.

Referee Report

2 major / 6 minor

Summary. The paper presents CR-Solver, a JAX-based, GPU-accelerated optimization solver for inverse kinematics, trajectory planning, and path following of tendon-driven continuum robots modeled with piecewise constant curvature (PCC). The method casts all three tasks into a single constrained nonlinear program (Eqs. 6-16), relaxes constraints into penalty terms with 'large weights' (Eq. 17), and uses a two-stage pipeline: parallel sampling/roadmap construction/IK to generate diverse initializations, followed by batched Levenberg-Marquardt refinement on the GPU. Experiments on simulated 3- and 4-segment robots report 100% IK success, 95%+ trajectory planning success, millimeter-level path-following accuracy, and large speedups over CPU baselines (Micsolver, CIDGIKc, RRT). The code is promised as open source.

Significance. If the reported results are trustworthy, CR-Solver would be a valuable, accessible software contribution: it is one of the first GPU-parallel toolchains for continuum-robot motion generation, provides a clean unification of three common tasks, and includes a clear ablation of the beam-search initialization. The use of pure Python/JAX and public code lowers the adoption barrier. However, the significance is currently tempered by three issues: the collision-free and constraint-satisfaction claims are only as good as the 10-point discretization and unpublished penalty weights; the experimental success rates are single-run point estimates on filtered benchmarks; and the baseline comparisons are narrow, partly author reimplementations. The paper is a useful systems contribution, but its headline guarantees need verification before the claims can be accepted at face value.

major comments (2)

[§V-B, Eq. (10), Eq. (17)] The central claim of constraint-aware, collision-free solutions is not established for the continuous robot. The collision check uses 10 discretization points per constant-curvature segment, and the success metric uses that same d_min. For a 1.0 m segment at θ=π, adjacent sample points are ~0.11 m apart; a 0.2 m-radius obstacle sphere centered 0.304 m from the chord line at its midpoint is 0.309 m from both sampled points but only 0.299 m from the arc midpoint, so with the 0.1 m robot radius the discretized check passes while the continuous backbone penetrates the obstacle. Because the penalty weights ρ_ineq/ρ_eq in Eq. (17) and the safety margin δ in Eq. (10) are not reported, and because Section VI concedes the penalty form 'does not guarantee adherence to all hard constraints,' the reported 100%/95% success rates need verification with a conservative or continuous collision check and
[§V, Tables I-III, Fig. 6] The success-rate and speedup claims are single-run point estimates. The pipeline involves random sampling (seeds, roadmap nodes, start/goal selection), so repeated runs would vary; without error bars or confidence intervals the 95-100% success rates are not statistically supported. The trajectory benchmark filters start/goal pairs for feasibility (Sec. V-C), making the rates conditional on that filter, and the baselines are partly author reimplementations, with Micsolver run sequentially on CPU while CR-Solver batches 1000 targets on the GPU in Tables I-II. Please report multiple seeds and per-query latencies in addition to batch totals, and state exactly which filtering steps are applied before reporting success rates.

minor comments (6)

[§I] The 'first solver in continuum robotics to leverage GPU acceleration' claim is made with a 'to the best of our knowledge' qualifier, but no supporting literature search is given. Please provide a focused related-work search or soften the claim.
[§III-B] Typo: 'mtendons' should be 'm tendons'. Also 'purePython' appears in several places.
[§V-B] If the 10-point discretization is used only for collision detection and not for forward kinematics, the text should say so explicitly; the phrase 'discretized with 10 points for forward kinematics computation and collision detection' is ambiguous and could imply the FK itself is approximate.
[§V-C, Fig. 6] The text says 'for three iterations' but the figure sweeps the optimization iteration budget; please reconcile and clarify the iteration budget in the CPU comparison.
[Table III, Fig. 7] The 'boxplots' appear to be per-waypoint errors from a single execution, not repeated trials; label them accordingly to avoid implying statistical repetition.
[Author line] The author line contains a stray colon and inconsistent spacing (e.g., 'Linqing Zhong 1: Linjiang Huang 1: Si Liu 1:'); please fix.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation; empirical system paper with validation caveats.

full rationale

CR-Solver is an empirical systems paper: its headline results (Tables I-III, Fig. 6) are measured runtimes, success rates, and tracking errors on self-constructed benchmarks, not quantities derived from fitted parameters. I checked the three mathematical components: the IK objective (Eq. 8), penalty reformulation (Eq. 17), and trajectory/path objectives (Eqs. 11/16) are standard optimization formulations; no constant is fitted on a subset of the data and then 'predicted'; no load-bearing premise depends on a self-citation (the cited baselines [10],[23] are Qiu et al., not the present authors). The Sec. VI admission that the penalty reformulation 'does not guarantee adherence to all hard constraints' and the 10-point-per-segment collision discretization are legitimate validation/correctness caveats, but they affect whether the benchmarks establish continuous collision-freedom, not whether the paper's argument reduces to its inputs by construction. There is therefore no equation- or citation-level circularity; score 0.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The central claims rest on the PCC simulation model, unstated optimizer penalties, and hand-chosen hyperparameters. No new physical entities are introduced; the two-stage GPU solver is software architecture, not an invented entity. The absence of reported weight values and sensitivity analyses is the main ledger concern.

free parameters (5)

Penalty weights ρ_ineq, ρ_eq = not reported
Eq. 17; must be large enough that the reported 100% success and collision-free claims hold. No values or sensitivity analysis are given.
Safety margin δ = not reported
Eqs. 10 and 15; controls distance-to-obstacle conservatism and is central to the collision-aware success rate.
Beam search hyperparameters (N_init, N_final, k_init, k_final) = N_init=128, N_final=8, budgets totaling ≤200 iterations
Algorithm 1; chosen by hand. The ablation only compares with and without beam search, not across these values.
Roadmap size N, k-nearest neighbors k, trajectory horizon H = not reported
Algorithm 2; success/runtime trade-off depends on these values, with no sensitivity analysis.
Collision discretization points per segment = 10
Sec. V-B; chosen without convergence or sensitivity analysis. Inter-segment collisions could be missed at this resolution.

assumptions (6)

domain assumption The PCC model (piecewise constant curvature segments, Eqs. 1-3) adequately represents tendon-driven continuum robots for motion generation.
All accuracy and collision results are computed on this simulated model; real tendon robots deviate due to backbone elasticity, tendon slack, and gravity.
domain assumption Eq. 4's tendon-length mapping ℓij = L_i - r_c θ_i cos(α_j - φ_i) is the correct actuation constraint.
Tendon speed/acceleration limits and actuation feasibility depend on this linearized routing model being accurate.
ad hoc to paper Discretizing each segment with 10 points is sufficient to guarantee collision-free motion.
The number is chosen in Sec. V-B without sensitivity analysis; collisions between sampled points can be missed.
domain assumption Convex decomposition of obstacles gives correct, smooth collision gradients in batched distance queries.
Relies on reference [34] and precomputation; the approximation can alter narrow passages and affect feasibility.
standard math Batched trust-region Levenberg-Marquardt converges to a solution of Eq. 17 within the iteration budget.
No convergence proof or failure analysis is provided; JAX autodiff and dense Cholesky are assumed correct.
ad hoc to paper The penalty weights ρ in Eq. 17 are large enough to make constraint violations negligible.
Values are not reported, and Section VI explicitly states that hard constraints are not guaranteed.

how reviews work

0 comments

Cite this review

Pith. "Pith review of CR-Solver: GPU-Accelerated Kinematics Solver for Tendon-driven Continuum Robots." pith.science (2026). https://pith.science/paper/ZCS3RIRU

@misc{pith2026260711340,
  author       = {Pith},
  title        = {Pith review of: CR-Solver: GPU-Accelerated Kinematics Solver for Tendon-driven Continuum Robots},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZCS3RIRU}},
  note         = {Machine review of arXiv:2607.11340}
}

read the original abstract

Continuum robots provide intrinsic compliance, high dexterity, and safe physical interaction, enabling navigation and manipulation in confined and unstructured environments. Despite recent advances in sensing and control, heightening the need for precise motion generation, most widely used planning libraries are grounded in rigid-body assumptions, creating a critical gap for fast and practical tools for continuum robots. To address this, we present CR-Solver, a two-stage, optimization-based solver for the motion generation of tendon-driven continuum robots. Our method unifies inverse kinematics, path following, and trajectory planning within a single constrained nonlinear optimization framework. Leveraging GPU-accelerated parallel optimization, CR-Solver delivers fast, accurate, and constraint-aware solutions. We validate our approach on three tasks, demonstrating significant speedups over traditional CPU-based solvers while achieving a consistently high success rate above 95% and millimeter-level accuracy. The solver is implemented in pure Python, reducing the barrier to adoption and offering a practical, extensible foundation for continuum robots' high-performance motion planning.

Figures

Figures reproduced from arXiv: 2607.11340 by the authors.

**Figure 2.** Kinematic model of a single segment in a tendon-driven [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Framework of our CR-Solver. We employ a GPU-accelerated two-stage pipeline: Stage I performs massively parallel seed sampling, roadmap searching, and parallel IK to generate diverse feasible initializations; Stage II uses a batched gradient-based solver on a GPU to refine selected candidates, solving inverse kinematics, trajectory planning, and path following under a unified constrained optimization. the homogeneou… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Inverse kinematics solutions for non-extendable (left) and [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 6.** Figure 6: Ablation study on trajectory planning with different initial [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

**Figure 7.** Figure 7: The results of tracking the ICRA-shaped paths. Top row: the robot traces the letters I, C, R, and A while avoiding obstacles, showing [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Review history (2 revisions) →

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 2 linked inside Pith

[1]

Continuum robots: An overview,

M. Russo, S. M. H. Sadati, X. Dong, A. Mohammad, I. D. Walker, C. Bergeles, K. Xu, and D. A. Axinte, “Continuum robots: An overview,”Advanced Intelligent Systems, vol. 5, no. 5, p. 2200367, 2023

2023
[2]

Design and characterization of a novel, continuum-robot surface for the human environment,

R. Sirohi, Y . Wang, S. Hollenberg, I. S. G. Keith, I. D. Walker, and E. Green, “Design and characterization of a novel, continuum-robot surface for the human environment,” in2019 IEEE 15th Interna- tional Conference on Automation Science and Engineering (CASE), pp. 1169–1174, IEEE, 2019

2019
[3]

A multi- modal sensor array for human–robot interaction and confined spaces exploration using continuum robots,

C. Abah, A. L. Orekhov, G. L. Johnston, and N. Simaan, “A multi- modal sensor array for human–robot interaction and confined spaces exploration using continuum robots,”IEEE sensors journal, vol. 22, no. 4, pp. 3585–3594, 2021

2021
[4]

Continuum robots for medical applications: A survey,

J. Burgner-Kahrs, D. C. Rucker, and H. Choset, “Continuum robots for medical applications: A survey,”IEEE transactions on robotics, vol. 31, no. 6, pp. 1261–1280, 2015

2015
[5]

Continuum robots for manipu- lation applications: A survey,

S. Kolachalama and S. Lakshmanan, “Continuum robots for manipu- lation applications: A survey,”Journal of Robotics, vol. 2020, no. 1, p. 4187048, 2020

2020
[6]

The open motion planning library,

I. A. Sucan, M. Moll, and L. E. Kavraki, “The open motion planning library,”IEEE Robotics & Automation Magazine, vol. 19, no. 4, pp. 72–82, 2012

2012
[7]

Curobo: Parallelized collision-free robot motion generation,

B. Sundaralingam, S. K. S. Hari, A. Fishman, C. Garrett, K. Van Wyk, V . Blukis, A. Millane, H. Oleynikova, A. Handa, F. Ramos,et al., “Curobo: Parallelized collision-free robot motion generation,” in2023 IEEE International Conference on Robotics and Automation (ICRA), pp. 8112–8119, IEEE, 2023

2023
[8]

Pyroki: A modular toolkit for robot kinematic optimization,

C. M. Kim, B. Yi, H. Choi, Y . Ma, K. Goldberg, and A. Kanazawa, “Pyroki: A modular toolkit for robot kinematic optimization,”arXiv preprint arXiv:2505.03728, 2025

arXiv 2025

Show all 34 references

[9]

Understanding urdf: A survey based on user experience,

D. Tola and P. Corke, “Understanding urdf: A survey based on user experience,” in2023 IEEE 19th International Conference on Automation Science and Engineering (CASE), pp. 1–7, IEEE, 2023

2023
[10]

An actuator space optimal kinematic path tracking framework for tendon- driven continuum robots: Theory, algorithm and validation,

K. Qiu, H. Zhang, J. Zhang, R. Xiong, H. Lu, and Y . Wang, “An actuator space optimal kinematic path tracking framework for tendon- driven continuum robots: Theory, algorithm and validation,”The International Journal of Robotics Research, vol. 44, no. 6, pp. 1006– 1034, 2025

2025
[11]

Planning high-quality motions for concentric tube robots in point clouds via parallel sampling and optimization,

A. Kuntz, M. Fu, and R. Alterovitz, “Planning high-quality motions for concentric tube robots in point clouds via parallel sampling and optimization,” in2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 2205–2212, IEEE, 2019

2019
[12]

Cidgikc: Distance-geometric inverse kinematics for continuum robots,

H. J. Zhang, M. Giamou, F. Mari ´c, J. Kelly, and J. Burgner- Kahrs, “Cidgikc: Distance-geometric inverse kinematics for continuum robots,”IEEE Robotics and Automation Letters, vol. 8, no. 11, pp. 7679–7686, 2023

2023
[13]

JAX: composable transformations of Python+NumPy programs,

J. Bradbury, R. Frostig, P. Hawkins, M. J. Johnson, C. Leary, D. Maclaurin, G. Necula, A. Paszke, J. VanderPlas, S. Wanderman- Milne, and Q. Zhang, “JAX: composable transformations of Python+NumPy programs,” 2018

2018
[14]

Forward and inverse problems in the mechanics of soft filaments,

M. Gazzola, L. H. Dudte, A. G. McCormick, and L. Mahadevan, “Forward and inverse problems in the mechanics of soft filaments,” Royal Society open science, vol. 5, no. 6, p. 171628, 2018

2018
[15]

Modeling and simulation of complex dynamic musculoskeletal architectures,

X. Zhang, F. K. Chan, T. Parthasarathy, and M. Gazzola, “Modeling and simulation of complex dynamic musculoskeletal architectures,” Nature communications, vol. 10, no. 1, p. 4825, 2019

2019
[16]

A dynamic model for concentric tube robots,

J. Till, V . Aloi, K. E. Riojas, P. L. Anderson, R. J. Webster III, and C. Rucker, “A dynamic model for concentric tube robots,”IEEE Transactions on Robotics, vol. 36, no. 6, pp. 1704–1718, 2020

2020
[17]

Dynamically closed-loop controlled soft robotic arm using a reduced order finite element model with state observer,

R. K. Katzschmann, M. Thieffry, O. Goury, A. Kruszewski, T.-M. Guerra, C. Duriez, and D. Rus, “Dynamically closed-loop controlled soft robotic arm using a reduced order finite element model with state observer,” in2019 2nd IEEE international conference on soft robotics (RoboSo...

2019
[18]

Design and kinematic modeling of constant curvature continuum robots: A review,

R. J. Webster III and B. A. Jones, “Design and kinematic modeling of constant curvature continuum robots: A review,”The International Journal of Robotics Research, vol. 29, no. 13, pp. 1661–1683, 2010

2010
[19]

Closed- form inverse kinematics for continuum manipulators,

S. Neppalli, M. A. Csencsits, B. A. Jones, and I. D. Walker, “Closed- form inverse kinematics for continuum manipulators,”Advanced robotics, vol. 23, no. 15, pp. 2077–2091, 2009
[20]

Kinematics of continuum robots with constant curvature bending and extension capabilities,

A. Garriga-Casanovas and F. Rodriguez y Baena, “Kinematics of continuum robots with constant curvature bending and extension capabilities,”Journal of Mechanisms and Robotics, vol. 11, no. 1, p. 011010, 2019

2019
[21]

Fabrikc: An efficient iterative inverse kinematics solver for continuum robots,

W. Zhang, Z. Yang, T. Dong, and K. Xu, “Fabrikc: An efficient iterative inverse kinematics solver for continuum robots,” in2018 IEEE/ASME International Conference on Advanced Intelligent Mecha- tronics (AIM), pp. 346–352, IEEE, 2018

2018
[22]

Fabrikx: Tackling the inverse kinematics problem of continuum robots with variable curvature,

D. Kolpashchikov, O. Gerget, and V . Danilov, “Fabrikx: Tackling the inverse kinematics problem of continuum robots with variable curvature,”Robotics, vol. 11, no. 6, p. 128, 2022

2022
[23]

An efficient multi-solution solver for the inverse kinematics of 3-section constant- curvature robots,

K. Qiu, J. Zhang, D. Sun, R. Xiong, H. Lu, and Y . Wang, “An efficient multi-solution solver for the inverse kinematics of 3-section constant- curvature robots,”arXiv preprint arXiv:2305.01458, 2023

2023 arXiv
[24]

Prob- abilistic roadmaps for path planning in high-dimensional configuration spaces,

L. E. Kavraki, P. Svestka, J.-C. Latombe, and M. H. Overmars, “Prob- abilistic roadmaps for path planning in high-dimensional configuration spaces,”IEEE transactions on Robotics and Automation, vol. 12, no. 4, pp. 566–580, 2002

2002
[25]

Randomized kinodynamic plan- ning,

S. M. LaValle and J. J. Kuffner Jr, “Randomized kinodynamic plan- ning,”The international journal of robotics research, vol. 20, no. 5, pp. 378–400, 2001

2001
[26]

Sampling-based algorithms for optimal motion planning,

S. Karaman and E. Frazzoli, “Sampling-based algorithms for optimal motion planning,”The international journal of robotics research, vol. 30, no. 7, pp. 846–894, 2011

2011
[27]

Finding locally optimal, collision-free trajectories with sequential convex optimization.,

J. Schulman, J. Ho, A. X. Lee, I. Awwal, H. Bradlow, and P. Abbeel, “Finding locally optimal, collision-free trajectories with sequential convex optimization.,” inRobotics: science and systems, vol. 9, pp. 1– 10, Berlin, Germany, 2013

2013
[28]

Itomp: Incremental trajectory optimization for real-time replanning in dynamic environments,

C. Park, J. Pan, and D. Manocha, “Itomp: Incremental trajectory optimization for real-time replanning in dynamic environments,” in Proceedings of the international conference on automated planning and scheduling, vol. 22, pp. 207–215, 2012

2012
[29]

Chomp: Gradient optimization techniques for efficient motion planning,

N. Ratliff, M. Zucker, J. A. Bagnell, and S. Srinivasa, “Chomp: Gradient optimization techniques for efficient motion planning,” in 2009 IEEE international conference on robotics and automation, pp. 489–494, IEEE, 2009

2009
[30]

Efficient rrt*-based safety-constrained motion planning for continuum robots in dynamic environments,

P. Luo, S. Yao, Y . Yue, J. Wang, H. Yan, and M. Q.-H. Meng, “Efficient rrt*-based safety-constrained motion planning for continuum robots in dynamic environments,” in2024 IEEE International Conference on Robotics and Automation (ICRA), pp. 9328–9334, IEEE, 2024

2024
[31]

Constrained motion planning of a cable-driven soft robot with compressible curvature modeling,

J. Lai, B. Lu, Q. Zhao, and H. K. Chu, “Constrained motion planning of a cable-driven soft robot with compressible curvature modeling,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 4813–4820, 2022

2022
[32]

A note on two problems in connexion with graphs,

E. W. Dijkstra, “A note on two problems in connexion with graphs,” in Edsger Wybe Dijkstra: his life, work, and legacy, pp. 287–290, 2022

2022
[33]

The levenberg-marquardt algorithm: implementation and theory,

J. J. Mor ´e, “The levenberg-marquardt algorithm: implementation and theory,” inNumerical analysis: proceedings of the biennial Conference held at Dundee, June 28–July 1, 1977, pp. 105–116, Springer, 2006

1977
[34]

Approximate convex decompo- sition for 3d meshes with collision-aware concavity and tree search,

X. Wei, M. Liu, Z. Ling, and H. Su, “Approximate convex decompo- sition for 3d meshes with collision-aware concavity and tree search,” ACM Transactions on Graphics (TOG), vol. 41, no. 4, pp. 1–18, 2022

2022

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.

[1] [1]

Continuum robots: An overview,

M. Russo, S. M. H. Sadati, X. Dong, A. Mohammad, I. D. Walker, C. Bergeles, K. Xu, and D. A. Axinte, “Continuum robots: An overview,”Advanced Intelligent Systems, vol. 5, no. 5, p. 2200367, 2023

2023

[2] [2]

Design and characterization of a novel, continuum-robot surface for the human environment,

R. Sirohi, Y . Wang, S. Hollenberg, I. S. G. Keith, I. D. Walker, and E. Green, “Design and characterization of a novel, continuum-robot surface for the human environment,” in2019 IEEE 15th Interna- tional Conference on Automation Science and Engineering (CASE), pp. 1169–1174, IEEE, 2019

2019

[3] [3]

A multi- modal sensor array for human–robot interaction and confined spaces exploration using continuum robots,

C. Abah, A. L. Orekhov, G. L. Johnston, and N. Simaan, “A multi- modal sensor array for human–robot interaction and confined spaces exploration using continuum robots,”IEEE sensors journal, vol. 22, no. 4, pp. 3585–3594, 2021

2021

[4] [4]

Continuum robots for medical applications: A survey,

J. Burgner-Kahrs, D. C. Rucker, and H. Choset, “Continuum robots for medical applications: A survey,”IEEE transactions on robotics, vol. 31, no. 6, pp. 1261–1280, 2015

2015

[5] [5]

Continuum robots for manipu- lation applications: A survey,

S. Kolachalama and S. Lakshmanan, “Continuum robots for manipu- lation applications: A survey,”Journal of Robotics, vol. 2020, no. 1, p. 4187048, 2020

2020

[6] [6]

The open motion planning library,

I. A. Sucan, M. Moll, and L. E. Kavraki, “The open motion planning library,”IEEE Robotics & Automation Magazine, vol. 19, no. 4, pp. 72–82, 2012

2012

[7] [7]

Curobo: Parallelized collision-free robot motion generation,

B. Sundaralingam, S. K. S. Hari, A. Fishman, C. Garrett, K. Van Wyk, V . Blukis, A. Millane, H. Oleynikova, A. Handa, F. Ramos,et al., “Curobo: Parallelized collision-free robot motion generation,” in2023 IEEE International Conference on Robotics and Automation (ICRA), pp. 8112–8119, IEEE, 2023

2023

[8] [8]

Pyroki: A modular toolkit for robot kinematic optimization,

C. M. Kim, B. Yi, H. Choi, Y . Ma, K. Goldberg, and A. Kanazawa, “Pyroki: A modular toolkit for robot kinematic optimization,”arXiv preprint arXiv:2505.03728, 2025

arXiv 2025