Pith. sign in

REVIEW 2 major objections 8 minor 117 references

Jointly searching the assignment of trajectories to arms and the Jacobian null spaces of redundant manipulators yields a resolution-complete planner for embodying multi-hand manipulation policies on multi-arm robots.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 06:01 UTC pith:S672BSZS

load-bearing objection Novel and useful problem formulation plus a sensible CBS-nullspace planner, but the resolution-completeness claim is graph-relative, not continuous, and the experiments need statistical support. the 2 major comments →

arxiv 2607.22020 v1 pith:S672BSZS submitted 2026-07-24 cs.RO

Embodying Multi-Hand Manipulation Policies by Searching the Assignment and Null Spaces

Yorai Shaoul , Jiaoyang Li , Maxim Likhachev This is my paper
classification cs.RO
keywords multi-hand manipulationmulti-arm robotJacobian null spaceconflict-based searchpose-constraint manifoldinverse kinematicsresolution completenessmotion planning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tackles the problem of executing learned multi-hand manipulation policies on physical multi-arm robots. Such policies output end-effector trajectories for abstract 'hands,' but they do not say which physical arm should execute which trajectory or how each arm should bend to avoid colliding with the others. The paper claims that both questions can be answered together by a single search: an on-manifold A* that explores the redundancy of each arm, wrapped in a conflict-based search that resolves both assignment conflicts and geometric collisions. The result is a planner that will find a feasible embodiment whenever one exists in the finite graph induced by its motion primitives, together with experiments showing it outperforms IK-tracking, pose-constrained roadmaps, and joint composite search.

Core claim

The paper's central claim is that the execution gap for multi-hand policies can be closed by treating embodiment as a joint search over two spaces previously handled separately: the discrete space of trajectory-to-arm assignments and the continuous Jacobian null spaces of redundant manipulators. Its algorithm, Ω-CBSA, builds on Conflict-Based Search: each robot runs a multi-goal on-manifold A* that tracks a chosen end-effector trajectory while generating successor configurations by IK projection followed by null-space steps, and a high-level constraint tree resolves assignment conflicts and inter-arm collisions with mutually disjunctive constraints. The theoretical result is Theorem 1: Ω-CBS

What carries the argument

The central object is the pose-constraint manifold M_t = {q | FK(q) = x_t}, the set of arm configurations realizing a given end-effector pose. The paper explores this manifold with null-space steps: from a nominal IK projection q', each basis direction n of the Jacobian null space is stepped by ±ε and corrected back onto the manifold by IK, generating new configurations that preserve the end-effector pose while changing the arm's shape. These local moves become the motion primitives of an A* search, and the whole scheme is wrapped in a CBS constraint tree whose two constraint types—assignment constraints (robot i may not select trajectory j) and geometric point constraints (robot i may not o

Load-bearing premise

That the finite graph produced by the specific null-space step size (ε=0.2), the IK projection, and the IK seed set actually contains a configuration trajectory that tracks the desired end-effector path; the theorem guarantees success only within that graph, and the paper gives no condition linking that graph to the continuous problem, so a feasible continuous solution may be missed if it is not representable at the chosen resolution.

What would settle it

Run Ω-CBSA with ε=0.2 on a two-arm problem where the only collision-free tracking of two crossing trajectories requires both arms to pass through a redundancy branch that is reachable only by a null-space motion larger than 0.2 at some step; if the planner returns 'no solution' while a continuous configuration trajectory exists, the practical completeness claim is falsified.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • A practitioner can take multi-hand policy outputs and run them on a multi-arm robot with a feasibility guarantee, rather than ad hoc single-arm IK pipelines, provided the solution is representable in the planner's primitive graph.
  • The unified assignment search means no separate task-allocation or greedy assignment step is needed; assignments are decided during motion search.
  • Redundancy is used constructively: null-space exploration allows an arm to change its body configuration to avoid inter-arm collisions without deviating from its prescribed end-effector path.
  • The prioritized variants provide faster, often successful planning at the cost of the same completeness guarantee.
  • Ω-CBSA's resolution-completeness extends to any number of arms and trajectories within the OM-AMRAMP formulation, as the proof is dimension-independent.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The completeness guarantee is conditioned on the resolution of the null-space step ε=0.2 and the IK seed set; if the ideal continuous solution passes between the discrete null-space steps, the planner could fail even though a continuous solution exists. A natural testable extension is to check completeness across decreasing ε to see if solutions emerge as resolution increases.
  • The method could be adapted to non-redundant arms (where null space is trivial) or to arms with kinematics that make the null-space projection ill-conditioned, in which case the framework degrades gracefully to IK-tracking with a potentially reduced success rate.
  • Because assignments are searched implicitly, the approach may also serve as a task-and-motion-planning primitive for problems where the number of trajectories exceeds the number of arms, allowing selection of a subset.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 8 minor

Summary. The paper addresses the problem of executing multi-hand, policy-generated end-effector trajectories on multi-arm robots. It formalizes OM-AMRAMP, in which one must find a bijective assignment of trajectories to robots and collision-free configuration trajectories that track the assigned end-effector poses. The proposed method, Omega-CBSA, combines a single-robot on-manifold A* search (Omega-A*) with Conflict-Based Search. Omega-A* generates successors by an IK projection followed by one null-space step of size epsilon, thereby exploring redundant configurations. Omega-CBSA adds assignment constraints and geometric point constraints within a CBS constraint tree. The paper claims resolution-completeness for OM-AMRAMP (Lemma 1, Lemma 2, Theorem 1) and reports simulated benchmark results and a qualitative physical demonstration with three arms.

Significance. If the claimed completeness result held for the continuous problem, this would be a valuable advance over ad hoc IK-tracking pipelines for multi-arm policy grounding. The algorithmic idea of jointly searching over trajectory-to-arm assignments and Jacobian null-space motions is attractive and the assignment-constraint disjunctiveness argument is clean and self-contained. The benchmark results, if reproducible with statistics, would also support practical usefulness. However, the central theoretical claim as stated is not supported: Theorem 1 is a completeness statement over a finite graph induced by the motion primitives, not over the continuous OM-AMRAMP problem, and the graph does not become dense as the resolution parameter tends to zero. The paper is therefore best viewed as proposing a useful search framework with a graph-level completeness property, rather than a resolution-complete continuous planner. With an honest restatement of the theoretical claim and strengthened experimental reporting, the paper could still be a meaningful contribution.

major comments (2)
  1. [Theoretical Analysis, Lemma 1 and Theorem 1] The completeness claim is relative to the finite graph G(epsilon, seeds, IK routine), not to the continuous OM-AMRAMP problem defined in the Background. Lemma 1's proof says that if a feasible trajectory exists 'in this induced graph' then Omega-A* will find it; this is essentially a restatement of the definition of reachability, not a resolution-completeness theorem. No density or clearance condition links epsilon to the existence of a continuous feasible solution. Moreover, the successor rule allows at most one null-space step of size ±epsilon per waypoint, so as epsilon→0 the graph collapses to the nominal IK-tracking trajectory rather than becoming dense in the constraint manifold M_t. A feasible continuous solution requiring a null-space deviation delta not in {0, ±epsilon} at some waypoint is unrepresentable, so the planner may fail while Theorem 1 remains true. The Abstract's clai
  2. [Experimental Analysis, Fig. 2 and Experimental Results] The main empirical claims—'consistently outperformed all baselines' and 'remains fast'—are based on a single aggregate over 450 problems with no error bars, confidence intervals, or statistical tests. Fig. 2 reports overall success rates and mean runtime (640ms, sigma=990ms) but not the number of trials per condition, variance across problem instances, or instance-level comparisons. The physical 3-arm demonstration is described only qualitatively (cloth rotation, box flip, planar pushing) with no quantitative metrics such as task success rate, execution time, or number of trials, despite the Introduction claiming a 'physical and simulated experimental study.' Please provide trial-level data, statistical comparisons, a sensitivity analysis of the fixed parameters (epsilon=0.2, w_h=10, w_c=0.1), and quantitative hardware results.
minor comments (8)
  1. [On-Manifold A*] The successor-generation description is ambiguous about whether the nominal projection q'=IK(x_{t+1},q) is itself added to OPEN as a successor, or only the null-space-explored q''=q'+epsilon n after IK correction. Please clarify.
  2. [On-Manifold A*] The initial-state cost is defined as g(s)=||FK(q)-x_1||, while subsequent edge costs are ||q''-q|| in joint space. These are different units and norms; please define a single, consistent cost-to-go for the search.
  3. [Background] The phrase 'on-manifold' is used throughout, but all states are produced by numerical IK and thus only lie approximately on the manifold FK(q)=x_t. Define the IK tolerance and state how it interacts with the theoretical guarantees.
  4. [Background] There is a typo in the Background section: 'thaF K' should be 'that FK'.
  5. [Lemma 2] The proof of geometric point-constraint disjunctiveness is imported from Shaoul et al. 2024b. Since the argument is short (if both robots occupy witness point p, they collide), including a self-contained one-paragraph proof would make Lemma 2 independent of the external reference.
  6. [Experimental Results] The symbol sigma is overloaded: it denotes the assignment bijection in the problem formulation and the standard deviation of runtime in the experiments. Please disambiguate.
  7. [Experimental Results] Figure 2's rightmost plot lacks clear axis labels and units for runtime; please add them.
  8. [Conclusion] The Conclusion claims 'real-time performance,' but the reported mean runtime is roughly 640ms with a maximum of 4.9s. Please specify the intended control period or otherwise qualify 'real-time.'

Circularity Check

2 steps flagged

Completeness guarantee is partly a graph-reachability tautology and partly rests on a load-bearing self-citation for geometric point-constraint disjunctiveness; assignment reasoning and experiments remain independent.

specific steps
  1. self definitional [Theoretical Analysis, Lemma 1 and proof]
    "If a feasible pose-constrained trajectory τ i following a fixed end-effector path X j ∈ {X1, . . . XN } exists for a R i under motion primitive discretization resolution, then Ω-A⋆ will find it. Proof. Ω-A⋆ searches a finite graph ... Since Ω-A⋆ systematically explores this finite graph, every reachable node will eventually be expanded. Therefore, if a feasible trajectory exists in this induced graph and starts from one of the initialized states, Ω-A⋆ will eventually reach and return it."

    Lemma 1's antecedent ('feasible ... under motion primitive discretization resolution') is identified in the proof with existence in the finite induced graph built from IK seeds and null-space primitives, and the proof only restates that exhaustive expansion reaches reachable nodes. This is a reachability tautology. Yet the paper concludes 'will succeed whenever a valid embodiment exists' as if this graph-restricted antecedent were continuous OM-AMRAMP feasibility. No density or clearance argument links the fixed ϵ=0.2 primitive set to the continuous constraint manifolds, so the headline completeness guarantee is the search's own reachability definition restated rather than a derived property of the original problem.

  2. self citation load bearing [Theoretical Analysis, Lemma 2 proof]
    "Geometric point constraints are mutually disjunctive, as shown in Shaoul et al. (2024b)."

    Theorem 1's proof relies on Lemma 2 for both assignment and geometric constraints. The assignment half is proved in-text, but the geometric half is not; it is deferred solely to Shaoul et al. (2024b), a prior paper by overlapping authors (Shaoul, Likhachev, Li). That citation is load-bearing: without the imported disjunctiveness of geometric point constraints, the CBS-completeness argument for collision constraints does not go through. The cited result is not machine-checked, code-reproduced, or independently argued in this manuscript, so the central multi-robot completeness claim rests on a self-citation rather than a self-contained derivation.

full rationale

No parameters are fitted, and the assignment-constraint portion of Lemma 2 is self-contained, so this is not a fitted-input-called-prediction case. The central algorithm also has independent empirical content in the experiments. However, Lemma 1's proof is a graph-reachability tautology relative to the motion-primitive discretization, and the paper's abstract and Theorem 1's closing sentence state the guarantee as if it applied to any valid continuous embodiment without a density/clearance link. In addition, the geometric point-constraint disjunctiveness needed for Theorem 1 is imported from the authors' own Shaoul et al. (2024b) paper without independent verification. These are partial circularity and load-bearing self-citation concerns, but the assignment-plus-CBS construction is not reduced to a fit or to a pure self-citation chain, so the score is moderate.

Axiom & Free-Parameter Ledger

2 free parameters · 5 axioms · 0 invented entities

The central completeness theorem rests on several domain assumptions: a finite-graph discretization of the continuous problem, local coverage of the constraint manifold by Jacobian null space plus IK correction, and imported mutual disjunctiveness of geometric constraints. The planner itself has two hand-set search parameters. No invented physical entities.

free parameters (2)
  • null-space step size ϵ = 0.2
    Hand-chosen primitive resolution; determines the finite graph over which resolution-completeness holds and affects whether continuous solutions are represented.
  • heuristic weights w_h, w_c = w_h=10, w_c=0.1
    Hand-tuned weights in the A* priority f=g+w_h*h+w_c*c; affect practical speed and search behavior but not completeness.
axioms (5)
  • domain assumption Solutions are represented on a finite-resolution discrete-time graph induced by a finite set of motion primitives; true continuous completeness is not claimed.
    Invoked in the Theoretical Analysis to define resolution-completeness; limits the guarantee to trajectories representable in the graph.
  • standard math The Jacobian null space plus IK projection locally spans the constraint manifold at non-singular configurations, so null-space moves explore redundant configurations.
    Used in Background and in Ω-A* successor generation; singularities or large steps could break exact manifold coverage.
  • domain assumption Geometric point constraints are mutually disjunctive as proven in Shaoul et al. 2024b; Theorem 1 relies on this external result.
    The current paper does not re-derive the geometric constraint property; it is load-bearing for CBS completeness.
  • domain assumption The robot can be moved from its current configuration to a sampled first-pose configuration τ_i^1 by an external planner without affecting the solution.
    The problem explicitly leaves reaching τ_1 to standard multi-arm planners; the algorithm starts from IK solutions for x_1, not from the robot's true start.
  • domain assumption Numerical IK converges to exact pose tracking, so every generated successor satisfies FK(q)=x_t within working tolerance.
    Successor generation relies on IK(x_{t+1}, q) returning a configuration on the manifold; convergence and tolerance are not analyzed.

pith-pipeline@v1.3.0-alltime-deepseek · 7556 in / 11334 out tokens · 118510 ms · 2026-08-01T06:01:30.908000+00:00 · methodology

0 comments
read the original abstract

Learned manipulation policies increasingly predict motions for abstract "hands" and are attractive in practice because they rely on easily collected demonstrations and transfer across robot platforms. Executing these trajectories on multi-arm robots, however, is not trivial. Multi-hand policy outputs must be assigned to physical arms, each arm must realize a configuration-space motion that tracks its prescribed end-effector trajectory, and all arms must respect kinematic limits and avoid collisions. In the absence of algorithms that directly address this problem, practitioners typically extend single-arm inverse-kinematics (IK) pipelines in an ad hoc way, with no guarantees of feasibility or safety. In this work, we close this execution gap with a search-based framework that is theoretically complete for grounding policy-generated multi-hand trajectories onto physical multi-arm systems. Building on Conflict-Based Search, our method explicitly searches over both the discrete assignment of trajectories to arms and the continuous Jacobian null spaces of redundant manipulators, using redundancy to avoid inter-arm collisions while tracking the prescribed motions. This unified treatment of assignment and null-space motion yields a practically efficient planner that safely realizes coordinated manipulation-policy outputs on multi-arm robots. See omcbsa.github.io for more.

Figures

Figures reproduced from arXiv: 2607.22020 by Jiaoyang Li, Maxim Likhachev, Yorai Shaoul.

Figure 1
Figure 1. Figure 1: Policy outputs often specify only end-effector tra [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Left images: illustrations of our test cases visualizing six robots and six end-effector trajectories. The left image shows all robots moving from right to left in a constrained space, and the right image shows six robots performing a box￾flipping motion. From middle to the right: overall success rates, success rates by robot count, and average overall runtime. Baselines To the best of our knowledge, no al… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

117 extracted references · 5 canonical work pages

  1. [1]

    2025 , url =

    Collaborative Multi-Robot Non-Prehensile Manipulation via Flow-Matching Co-Generation , author =. 2025 , url =

    2025

  2. [2]

    2017 22nd IEEE International conference on emerging technologies and factory automation (ETFA) , pages=

    Cartesian path planning for arc welding robots: Evaluation of the descartes algorithm , author=. 2017 22nd IEEE International conference on emerging technologies and factory automation (ETFA) , pages=. 2017 , organization=

    2017

  3. [3]

    Applied Sciences , VOLUME =

    Gan, Zhi Lon and Musa, Siti Nurmaya and Yap, Hwa Jen , TITLE =. Applied Sciences , VOLUME =. 2023 , NUMBER =

    2023

  4. [4]

    Proceedings of the International Conference on Automated Planning and Scheduling , author=

    Provable Indefinite-Horizon Real-Time Planning for Repetitive Tasks , volume=. Proceedings of the International Conference on Automated Planning and Scheduling , author=. 2021 , month=. doi:10.1609/icaps.v29i1.3540 , abstractNote=

    work page doi:10.1609/icaps.v29i1.3540 2021

  5. [5]

    The International Journal of Robotics Research , volume=

    Provably constant-time planning and replanning for real-time grasping objects off a conveyor belt , author=. The International Journal of Robotics Research , volume=. 2021 , publisher=

    2021

  6. [6]

    2021 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    Alternative paths planner (app) for provably fixed-time manipulation planning in semi-structured environments , author=. 2021 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2021 , organization=

    2021

  7. [7]

    IEEE Rob

    The Open Motion Planning Library , Author =. IEEE Rob. & Aut. Mag. , Year =

  8. [8]

    and Svestka, P

    Kavraki, L.E. and Svestka, P. and Latombe, J.-C. and Overmars, M.H. , journal=. Probabilistic roadmaps for path planning in high-dimensional configuration spaces , year=

  9. [9]

    and Kolountzakis, M.N

    Kavraki, L.E. and Kolountzakis, M.N. and Latombe, J.-C. , journal=. Analysis of probabilistic roadmaps for path planning , year=

  10. [10]

    2012 IEEE International Conference on Robotics and Automation , pages=

    A robot path planning framework that learns from experience , author=. 2012 IEEE International Conference on Robotics and Automation , pages=. 2012 , organization=

    2012

  11. [11]

    Autonomous Robots , year=

    Fast motion planning from experience: trajectory prediction for speeding up movement generation , author=. Autonomous Robots , year=

  12. [12]

    Proceedings of the International Conference on Automated Planning and Scheduling , author=

    Subgoal Graphs for Optimal Pathfinding in Eight-Neighbor Grids , volume=. Proceedings of the International Conference on Automated Planning and Scheduling , author=. 2013 , month=. doi:10.1609/icaps.v23i1.13568 , abstractNote=

    work page doi:10.1609/icaps.v23i1.13568 2013

  13. [13]

    2006 , publisher=

    Planning algorithms , author=. 2006 , publisher=

    2006

  14. [14]

    arXiv preprint arXiv:2205.04422 , year=

    Motion planning around obstacles with convex optimization , author=. arXiv preprint arXiv:2205.04422 , year=

    Pith/arXiv arXiv

  15. [15]

    Rapidly-exploring random trees : a new tool for path planning

    LaValle, S. Rapidly-exploring random trees : a new tool for path planning. Research Report 9811. 1998

    1998

  16. [16]

    The international journal of robotics research , volume=

    Sampling-based algorithms for optimal motion planning , author=. The international journal of robotics research , volume=. 2011 , publisher=

    2011

  17. [17]

    Kuffner, J.J. and. Proceedings 2000. doi:10.1109/ROBOT.2000.844730 , shorttitle =

    arXiv 2000

  18. [18]

    Gammell and Siddhartha S

    Jonathan D. Gammell and Siddhartha S. Srinivasa and Timothy D. Barfoot , title =. 2014. doi:10.1109/iros.2014.6942976 , url =

    arXiv 2014

  19. [19]

    The Repetition Roadmap for Repetitive Constrained Motion Planning , year=

    Lehner, Peter and Albu-Schäffer, Alin , journal=. The Repetition Roadmap for Repetitive Constrained Motion Planning , year=

  20. [20]

    and Orthey, Andreas and Driess, Danny and Oguz, Ozgur S

    Hartmann, Valentin N. and Orthey, Andreas and Driess, Danny and Oguz, Ozgur S. and Toussaint, Marc , journal=. Long-Horizon Multi-Robot Rearrangement Planning for Construction Assembly , year=

  21. [21]

    Integrated Task and Motion Planning , journal =

    Garrett, Caelan Reed and Chitnis, Rohan and Holladay, Rachel and Kim, Beomjoon and Silver, Tom and Kaelbling, Leslie Pack and Lozano-P\'. Integrated Task and Motion Planning , journal =. 2021 , doi =. https://doi.org/10.1146/annurev-control-091420-084139 , abstract =

    doi:10.1146/annurev-control-091420-084139 2021

  22. [22]

    arXiv preprint arXiv:2011.02608 , year=

    Learning a decentralized multi-arm motion planner , author=. arXiv preprint arXiv:2011.02608 , year=

    Pith/arXiv arXiv 2011

  23. [23]

    Science Robotics , volume=

    Can robots assemble an IKEA chair? , author=. Science Robotics , volume=. 2018 , publisher=

    2018

  24. [24]

    2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Robust task and motion planning for long-horizon architectural construction planning , author=. 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2020 , organization=

    2020

  25. [25]

    Robotics and biomimetics , volume=

    Integrated assembly and motion planning using regrasp graphs , author=. Robotics and biomimetics , volume=. 2016 , publisher=

    2016

  26. [26]

    2023 , eprint=

    Certified Polyhedral Decompositions of Collision-Free Configuration Space , author=. 2023 , eprint=

    2023

  27. [27]

    2023 , eprint=

    Growing Convex Collision-Free Regions in Configuration Space using Nonlinear Programming , author=. 2023 , eprint=

    2023

  28. [28]

    2023 , eprint=

    Shortest Paths in Graphs of Convex Sets , author=. 2023 , eprint=

    2023

  29. [29]

    2022 , eprint=

    Motion Planning around Obstacles with Convex Optimization , author=. 2022 , eprint=

    2022

  30. [30]

    2023 , eprint=

    Non-Euclidean Motion Planning with Graphs of Geodesically-Convex Sets , author=. 2023 , eprint=

    2023

  31. [31]

    A robot path planning framework that learns from experience , year=

    Berenson, Dmitry and Abbeel, Pieter and Goldberg, Ken , booktitle=. A robot path planning framework that learns from experience , year=

  32. [32]

    and Moll, Mark and Okada, Kei and Correll, Nikolaus , booktitle=

    Coleman, David and Şucan, Ioan A. and Moll, Mark and Okada, Kei and Correll, Nikolaus , booktitle=. Experience-based planning with sparse roadmap spanners , year=

  33. [33]

    Proceedings of Robotics: Science and Systems , YEAR =

    Michael Phillips AND Benjamin Cohen AND Sachin Chitta AND Maxim Likhachev , TITLE =. Proceedings of Robotics: Science and Systems , YEAR =

  34. [34]

    AAAI/IAAI , pages=

    Complete anytime beam search , author=. AAAI/IAAI , pages=

  35. [35]

    Information Processing Letters , volume=

    Incremental beam search , author=. Information Processing Letters , volume=. 2013 , publisher=

    2013

  36. [36]

    ICAPS 2023 Heuristics and Search for Domain-Independent Planning Workshop , year=

    Triangle Search: An Anytime Beam Search , author=. ICAPS 2023 Heuristics and Search for Domain-Independent Planning Workshop , year=

    2023

  37. [37]

    International Joint Conference on Artificial Intelligence , pages=

    Awa-a window constrained anytime heuristic search algorithm , author=. International Joint Conference on Artificial Intelligence , pages=. 2007 , organization=

    2007

  38. [38]

    Anytime search in dynamic graphs , journal =

    Maxim Likhachev and Dave Ferguson and Geoff Gordon and Anthony Stentz and Sebastian Thrun , keywords =. Anytime search in dynamic graphs , journal =. 2008 , issn =. doi:https://doi.org/10.1016/j.artint.2007.11.009 , url =

    work page doi:10.1016/j.artint.2007.11.009 2008

  39. [39]

    Journal of Artificial Intelligence Research , volume=

    Anytime heuristic search , author=. Journal of Artificial Intelligence Research , volume=

  40. [40]

    , author=

    Anytime Focal Search with Applications. , author=. IJCAI , pages=

  41. [41]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Anytime Nonparametric A* , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2011 , month=. doi:10.1609/aaai.v25i1.7819 , abstractNote=

    doi:10.1609/aaai.v25i1.7819 2011

  42. [42]

    Anytime incremental planning with E-Graphs , year=

    Phillips, Mike and Dornbush, Andrew and Chitta, Sachin and Likhachev, Maxim , booktitle=. Anytime incremental planning with E-Graphs , year=

  43. [43]

    Proceedings of the International Symposium on Combinatorial Search , volume=

    A-MHA*: anytime multi-heuristic A , author=. Proceedings of the International Symposium on Combinatorial Search , volume=

  44. [44]

    The International Journal of Robotics Research , volume=

    Multi-heuristic a , author=. The International Journal of Robotics Research , volume=. 2016 , publisher=

    2016

  45. [45]

    Proceedings of the 24th International Conference on Artificial Intelligence , pages =

    Phillips, Mike and Narayanan, Venkatraman and Aine, Sandip and Likhachev, Maxim , title =. Proceedings of the 24th International Conference on Artificial Intelligence , pages =. 2015 , isbn =

    2015

  46. [46]

    , author=

    Multiple Sequence Alignment Using Anytime A*. , author=. AAAI/IAAI , pages=

  47. [47]

    Artificial Intelligence , volume=

    Conflict-based search for optimal multi-agent pathfinding , author=. Artificial Intelligence , volume=. 2015 , publisher=

    2015

  48. [48]

    Proceedings of the International Symposium on Combinatorial Search , volume=

    Suboptimal variants of the conflict-based search algorithm for the multi-agent pathfinding problem , author=. Proceedings of the International Symposium on Combinatorial Search , volume=

  49. [49]

    Artificial Intelligence: 5th RAAI Summer School, Dolgoprudny, Russia, July 4--7, 2019, Tutorial Lectures , pages=

    Multi-agent path finding--an overview , author=. Artificial Intelligence: 5th RAAI Summer School, Dolgoprudny, Russia, July 4--7, 2019, Tutorial Lectures , pages=. 2019 , publisher=

    2019

  50. [50]

    Palo Alto, CA , year=

    Principles of Artificial Intelligence Tioga Publishing , author=. Palo Alto, CA , year=

  51. [51]

    Algorithmica , volume=

    On multiple moving objects , author=. Algorithmica , volume=. 1987 , publisher=

    1987

  52. [52]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Searching with consistent prioritization for multi-agent path finding , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  53. [53]

    Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence,

    Priority Inheritance with Backtracking for Iterative Multi-agent Path Finding , author=. Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence,. 2019 , month=. doi:10.24963/ijcai.2019/76 , url=

    work page doi:10.24963/ijcai.2019/76 2019

  54. [54]

    2023 , author=

    LaCAM: Search-Based Algorithm for Quick Multi-Agent Pathfinding , booktitle=. 2023 , author=

    2023

  55. [55]

    Autonomous Robots , volume=

    drrt*: Scalable and informed asymptotically-optimal multi-robot motion planning , author=. Autonomous Robots , volume=. 2020 , publisher=

    2020

  56. [56]

    IEEE Robotics and Automation Letters , volume=

    Representation-optimal multi-robot motion planning using conflict-based search , author=. IEEE Robotics and Automation Letters , volume=. 2021 , publisher=

    2021

  57. [57]

    arXiv preprint arXiv:2309.10665 , year=

    Fast-dRRT*: Efficient Multi-Robot Motion Planning for Automated Industrial Manufacturing , author=. arXiv preprint arXiv:2309.10665 , year=

    Pith/arXiv arXiv

  58. [58]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    MAPF-LNS2: fast repairing for multi-agent path finding via large neighborhood search , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  59. [59]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Multi-agent path finding for large agents , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  60. [60]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Eecbs: A bounded-suboptimal search for multi-agent path finding , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  61. [61]

    2023 , eprint=

    Constant-time Motion Planning with Anytime Refinement for Manipulation , author=. 2023 , eprint=

    2023

  62. [62]

    1988 , publisher=

    The complexity of robot motion planning , author=. 1988 , publisher=

    1988

  63. [63]

    , journal=

    Pearl, Judea and Kim, Jin H. , journal=. Studies in Semi-Admissible Heuristics , year=

  64. [64]

    Proceedings of the AAAI Conference on Artificial Intelligence , author=

    Effective Integration of Weighted Cost-to-Go and Conflict Heuristic within Suboptimal CBS , volume=. Proceedings of the AAAI Conference on Artificial Intelligence , author=. 2023 , month=. doi:10.1609/aaai.v37i10.26381 , number=

    work page doi:10.1609/aaai.v37i10.26381 2023

  65. [65]

    Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence , pages=

    Iterative-deepening conflict-based search , author=. Proceedings of the Twenty-Ninth International Conference on International Joint Conferences on Artificial Intelligence , pages=

  66. [66]

    Artificial Intelligence , volume=

    Lifelong planning A∗ , author=. Artificial Intelligence , volume=. 2004 , publisher=

    2004

  67. [67]

    IEEE Transactions on Robotics , volume=

    Fast replanning for navigation in unknown terrain , author=. IEEE Transactions on Robotics , volume=. 2005 , publisher=

    2005

  68. [68]

    , author=

    A Generalized Framework for Lifelong Planning A* Search. , author=. ICAPS , pages=

  69. [69]

    and Chitta, Sachin and Likhachev, Maxim , booktitle=

    Cohen, Benjamin J. and Chitta, Sachin and Likhachev, Maxim , booktitle=. Search-based planning for manipulation with motion primitives , year=

  70. [70]

    arXiv preprint arXiv:1404.3785 , year=

    Reducing the barrier to entry of complex robotic software: a moveit! case study , author=. arXiv preprint arXiv:1404.3785 , year=

    Pith/arXiv arXiv

  71. [71]

    2022 , eprint=

    Cooperative Task and Motion Planning for Multi-Arm Assembly Systems , author=. 2022 , eprint=

    2022

  72. [72]

    Kiril Solovey and Oren Salzman and Dan Halperin , title =. Int. J. Robotics Res. , volume =. 2016 , url =. doi:10.1177/0278364915615688 , timestamp =

    doi:10.1177/0278364915615688 2016

  73. [73]

    Stuckey and Hang Ma and Graeme Gange and Sven Koenig , keywords =

    Jiaoyang Li and Daniel Harabor and Peter J. Stuckey and Hang Ma and Graeme Gange and Sven Koenig , keywords =. Pairwise symmetry reasoning for multi-agent path finding search , journal =. 2021 , issn =. doi:https://doi.org/10.1016/j.artint.2021.103574 , url =

    arXiv 2021

  74. [74]

    FCL: A general purpose library for collision and proximity queries , year=

    Pan, Jia and Chitta, Sachin and Manocha, Dinesh , booktitle=. FCL: A general purpose library for collision and proximity queries , year=

  75. [75]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Generalized and sub-optimal bipartite constraints for conflict-based search , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  76. [76]

    International Conference on Machine Learning , pages=

    Planning with Diffusion for Flexible Behavior Synthesis , author=. International Conference on Machine Learning , pages=. 2022 , organization=

    2022

  77. [77]

    34th International Conference on Automated Planning and Scheduling , year =

    Accelerating Search-Based Planning for Multi-Robot Manipulation by Leveraging Online-Generated Experiences , author =. 34th International Conference on Automated Planning and Scheduling , year =

  78. [78]

    Frontiers in Robotics and AI , volume=

    Let's push things forward: A survey on robot pushing , author=. Frontiers in Robotics and AI , volume=. 2020 , publisher=

    2020

  79. [79]

    Manipulator grasping and pushing operations , author=

  80. [80]

    Proceedings 1995 IEEE/RSJ international conference on intelligent robots and systems

    Cooperative multi-robot box-pushing , author=. Proceedings 1995 IEEE/RSJ international conference on intelligent robots and systems. Human robot interaction and cooperative robots , volume=. 1995 , organization=

    1995

Showing first 80 references.