Pith. sign in

REVIEW 3 major objections 4 minor 73 references

Manifold-Guided Motion Planning for Tight Assemblies

T0 review · 3 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper claims that biasing samples toward the critical contact manifold makes sampling-based planning for tight assemblies probabilistically complete and fast enough to solve the Elk puzzle automatically for the first time.

desk verdict The Elk result and the speedup look real, but the completeness proof has a load-bearing gap: the algorithm prunes away the wide-clearance regions it claims to handle, so the main theorem does not follow as written. read the letter →

arxiv 2607.17898 v1 pith:WFJFRSO7 submitted 2026-07-20 cs.RO

Dror Livnat , Michael M. Bilevich , Michal Kleinbort , Dan Halperin This is my paper · ORCID
classification cs.RO
keywords motionplanningtightassemblycriticalmanifoldsampling-basedprobabilisticcompletenesssigneddistancefunctionadaptivesubdivisionRRT
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 targets motion planning for tight assemblies, where valid paths must squeeze through near-zero-clearance configurations. The central claim is that such paths lie on or near a critical manifold—configurations where the two parts touch—so a planner can concentrate its random samples in a shrinking neighborhood of that manifold using adaptive subdivision, while retaining enough exploration for wider stretches. The paper proves that the resulting planner, CMG-RRT, is probabilistically complete: the probability of failure after k iterations decays exponentially under standard clearance assumptions. Empirically, it reports 100% success on a 16-instance rotational assembly benchmark, averaging 3.0 minutes, and the first fully automatic solution of the Elk disentanglement puzzle, averaging 156 minutes. A sympathetic reader would care because this addresses the longstanding narrow-passage bottleneck of sampling-based planners without relying on handcrafted geometric features.

What carries the argument

The critical manifold is the subset of SE(3) consisting of poses in which the moving part has at least one contact point with the static part. Around it, the algorithm maintains a hierarchical subdivision: boxes in configuration space are bisected along their longest axis, and a box is discarded if its center's signed distance to the obstacle is at least D/2, where D is the box diagonal. Sampling selects a surviving box uniformly and then a configuration uniformly inside it; a random rotation step restores convexity of the pseudo-metric balls used in the proof. The proof itself uses a pseudo-metric d_C that blends Euclidean translation distance with circular distances on the three Euler angl

What would settle it

Construct a valid path with a long wide-clearance segment connecting two tight tunnels, run CMG-RRT, and after refinement check whether any box covering that wide segment survives the |F(center)| < D/2 test. If all such boxes are discarded, sampling support on the segment is empty and no sequence of iterations can cross it, contradicting the claimed exponential decay for that instance.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the hard part of tight assembly planning is not the local extension step but the distribution of samples. CMG-RRT replaces uniform sampling in SE(3) with an adaptive subdivision that keeps only axis-aligned boxes whose center's signed distance to the other part is less than half the box diagonal, so surviving boxes shrink around the critical manifold. Sampling uniformly from these boxes, interleaved with random rotations of the whole system, concentrates effort near contact. The paper proves that under standard clearance assumptions, the probability that CMG-RRT fails to reach the goal after k iterations is at most a e^{-bk}, and it demonstrates empirica

Load-bearing premise

The completeness proof assumes that after refinement every ball along the solution path still contains at least one box kept by the pruning rule, but the rule deliberately discards boxes in wide-clearance regions, so on a mixed wide-tight path the assumed positive sampling probability can drop to zero.

Editorial extensions

If this is right

  • Probabilistic completeness transfers to SE(3) with Euler-angle parametrization, giving an exponential decay bound on failure probability rather than only a qualitative guarantee.
  • Tight assembly planning no longer needs problem-specific geometric cues; the only required oracle is a signed-distance query, which is standard in collision-checking pipelines.
  • Mixed wide-tight paths become tractable because refinement is interleaved with search: easy stretches pay little overhead, while hard stretches automatically trigger finer sampling.
  • Planned paths can be physically executed: the paper reports a qualitative two-arm robotic execution of a CMG-RRT trajectory via a joint-space IK conversion.

Reading between the lines

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

  • If the central claim holds, the same subdivision-and-prune principle should extend to multi-part assemblies, where coordinated contacts among three or more parts implicitly restrict feasible motion to a small, structured subset of configuration space.
  • The proof's dependence on every ball containing a kept box suggests a testable extension: modify the pruning rule to retain a minimal set of wide-clearance boxes, then check whether the exponential failure bound becomes provable for paths with long wide segments.
  • The critical-manifold bias is conceptually similar to medial-axis sampling but in configuration space; one could test whether the |F(center)| < D/2 rule approximates the free-space medial axis in tight regions, which would link this method to a broader family of topology-guided samplers.
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.

Referee Report

3 major / 4 minor

Summary. The paper presents CMG-RRT, a sampling-based motion planner for tight assembly tasks. The planner maintains an adaptive subdivision of SE(3) into boxes, prunes boxes whose centers are not near the contact critical manifold using an SDF oracle, and samples only from the remaining boxes. The authors prove probabilistic completeness (Theorem 4.3) and report experiments on 16 rotational assembly instances plus the Elk puzzle, including a claimed first automatic solution of Elk. The paper also contributes a randomized-rotation scheme intended to make local balls convex, and it provides open-source code.

Significance. If the completeness theorem and the experimental results hold, this would be a valuable contribution: it targets mixed wide-tight assembly problems, gives an adaptive manifold-guided sampling scheme, and demonstrates strong empirical performance, including the first automatic Elk solution and a claimed 100% success rate on the rotational benchmark subset. The open-source release is a concrete strength. However, the central theoretical guarantee is not established by the submitted proof. The completeness argument fails for the algorithm as written, and the exponential-rate bound relies on assumptions that are not met. Since the paper's main contribution includes the probabilistic completeness proof, the manuscript requires substantial revision.

major comments (3)
  1. [Section 4, proof of Lemma 4.2; Algorithm 2] The assertion that after sufficient refinement 'each such ball will contain at least one box b∈B' is not supported by Algorithm 2. A child box is retained only when the SDF at its center satisfies d < D/2 (Algorithm 2, line 8; the text in §3.2 writes |F| < D/2, but the pseudocode omits the absolute value). For any path segment lying in a wide-clearance region (|F_M2→M1| ≫ D/2), every sufficiently small box centered there is discarded. The paper explicitly targets 'mixed wide-tight' paths (Section 1, Figure 1), which contain wide free-space segments. On those segments the retained box set has empty intersection with the covering balls, so p_sample = 0 and the proof's positivity condition p = p_rotate · p_sample > 0 fails. Moreover, because Algorithm 1 interleaves refinement with search, the box set can lose coverage of wide regions mid-run; nothing guarantees that B remains nonempty, alth
  2. [Theorem 4.3 and Algorithm 1] The proof treats p as a fixed constant independent of k, but p_sample depends on the current box set B: refinement changes the box diagonal D, the number of boxes, and the total volume of B, and RandomRotate changes the coordinate frame from which samples are drawn. No uniform lower bound on p_sample over the course of the run is established. In fact, as D decreases, p_sample = |b|/|B| can shrink arbitrarily, and on wide-clearance segments it becomes exactly zero. Therefore the Bernoulli-trial argument and the exponential bound a e^{-bk} do not follow.
  3. [Section 4, proof of Lemma 4.2] The proof assumes that q_near,q_rand ∈ B_ν(q_i) ⊆ F^δ and concludes that the straight segment q_near q_rand lies in F^δ. Membership of q_i in F^δ only gives F_M2→M1(q_i) > -δ. Points at d_C-distance ν from q_i can have substantially deeper penetration unless an explicit Lipschitz/clearance assumption is stated. Thus the inclusion B_ν(q_i) ⊆ F^δ is not justified by the definitions in Section 2.4, and the local extension step may not be collision-free.
minor comments (4)
  1. [Algorithm 2, line 8] The pseudocode uses d < D/2 while the text in Section 3.2 says |F_M2→M1(Center(b))| < D/2. The pseudocode as written retains boxes deep inside the obstacle, since for large negative SDF values d < D/2 holds; this is inconsistent with the intended manifold bias and should be corrected.
  2. [Algorithm 1 and Lemma 4.1] RandomRotate() is called without arguments, but the proof of Lemma 4.1 applies a rotation R to a specific configuration q to obtain q'=R·q. Please clarify whether RandomRotate rotates the entire coordinate frame, the robot, or the box subdivision, and how this maps to the proof's notation.
  3. [Section 5.2, Table 1] The claim of 100% success for CMG-RRT, especially for the Elk puzzle, should state the number of runs and the variance across runs. For Elk only an average time of 156 minutes is given; without trial counts the success rate is not fully interpretable.
  4. [Section 4, general notation] The paper uses both 'clearance assumptions' and the allowance δ with F^δ = {F > -δ}. A 'clearance' assumption normally means a lower bound on distance to obstacles, not an allowance of penetration. Please define explicitly what assumption on γ and the local neighborhood is actually needed.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the completeness proof is a standard RRT-style argument and the empirical claims are external; the known proof gap is a correctness issue, not circularity.

full rationale

The paper's central derived claim, Theorem 4.3, is a probabilistic completeness bound obtained by a standard RRT-style ball-covering argument. The lower bound p = p_rotate * p_sample is not a restatement of any fitted input: p_rotate follows from a geometric convexity calculation (Lemma 4.1) and p_sample is asserted from box coverage (Lemma 4.2). There is no fitted parameter renamed as a prediction, and the empirical successes (100% success, AST 3.0 min on the 16-instance subset; 156 min on Elk) are direct measurements on external benchmarks from Tian et al. [19] and Zhang et al. [20], not outputs forced by construction. The paper does cite the authors' own prior work, notably [61] for the RRT-completeness framework and [28] for TR-RRT, but these citations are not load-bearing in a circular sense: the Chernoff-style tail bound is reproduced in the paper and is standard, and the contribution adapts it to rotations and subdivision boxes rather than merely invoking an unverified self-result. There is a genuine proof gap, not a circularity: Algorithm 2 retains a box only when |F_M2→M1(Center(b))| < D/2, while Lemma 4.2 asserts, without proof, that 'each such ball will contain at least one box b∈B' once D ≤ ν/5. For the paper's own mixed wide-tight setting, wide-clearance segments have no configurations with |F| < D/2 as D shrinks, so p_sample can be zero on those segments; this undermines Theorem 4.3's hypothesis but is a soundness/correctness issue, not equivalence-of-input-and-output. The limitation section also acknowledges grid-SDF approximation issues, which are implementation caveats rather than circular steps. Overall, no specific reduction from the claimed result back to its inputs is exhibited, so the correct circularity score is 0.

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

No new physical or mathematical entities are postulated; the 'critical manifold' is a named geometric object, not an additional degree of freedom. The central proof relies on a hidden coverage assumption (the second axiom) that is not stated as an assumption and is contradicted by the paper's mixed wide-tight framing.

free parameters (6)
  • δ (allowance/contact threshold) = 0.005
    Set in §5.1; defines F^δ and contact detection, and enters the proof via ν = min(δ, η). No sensitivity analysis is given.
  • η (RRT step size)
    Required by Algorithm 1 and used in the proof's ν = min(δ, η); no numerical value is reported.
  • N_rot (random rotation interval)
    Used in Algorithm 1 line 9; not specified. The proof appears to require a random rotation at every transition (Lemma 4.2), so this matters for completeness.
  • factor_refine = 10
    Algorithm 1 line 2; schedule parameter balancing refinement and search. Chosen by hand.
  • SDF grid resolution = 240^3
    §5.1; trilinear-interpolated grid for F_M1. Approximation quality affects both refinement pruning and collision detection.
  • Boundary sample count = 10,000
    §5.1; point set on robot boundary used for collision detection. Density is not justified and affects contact classification.
assumptions (6)
  • domain assumption There exists a solution path γ ⊂ F^δ with positive clearance such that small d_C-balls around its points are contained in F^δ.
    Implied by 'standard clearance assumptions' but never formalized; required for Lemma 4.2's convex-ball containment.
  • ad hoc to paper Every d_C-ball of radius ν/5 along γ contains a retained box b ∈ B after refinement.
    This is the load-bearing assertion in the proof of Lemma 4.2; it is not proven and is generally false for wide-clearance segments because Algorithm 2 prunes boxes with |F(center)| ≥ D/2.
  • domain assumption The SDF grid and 10,000-point boundary sampling correctly classify collision and contact for the models tested.
    Collision detection in §5.1 is approximate; false negatives/positives would affect both planning validity and empirical results.
  • domain assumption Random rotation of the entire system preserves feasibility and can be applied without changing the path's existence.
    Used in Lemma 4.1 and Algorithm 1's RandomRotate; not explicitly verified for the rpy box subdivision.
  • standard math The contact manifold has effective dimension at most 5, so subdivision complexity is O(1/δ^k), k≤5.
    Imported from the authors' own [59]; supports the refinement scheme's efficiency claim.
  • standard math The rpy Euler-angle parametrization with Haar volume element 1/(8π^2) cosθ dϕ dθ dψ correctly models uniform rotation sampling.
    Used in Lemma 4.1's probability computation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Manifold-Guided Motion Planning for Tight Assemblies." pith.science (2026). https://pith.science/paper/WFJFRSO7

@misc{pith2026260717898,
  author       = {Pith},
  title        = {Pith review of: Manifold-Guided Motion Planning for Tight Assemblies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WFJFRSO7}},
  note         = {Machine review of arXiv:2607.17898}
}
read the original abstract

Motion planning for rigid-body assembly poses a fundamental challenge in robotics due to tight geometric constraints. In such scenarios, feasible motions often require passing through (near-)zero clearance configurations in which the parts are tightly constrained by contact. In this work, we introduce Critical-Manifold Guided RRT (CMG-RRT), a sampling-based planner designed specifically for tight assembly problems. Our key observation is that in tight assemblies, valid solution paths lie on or near a critical manifold: the subset of configuration space consisting of poses with at least one contact point between parts. CMG-RRT guides exploration by adaptively biasing sampling toward neighborhoods of the critical manifold using a hierarchical subdivision of the configuration space. We prove that CMG-RRT is probabilistically complete under standard clearance assumptions. Empirical evaluation on challenging rotational assembly benchmarks demonstrates a 100% success rate across all tested instances, including, to the best of our knowledge, the first fully automatic solution of the Elk disentanglement puzzle. Our open source software is available through our project page: https://www.cgl.cs.tau.ac.il/projects/tight-assembly-planning.

Figures

Figures reproduced from arXiv: 2607.17898 by the authors.

Figure 1
Figure 1. Snapshots of the Elk disentanglement puzzle, previously unsolved (to the best of our knowledge), showing a complete solution generated by our algorithm. From left to right: the start configuration, traversal through four narrow C-space tunnels interleaved with wide free-space regions, and the goal configuration. 1 Introduction Sampling-based motion planning is a fundamental paradigm in robotics, used in appli￾cation… view at source ↗
Figure 2
Figure 2. Dataset provided by Tian et al. [19]. Excerpts from the more challenging [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustrations of projections of B C ν (·) balls onto [−π, π) × R. The center of the yellow ball is located in a subset of [−π, π) × R such that the ball is not convex. Therefore, there exist points q1, q2 in the ball such that the straight line segment (in blue) connecting them does not lie in the ball. With high probability (specified in Lemma 4.1), a random rotation will shift the center of the ball to a new posit… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of the Refine (subdivision) procedure in 3D. In CMG-RRT, this refinement takes place in 6D [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Illustration of Lemma 4.2. Note that the balls are not necessarily convex. [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 6
Figure 6. Figure 6: Execution of a CMG-RRT planned trajectory for the az puzzle using two UR5e robotic arms. Arms trajectories are based on a separate continuous IK derived from the exact CMG-RRT trajectory for the free-flying objects [PITH_FULL_IMAGE:figures/full_fig_p015_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 7 linked inside Pith

  1. [1]

    S. M. LaValle,Planning Algorithms. Cambridge University Press, 2006

    2006

  2. [2]

    Probabilistic Roadmaps for path planning in high-dimensional configuration spaces,

    L. E. Kavraki, P. ˇSvestka, J.-C. Latombe, and M. H. Overmars, “Probabilistic Roadmaps for path planning in high-dimensional configuration spaces,”IEEE Transactions on Robotics and Automation, vol. 12, no. 4, pp. 566–580, 1996

    1996

  3. [3]

    Rapidly-Exploring Random Trees: a new tool for path planning,

    S. LaValle, “Rapidly-Exploring Random Trees: a new tool for path planning,” Research Report 9811, 1998

    1998

  4. [4]

    Rapidly-exploring random trees: Progress and prospects: Steven m. lavalle, iowa state university, a james j. kuffner, jr., university of tokyo, tokyo, japan,

    S. M. LaValle and J. J. Kuffner, “Rapidly-exploring random trees: Progress and prospects: Steven m. lavalle, iowa state university, a james j. kuffner, jr., university of tokyo, tokyo, japan,”Algorithmic and computational robotics, pp. 303–307, 2001

    2001

  5. [5]

    Recent advances in Rapidly-Exploring Random Tree: A review,

    T. Xu, “Recent advances in Rapidly-Exploring Random Tree: A review,”He- liyon, vol. 10, no. 11, 2024. 15

    2024

  6. [6]

    Path planning in expansive configu- ration spaces,

    D. Hsu, J.-C. Latombe, and R. Motwani, “Path planning in expansive configu- ration spaces,”The International Journal of Robotics Research, vol. 18, no. 5, pp. 495–512, 1999

    1999

  7. [7]

    Analysis of probabilistic roadmaps for path planning,

    L. E. Kavraki, P. ˇSvestka, J.-C. Latombe, and M. H. Overmars, “Analysis of probabilistic roadmaps for path planning,”IEEE Transactions on Robotics and Automation, vol. 14, no. 1, pp. 166–171, 1998

    1998

  8. [8]

    On the power of manifold samples in exploring configuration spaces and the dimensionality of narrow passages,

    O. Salzman, M. Hemmer, and D. Halperin, “On the power of manifold samples in exploring configuration spaces and the dimensionality of narrow passages,” IEEE Trans Autom. Sci. Eng., vol. 12, no. 2, pp. 529–538, 2015

    2015

Show all 73 references
  1. [9]

    The bridge test for sampling narrow pas- sages with probabilistic roadmap planners,

    D. Hsu, T. Jiang, J. Reif, and Z. Sun, “The bridge test for sampling narrow pas- sages with probabilistic roadmap planners,” inIEEE International Conference on Robotics and Automation (ICRA), 2003, pp. 4420–4426

    2003

  2. [10]

    Narrow passage sampling for proba- bilistic roadmap planners,

    Z. Sun, D. Hsu, T. Jiang, and J. Reif, “Narrow passage sampling for proba- bilistic roadmap planners,” inIEEE International Conference on Robotics and Automation (ICRA), 2005, pp. 4423–4428

    2005

  3. [11]

    Path planning in complex 3d environments using a probabilistic roadmap method,

    F. Yan, Y.-S. Liu, and J.-Z. Xiao, “Path planning in complex 3d environments using a probabilistic roadmap method,”International Journal of Automation and computing, vol. 10, no. 6, pp. 525–533, 2013

    2013

  4. [12]

    A hybrid approach for complete motion planning,

    L. Zhang, Y. J. Kim, and D. Manocha, “A hybrid approach for complete motion planning,” in2007 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2007, pp. 7–14

    2007

  5. [13]

    Soft subdivision search in motion planning,

    C. K. Yap, “Soft subdivision search in motion planning,” inProceedings, 1st Workshop on Robotics Challenge and Vision (RCV 2013), 2013

    2013

  6. [14]

    Soft subdivision search in motion planning, ii: axiomatics,

    ——, “Soft subdivision search in motion planning, ii: axiomatics,” inInterna- tional Workshop on Frontiers in Algorithmics. Springer, 2015, pp. 7–22

    2015

  7. [15]

    Theory and explicit design of a path planner for an se (3) robot,

    Z. Zhang, Y.-J. Chiang, and C. Yap, “Theory and explicit design of a path planner for an se (3) robot,”arXiv preprint arXiv:2407.05135, 2024

    2024 arXiv

  8. [16]

    Compliant manipulation planning,

    Z. Friedman and R. Paul, “Compliant manipulation planning,”IEEE Transac- tions on Robotics and Automation, vol. 12, no. 4, pp. 554–564, 1996

    1996

  9. [17]

    Active compliant motion: A survey,

    T. Lefebvre, H. Bruyninckx, and J. D. Schutter, “Active compliant motion: A survey,”Advanced Robotics, vol. 19, no. 5, pp. 479–499, 2005

    2005

  10. [18]

    Compliant robot motion i: A formalism for specifying compliant motion tasks,

    J. D. Schutter and J. V. Brussel, “Compliant robot motion i: A formalism for specifying compliant motion tasks,”The International Journal of Robotics Research, vol. 7, no. 4, pp. 3–17, 1988

    1988

  11. [19]

    Assemble Them All: Physics-Based Planning for Generalizable Assembly By Disassembly,

    Y. Tian, J. Xu, Y. Li, J. Luo, S. Sueda, H. Li, K. D. Willis, and W. Matusik, “Assemble Them All: Physics-Based Planning for Generalizable Assembly By Disassembly,”ACM Transactions on Graphics (TOG), vol. 41, no. 6, pp. 1–11, 2022

    2022

  12. [20]

    C-space tunnel discovery for puzzle path planning,

    X. Zhang, R. Belfer, P. G. Kry, and E. Vouga, “C-space tunnel discovery for puzzle path planning,”ACM Trans. Graph., vol. 39, no. 4, Aug 2020

    2020

  13. [21]

    Adaptive experience sampling for motion planning using the generator–critic framework,

    Y. Lee, C. Chamzas, and L. E. Kavraki, “Adaptive experience sampling for motion planning using the generator–critic framework,” inIEEE International Conference on Robotics and Automation (ICRA), 2022, pp. 7104–7110. 16

    2022

  14. [22]

    Sample-driven connectivity learning for motion plan- ning in narrow passages,

    S. Li and N. T. Dantam, “Sample-driven connectivity learning for motion plan- ning in narrow passages,” inProceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2023, pp. 5681–5687

    2023

  15. [23]

    Manipulation planning on constraint manifolds,

    D. Berenson, S. Srinivasa, D. Ferguson, and J. Kuffner, “Manipulation planning on constraint manifolds,” inIEEE International Conference on Robotics and Automation (ICRA), 2009, pp. 625–632

    2009

  16. [24]

    Path planning under kinematic constraints by Rapidly- Exploring Random Trees,

    L. Jaillet and J. Porta, “Path planning under kinematic constraints by Rapidly- Exploring Random Trees,”IEEE Transactions on Robotics, vol. 29, no. 1, pp. 105–117, 2013

    2013

  17. [25]

    A general method for sampling on im- plicit manifolds,

    C. Voss, M. Moll, and L. E. Kavraki, “A general method for sampling on im- plicit manifolds,” inIEEE International Conference on Robotics and Automation (ICRA), 2017, pp. 2420–2426

    2017

  18. [26]

    Exploring implicit configuration spaces,

    Z. Kingston and M. Moll, “Exploring implicit configuration spaces,”The Inter- national Journal of Robotics Research, vol. 38, no. 8, pp. 998–1012, 2019

    2019

  19. [27]

    Sampling-based methods for mo- tion planning with constraints,

    Z. Kingston, M. Moll, and L. E. Kavraki, “Sampling-based methods for mo- tion planning with constraints,”Annual Review of Control, Robotics, and Au- tonomous Systems, vol. 1, pp. 159–185, 2018

    2018

  20. [28]

    Tight Motion Planning by Rie- mannian Optimization for Sliding and Rolling with Finite Number of Contact Points,

    D. Livnat, M. M. Bilevich, and D. Halperin, “Tight Motion Planning by Rie- mannian Optimization for Sliding and Rolling with Finite Number of Contact Points,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 14 333–14 340

    2024

  21. [29]

    Reporting neighbors in high-dimensional euclidean space,

    D. Aiger, H. Kaplan, and M. Sharir, “Reporting neighbors in high-dimensional euclidean space,”SIAM Journal on Computing, vol. 43, no. 4, pp. 1363–1395, 2014

    2014

  22. [30]

    Efficient high-quality motion plan- ning by fast all-pairs r-nearest-neighbors,

    M. Kleinbort, O. Salzman, and D. Halperin, “Efficient high-quality motion plan- ning by fast all-pairs r-nearest-neighbors,” in2015 IEEE International Confer- ence on Robotics and Automation (ICRA). IEEE, 2015, pp. 2985–2990

    2015

  23. [31]

    Perspectives on Euler angle singulari- ties, gimbal lock, and the orthogonality of applied forces and applied moments,

    E. G. Hemingway and O. M. O’Reilly, “Perspectives on Euler angle singulari- ties, gimbal lock, and the orthogonality of applied forces and applied moments,” Multibody system dynamics, vol. 44, no. 1, pp. 31–56, 2018

    2018

  24. [32]

    Special orthogonal group SO(3), Euler angles, angle-axis, Ro- driguez vector and unit-quaternion: Overview, mapping and challenges,

    H. A. Hashim, “Special orthogonal group SO(3), Euler angles, angle-axis, Ro- driguez vector and unit-quaternion: Overview, mapping and challenges,”arXiv preprint arXiv:1909.06669, 2019

    1909 arXiv

  25. [33]

    J. M. Selig,Geometrical Methods in Robotics. Springer Science & Business Media, 2013

    2013

  26. [34]

    Representing attitude: Euler angles, unit quaternions, and rotation vectors,

    J. Diebelet al., “Representing attitude: Euler angles, unit quaternions, and rotation vectors,”Matrix, vol. 58, no. 15-16, pp. 1–35, 2006

    2006

  27. [35]

    The Haar Measure in Solid Mechanics,

    C. Ecker and B. Kolev, “The Haar Measure in Solid Mechanics,”arXiv preprint arXiv:2410.03371, 2024

    2024 arXiv

  28. [36]

    Finite metric spaces–combinatorics, geometry and algorithms,

    N. Linial, “Finite metric spaces–combinatorics, geometry and algorithms,”arXiv preprint math/0304466, 2003

    2003 arXiv

  29. [37]

    M. O. Searcoid,Metric Spaces. Springer, 2007. 17

    2007

  30. [38]

    N. R. Howes,Modern Analysis and Topology. Springer Science & Business Media, 1995

    1995

  31. [39]

    3d distance fields: A survey of techniques and applications,

    M. W. Jones, J. A. Baerentzen, and M. Sramek, “3d distance fields: A survey of techniques and applications,”IEEE Transactions on visualization and Computer Graphics, vol. 12, no. 4, pp. 581–599, 2006

    2006

  32. [40]

    A survey and benchmark of automatic surface reconstruction from point clouds,

    R. Sulzer, R. Marlet, B. Vallet, and L. Landrieu, “A survey and benchmark of automatic surface reconstruction from point clouds,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  33. [41]

    Computing local signed distance fields for large polygonal models,

    B. Chang, D. Cha, and I. Ihm, “Computing local signed distance fields for large polygonal models,” inComputer Graphics Forum, vol. 27. Wiley Online Library, 2008, pp. 799–806

    2008

  34. [42]

    Generating signed distance fields on the GPU with ray maps,

    B. Krayer and S. M¨ uller, “Generating signed distance fields on the GPU with ray maps,”The Visual Computer, vol. 35, no. 6, pp. 961–971, 2019

    2019

  35. [43]

    PSDF fusion: Probabilistic signed distance function for on-the-fly 3D data fusion and scene reconstruction,

    W. Dong, Q. Wang, X. Wang, and H. Zha, “PSDF fusion: Probabilistic signed distance function for on-the-fly 3D data fusion and scene reconstruction,” in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 701–717

    2018

  36. [44]

    Signed distance transform using graphics hardware,

    C. Sigg, R. Peikert, and M. Gross, “Signed distance transform using graphics hardware,” inIEEE Visualization, 2003. VIS 2003.IEEE, 2003, pp. 83–90

    2003

  37. [45]

    A fast sweeping method for Eikonal equations,

    H. Zhao, “A fast sweeping method for Eikonal equations,”Mathematics of com- putation, vol. 74, no. 250, pp. 603–627, 2005

    2005

  38. [46]

    On the effectiveness of eeight- encoded neural implicit 3D shapes,

    T. Davies, D. Nowrouzezahrai, and A. Jacobson, “On the effectiveness of eeight- encoded neural implicit 3D shapes,”arXiv preprint arXiv:2009.09808, 2020

    2009 arXiv

  39. [47]

    Learning models as func- tionals of signed-distance fields for manipulation planning,

    D. Driess, J.-S. Ha, M. Toussaint, and R. Tedrake, “Learning models as func- tionals of signed-distance fields for manipulation planning,” inConference on robot learning. PMLR, 2022, pp. 245–255

    2022

  40. [48]

    Neural NMPC through signed distance field encoding for collision avoidance,

    M. Jacquet, M. Harms, and K. Alexis, “Neural NMPC through signed distance field encoding for collision avoidance,”The International Journal of Robotics Research, p. 02783649251401223, 2025

    2025

  41. [49]

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

    Y. Li, Y. Zhang, A. Razmjoo, and S. Calinon, “Representing robot geometry as distance fields: Applications to whole-body manipulation,” in2024 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2024, pp. 15 351–15 357

    2024

  42. [50]

    Deepsdf: Learning continuous signed distance functions for shape representation,

    J. J. Park, P. Florence, J. Straub, R. Newcombe, and S. Lovegrove, “Deepsdf: Learning continuous signed distance functions for shape representation,” inPro- ceedings of the IEEE/CVF conference on computer vision and pattern recogni- tion, 2019, pp. 165–174

    2019

  43. [51]

    Diffusion-based signed distance fields for 3D shape generation,

    J. Shim, C. Kang, and K. Joo, “Diffusion-based signed distance fields for 3D shape generation,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2023, pp. 20 887–20 897

    2023

  44. [52]

    ContactSDF: Signed Distance Functions as multi-contact models for dexterous manipulation,

    W. Yang and W. Jin, “ContactSDF: Signed Distance Functions as multi-contact models for dexterous manipulation,”IEEE Robotics and Automation Letters, 2025. 18

    2025

  45. [53]

    3D collision detection: A survey,

    P. Jim´ enez, F. Thomas, and C. Torras, “3D collision detection: A survey,”Com- puters & Graphics, vol. 25, no. 2, pp. 269–285, 2001

    2001

  46. [54]

    Continuous collision detec- tion between points and signed distance fields,

    J. Bender, C. Duriez, F. Jaillet, and G. Zachmann, “Continuous collision detec- tion between points and signed distance fields,” inWorkshop on virtual reality interaction and physical simulation, vol. 8, 2014

    2014

  47. [55]

    Neural implicit surfaces for effi- cient and accurate collisions in physically based simulations,

    H. Bertiche, M. Madadi, and S. Escalera, “Neural implicit surfaces for effi- cient and accurate collisions in physically based simulations,”arXiv preprint arXiv:2110.01614, 2021

    2021 arXiv

  48. [56]

    Hierarchical HP-Adaptive signed distance fields,

    D. Koschier, C. Deul, and J. Bender, “Hierarchical HP-Adaptive signed distance fields,” inSymposium on Computer Animation, 2016, pp. 189–198

    2016

  49. [57]

    Local optimization for robust signed distance field collision,

    M. Macklin, K. Erleben, M. M¨ uller, N. Chentanez, S. Jeschke, and Z. Corse, “Local optimization for robust signed distance field collision,”Proceedings of the ACM on Computer Graphics and Interactive Techniques, vol. 3, no. 1, pp. 1–17, 2020

    2020

  50. [58]

    6-DOF Haptic rendering using continuous collision detec- tion between points and signed distance fields,

    H. Xu and J. Barbiˇ c, “6-DOF Haptic rendering using continuous collision detec- tion between points and signed distance fields,”IEEE transactions on haptics, vol. 10, no. 2, pp. 151–161, 2016

    2016

  51. [59]

    A note on the time complexity of using subdivi- sion methods for the approximation of fibers,

    M. M. Bilevich and D. Halperin, “A note on the time complexity of using subdivi- sion methods for the approximation of fibers,”arXiv preprint arXiv:2503.01626, 2025

    2025 arXiv

  52. [60]

    Adaptively sam- pled distance fields: A general representation of shape for computer graphics,

    S. F. Frisken, R. N. Perry, A. P. Rockwood, and T. R. Jones, “Adaptively sam- pled distance fields: A general representation of shape for computer graphics,” inProceedings of SIGGRAPH, 2000, pp. 249–254

    2000

  53. [61]

    Prob- abilistic completeness of RRT for geometric and kinodynamic planning with forward propagation,

    M. Kleinbort, K. Solovey, Z. Littlefield, K. E. Bekris, and D. Halperin, “Prob- abilistic completeness of RRT for geometric and kinodynamic planning with forward propagation,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. i–vii, 2018

    2018

  54. [62]

    Revisiting the asymptotic optimality of RRT,

    K. Solovey, L. Janson, E. Schmerling, E. Frazzoli, and M. Pavone, “Revisiting the asymptotic optimality of RRT,” in2020 IEEE international conference on robotics and automation (ICRA). IEEE, 2020, pp. 2189–2195

    2020

  55. [63]

    Openmp: an industry standard api for shared- memory programming,

    L. Dagum and R. Menon, “Openmp: an industry standard api for shared- memory programming,”IEEE computational science and engineering, vol. 5, no. 1, pp. 46–55, 1998

    1998

  56. [64]

    2D and 3D fast intersection and distance computation,

    P. Alliez, S. Tayeb, and C. Wormser, “2D and 3D fast intersection and distance computation,” inCGAL User and Reference Manual, 6.1.1 ed. CGAL Editorial Board, 2026. [Online]. Available: https://doc.cgal.org/6.1.1/Manual/packages. html#PkgAABBTree

    2026

  57. [65]

    The K-Means algorithm: A compre- hensive survey and performance evaluation,

    M. Ahmed, R. Seraj, and S. M. S. Islam, “The K-Means algorithm: A compre- hensive survey and performance evaluation,”Electronics, vol. 9, no. 8, p. 1295, 2020

    2020

  58. [66]

    Some methods for classification and analysis of multivariate ob- servations,

    J. MacQueen, “Some methods for classification and analysis of multivariate ob- servations,” inProceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, L. M. Le Cam and J. Neyman, Eds., no. 14. Oak- land, CA, USA, 1967, pp. 281–297. 19

    1967

  59. [67]

    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

  60. [68]

    Batch informed trees (bit*): Informed asymptotically optimal anytime search,

    J. D. Gammell, T. D. Barfoot, and S. S. Srinivasa, “Batch informed trees (bit*): Informed asymptotically optimal anytime search,”The International Journal of Robotics Research, vol. 39, no. 5, pp. 543–567, 2020

    2020

  61. [69]

    Kinodynamic motion planning by interior- exterior cell exploration,

    I. A. S ¸ucan and L. E. Kavraki, “Kinodynamic motion planning by interior- exterior cell exploration,” inAlgorithmic foundation of robotics VIII: selected contributions of the eight international workshop on the algorithmic foundations of robotics. Springer, 2009, pp. 449–464

    2009

  62. [70]

    Rrt-connect: An efficient approach to single- query path planning,

    J. J. Kuffner and S. M. LaValle, “Rrt-connect: An efficient approach to single- query path planning,” inProceedings 2000 ICRA. Millennium conference. IEEE international conference on robotics and automation. Symposia proceedings (Cat. No. 00CH37065), vol. 2. IEEE, 2000, pp. 995–1001

    2000

  63. [71]

    The Open Motion Planning Library,

    I. A. S ¸ucan, M. Moll, and L. E. Kavraki, “The Open Motion Planning Library,” IEEE Robotics & Automation Magazine, vol. 19, no. 4, pp. 72–82, December 2012, https://ompl.kavrakilab.org

    2012

  64. [72]

    Efficient physics-based planning: Sampling search via non-deterministic tactics and skills,

    S. Zickler and M. M. Veloso, “Efficient physics-based planning: Sampling search via non-deterministic tactics and skills,” inAAMAS (1), 2009, pp. 27–33

    2009

  65. [73]

    A full-cycle assembly operation: From digital planning to trajectory execution using a robotic arm,

    D. Livnat, Y. Lavi, and D. Halperin, “A full-cycle assembly operation: From digital planning to trajectory execution using a robotic arm,” in2025 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2025, pp. 9184–9191. 20

    2025

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.