Pith. sign in

REVIEW 3 major objections 6 minor 43 references

Coordinated Multi-Robot Disassembly for Makespan Optimization of Large-Scale Assemblies

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a coordinated scheduler, CoMuDi, lets teams of up to nine robots disassemble assemblies of up to 49 parts with makespan decreasing as the team grows, while outperforming a fixed-time RRT* baseline in success rate and…

desk verdict A coherent MR-TAMP systems paper with a real gap in continuous collision checking during object extraction; worth reviewing but needs fixes. read the letter →

arxiv 2608.05830 v1 pith:FTRK66MJ submitted 2026-08-06 cs.RO

Niklas Hargus , Andreas Orthey , Marc Toussaint This is my paper · ORCID
classification cs.RO
keywords multi-robotdisassemblymakespanoptimizationtaskandmotionplanningspace-timeRRT*temporalconstraintpropagationdependencygraphobjectextractionmobilemanipulators
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 introduces CoMuDi, a planning framework that coordinates a team of robots to take apart large assemblies. The central claim is that a prioritized task queue, propagation of temporal constraints along a dependency graph, and a space-time motion planner per task let up to nine simulated robots disassemble assemblies of up to 49 parts while the makespan falls as the team grows. The authors argue this matters because disassembly is the bottleneck for recycling, repair, and remanufacturing, where several robots must work in close quarters without colliding and without standing idle. Across six test scenarios, CoMuDi produced plans with low idle time, and in a direct comparison on a tower assembly it achieved a higher success rate and lower makespan than a fixed-time RRT* baseline. The paper concludes that CoMuDi reliably solves large-scale disassembly problems and is a step toward general-purpose robotic disassembly systems.

What carries the argument

The central object is CoMuDi itself, a prioritized sequential task-and-motion-planning framework. Its load-bearing pieces are: a dependency graph whose leaves define which parts are currently removable; a per-robot FIFO task queue that assigns leaves to free robots; temporal constraint propagation that passes attach and detach times from a completed vertex to its successors, letting successor pick and place phases start earlier; a generalized disassembly task that bundles approach, attach, optional extraction and insertion along precomputed waypoint paths, detach, and exit into one arrival-time-minimizing query; and ST-RRT*, the space-time RRT* planner that minimizes arrival time per query while treating other robots' committed trajectories as dynamic obstacles. The mechanism works by interleaving these levels: queues decide what each robot does next, temporal propagation tightens the earliest start times, and space-time planning makes each individual motion as short as possible, so the global makespan falls without centralized joint planning.

What would settle it

Give CoMuDi an assembly in which one part is geometrically interlocked so that it requires simultaneous pulling by two robots, or in which a precomputed removal path clips the surrounding geometry, and run the planner: if it returns a plan that fails at execution, or silently ignores the multi-robot extraction requirement, then the claim that CoMuDi reliably coordinates disassembly of arbitrary input assemblies fails.

Watch

Extended reading notes

Core claim

In the paper's own terms, the discovery is that a coordinated multi-robot disassembly planner can turn an assembly, a dependency graph, and per-part removal and insertion paths into collision-free trajectories for a variable-size robot team while optimizing the makespan. CoMuDi does this by generalizing the disassembly of one part into a single compound task — approach, attach, optional extraction along a precomputed path, transport, optional insertion, detach, and exit — and scheduling those tasks through a prioritized FIFO queue. Temporal constraints derived from attach and detach times are propagated from completed vertices to their successors in the dependency graph, so a later task can begin its pick phase as soon as the predecessor part is picked rather than waiting for full placement. Each individual move is planned with ST-RRT*, which treats time as a planning dimension and minimizes arrival time, and the experiments show the combination yields decreasing makespan with up to nine robots on assemblies of up to 49 parts, with lower makespan and 100% success versus fixed-time RRT* in the tower scenario.

Load-bearing premise

The plan's correctness rests on the given dependency graph and per-part removal and insertion paths being right, and on every part being removable by one robot alone; CoMuDi takes these as inputs and does not check them.

Editorial extensions

If this is right

  • Makespan decreases monotonically as robot count grows from 1 to 9 in all six test scenarios, so adding robots does not degrade coordination despite the tighter workspace.
  • Three to five robots often sit on the Pareto frontier of makespan versus computation time, giving practitioners a default team size before computation cost climbs.
  • Using a planner that adapts its time horizon in space-time (ST-RRT*) substantially outperforms fixed-time RRT* in both success rate and makespan in crowded conditions.
  • The scheduling timelines show robots remain idle only when the scenario structure forces waits, indicating the temporal propagation is doing the intended work.
  • Because CoMuDi treats removal and insertion paths as inputs, it can be paired with any disassembly-sequence planner that supplies such paths, broadening its applicability.

Reading between the lines

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

  • The same queue-plus-temporal-propagation machinery should transfer to assembly (the reverse operation) and to other precedence-constrained manipulation problems, since the scheduling logic does not depend on task direction.
  • The dependency graph and the per-part extraction paths are the real bottleneck; a system that synthesizes them automatically from geometry would turn CoMuDi into a complete pipeline rather than a coordinator.
  • A physical implementation would likely stress the two simplifying assumptions — omnidirectional gripper and non-physical simulation — so the planner's real-world value hinges on how well those assumptions hold outside simulation.
  • A depth-aware assignment policy, which the paper lists as a limitation, could plausibly turn the motor and gearbox scenarios from chance-driven to two-sided disassembly and further lower makespan.
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 / 6 minor

Summary. The paper proposes CoMuDi, a coordinated multi-robot task and motion planning framework for disassembly. CoMuDi takes a robot team, an assembly of objects, and a dependency graph as input, creates compound tasks (move, attach/detach, extraction/insertion, exit), and coordinates robots through a prioritized task queue, temporal constraint propagation, and a time-dependent collision-checking scheme. Individual robot motions are planned with ST-RRT*, and the system is evaluated on six disassembly scenarios with up to 49 parts and up to 9 robots, reporting makespan, computation time, Pareto fronts, scheduling timelines, failure rates, and a comparison between ST-RRT* and RRT* on the tower scenario. The central claims are that CoMuDi returns collision-free disassembly trajectories and reliably solves large-scale assemblies with decreasing makespan as the robot count grows.

Significance. If the claims are established, CoMuDi would be a useful contribution to multi-robot disassembly TAMP, combining existing modules (ST-RRT*, scale-invariant sampling, prioritized planning) into a coherent framework and providing an extensive six-scenario evaluation with public videos. The implementation in OMPL and the explicit documentation of tunable parameters are strengths. However, the central assertions—that CoMuDi returns collision-free trajectories and reliably solves large-scale assemblies—are not fully supported by the current evidence: continuous path segments in object extraction are never collision-checked, and the main benchmark reports non-negligible failure rates in several scenarios. The significance is therefore conditional on closing these gaps and recalibrating the claims.

major comments (3)
  1. [§IV-D, Eq. (4)–(6), Alg. 4] The claim that CoMuDi returns collision-free trajectories is not secured by Algorithm 4. The space-time waypoint problem is decomposed into an IK subproblem (Eq. 5) and a temporal subproblem (Eq. 6), and the validity check at Alg. 4 lines 14–16 tests only the single configuration q_i at a single time t. The continuous configuration-space segment from q_{i-1} at t_{i-1} to q_i at t_i, including the swept volume of the carried object, is never checked against static obstacles or other robots' trajectories. Because later robots treat the interpolated trajectory as the dynamic obstacle (Alg. 8), an accepted plan can be physically infeasible. The statement that 'the waypoint density of the precomputed paths is sufficient' is an empirical assertion, and no experiment verifies intermediate path segments. The paper should either rephrase the central contribution as discrete-time validation or augment the algorithm with continuous collision checking on the connecting segments.
  2. [§VI-E, Figs. 12–13] The planner comparison that supports the claim that CoMuDi with ST-RRT* is superior to RRT* is not yet a convincing fair benchmark. The RRT* time windows (5s and 10s) are stated to have been 'empirically selected based on the convergence behavior of ST-RRT*' (§VI-E), which introduces a bias in favor of the method being advocated; the windows are also not known a priori in practice. In addition, the comparison is run only on the tower scenario, so the broader claim that the combination 'consistently' yields high success rates and low makespan across the six scenarios is not directly supported by the comparison.
  3. [§VI-D.4, Fig. 11, §VII] The success rates in the main benchmark are lower than the abstract and conclusion imply. For example, the crate scenario shows success around 30–56% depending on robot count, the motor scenario 36–64%, and the tower 70–89%. The text states that failure modes 'do not stop the algorithm, but they represent wasted computation time,' yet the reported success percentages indicate that a substantial fraction of runs do not complete the disassembly within the trial. The claims of 'reliably solve' (abstract) and 'consistently to successful results' (§VII) should be reconciled with these numbers, or the claims should be qualified to reflect the measured success rates and the precise retry mechanism should be described.
minor comments (6)
  1. [§VI-B] The parameter name 'numberO-fExitConfigurationsto' appears to be a typo for 'numberOfExitConfigurations'.
  2. [§VI-E] The reference to 'Fig. 8' for the convergence behavior of ST-RRT* is confusing, since Fig. 8 displays makespan and computation time per scenario rather than a convergence study; please clarify the intended reference.
  3. [Fig. 11 caption] The legend text 'Success Exit fail Pull fail Plan to object fail Plan to goal fail' lacks separators between categories; please format the legend so each category is clearly distinct.
  4. [§VII-A and §VIII] The conclusion's phrase 'near optimal results' is too strong given the paper's own limitations list (depth-aware assignment, clearance cost, grouping, global optimality); a more modest phrasing such as 'good makespan reduction in the tested scenarios' would be appropriate.
  5. [§III-A] The paper assumes the dependency graph, per-object removal paths, and insertion paths are given and correct, and that each task is executable by a single robot; this scope should be stated explicitly in the problem formulation so that the contribution is not interpreted as verifying these inputs.
  6. [Abstract and §I] The phrase 'thereby helping us minimize overall makespan' is imprecise: the algorithm minimizes arrival time for individual tasks, and the global makespan is only empirically reduced; the text should avoid implying a formal makespan-minimization guarantee.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: CoMuDi is an empirical planning system; its reported makespan and success results are measured simulation outputs, not quantities reconstructed from its inputs by construction.

full rationale

The paper does not claim a mathematical derivation of its final makespan from the dependency graph or precomputed paths. CoMuDi takes as input a robot team, an assembly, and a dependency graph (Alg. 1), then executes a sequence of generalized disassembly tasks (Alg. 2) and measures the resulting makespan and success rate in simulation. No equation in the paper reduces the reported makespan to a fitted parameter or to the input data. The object extraction routine (Alg. 4) checks only discrete waypoint states, which is a soundness gap in the claimed collision-free trajectories, but that is not circularity because the check is not equivalent to the output by definition. The RRT* time windows are empirically selected, but they are benchmark settings, not fitted parameters renamed as predictions. The self-group citations—ST-RRT* [10], scale-invariant sampling [2], and multi-robot assembly [12]—are used as independently published algorithmic modules, and the paper's central contribution, the prioritized task queue and temporal constraint propagation, is evaluated directly by simulation experiments rather than justified solely by those citations. No uniqueness theorem or prior result is invoked to forbid alternative choices in a way that makes the conclusion depend on the authors' own prior work. Thus the central claim is an empirical system claim with independent experimental content, and no circular step can be exhibited with a specific equation or fitted parameter reduction.

Assumptions & free parameters 10 free parameters · 6 assumptions · 0 invented entities

The planner builds on precomputed inputs, several hand-set constants, and deliberately simplified physical assumptions. It introduces no new physical entities. The algorithm depends on ST-RRT* and scale-invariant sampling modules from the same research group, which have independent published support but are not re-verified here.

free parameters (10)
  • maxTryNumber = 1000
    Maximum surface-sampling and IK attempts in grasp pose generation (Alg. 9, Sec VI.B); hand-set.
  • maxNumGoals = 5
    Maximum number of goal configurations collected before grasp pose generation returns (Alg. 9, Sec VI.B).
  • delta_t (collision sweep) = 0.1
    Time increment in forward collision sweeps (Alg. 10) and waypoint time search (Sec VI.B); affects collision-checking resolution.
  • n_retry = 3
    Maximum number of planning attempts for move and exit tasks (Sec VI.B).
  • numberOfExitConfigurations = 10
    Number of candidate exit configurations sampled per retry in the exit task (Alg. 5, Sec VI.B).
  • t_max (extraction) = 10
    Upper bound on the time window when searching for a valid waypoint time in the object extraction task (Alg. 4, Sec VI.B).
  • epsilon = 0.1
    Minimum time separation required between consecutive waypoint states (Sec VI.B).
  • lambda = 1
    Weight on joint-displacement regularization in the IK subproblem (Eq. 5, Sec VI.B).
  • t_timelimit = 10s
    Maximum computation time used as termination condition for each planner query (Sec VI.A).
  • RRT* time windows = 5s and 10s
    Empirically selected based on the convergence behavior of ST-RRT* in the tower scenario (Sec VI.E); directly affects the fairness of the planner comparison.
assumptions (6)
  • domain assumption The dependency graph G is a precomputed, correct DAG encoding all precedence constraints.
    CoMuDi receives G as input (Sec III-A) and never validates it; an incorrect graph produces invalid disassembly orders.
  • domain assumption Each disassembly task can be executed by a single robot; no cooperative multi-robot manipulation is required.
    Task assignment in Alg. 1 pairs one robot to one leaf object; simultaneous holding or insertion by multiple robots is not modeled.
  • domain assumption The omnidirectional gripper abstraction makes grasp planning unnecessary.
    Sec III-A(b) states the end-effector can grasp from any angle; this removes grasp feasibility from the planner and is load-bearing for the claimed success.
  • domain assumption Non-physical simulation is a valid proxy for execution; contact effects such as stick-slip and object resting are ignored.
    Stated in Sec III-A(b); if contacts invalidate planned trajectories, the plans are not directly executable in reality.
  • domain assumption Removal and insertion paths are precomputed correctly and provided as input.
    Sec IV states both paths are precomputed from assembly geometry and given to CoMuDi; the extraction task assumes they exist and are feasible.
  • domain assumption ST-RRT* provides the claimed asymptotically optimal arrival-time planning behavior from [10].
    CoMuDi delegates move, goal, and exit planning to ST-RRT*; the makespan results inherit its behavioral properties without being re-proven in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Coordinated Multi-Robot Disassembly for Makespan Optimization of Large-Scale Assemblies." pith.science (2026). https://pith.science/paper/FTRK66MJ

@misc{pith2026260805830,
  author       = {Pith},
  title        = {Pith review of: Coordinated Multi-Robot Disassembly for Makespan Optimization of Large-Scale Assemblies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FTRK66MJ}},
  note         = {Machine review of arXiv:2608.05830}
}
read the original abstract

Multi-robot task and motion planning for disassembly tasks requires robots to operate in confined workspaces while coordinating their motions with other robots. To tackle this problem, we propose a planning method called coordinated multi-robot disassembly (CoMuDi). CoMuDi coordinates a team of robots for disassembly tasks. The input is a team of robots, an assembly of objects, and a dependency graph. Based on this information, we create compound tasks for pick, place, and exit motions. By propagating temporal constraints, we ensure that each robot can start and end their tasks as early as possible while avoiding collisions with nearby robots. By integrating the space-time RRT* planner (ST-RRT*) into CoMuDi, we ensure that individual tasks minimize arrival time and thereby help us minimize overall makespan. We compare the performance of CoMuDi using both ST-RRT* and RRT* planners with varying time bounds, demonstrating that the combination of CoMuDi and ST-RRT* leads to a higher success rate while minimizing makespan. Finally, we evaluate CoMuDi on six assemblies with up to 49 pieces and up to 9 robots. In those scenarios, we show that CoMuDi returns robot paths that exhibit low idle times, thereby demonstrating that CoMuDi can reliably solve large-scale assemblies.

Figures

Figures reproduced from arXiv: 2608.05830 by the authors.

Figure 1
Figure 1. We develop a coordinated disassembly multi-robot planner to solve large-scale multi-robot task and motion planning problems. Pictured is our framework coordinating 9 mobile manipulators which are disassembling a gearbox with 33 pieces. Left: Gearbox in assembled state, robots are approaching pick positions. Middle: Robots during disassembly with some robots transporting parts, while others are returning for the next… view at source ↗
Figure 2
Figure 2. Overview of the temporal multi-robot TAMP framework for disassembly tasks. As input, we require a team of robots and the assembly consisting of objects, together with a dependency tree specifying the possible execution order of the objects. This tree is used in the high-level coordinator to extract the next possible tasks by extracting the leaves. II. RELATED WORK Our framework for multi-robot task and motion planni… view at source ↗
Figure 3
Figure 3. Example of an assembly with a lid, two bolts, and panel. The dependency graph shows the assembly constraints. depend on the lid. All leaf nodes in the graph have no dependencies on other nodes. b) Assumptions: The goal of this paper is to coordinate a multi-robot team for disassembly. To study this in isolation, we make two assumptions to simplify the problem: • The end-effector of each robot is modeled as an omnidi… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Example timeline with pick/place time constraints for two robots A and B. new leaf nodes are available for manipulation. This is a conservative approach, since a successor object can begin its planning before its predecessor has completed its full task sequence. Waitin…
Figure 5
Figure 5. Figure 5: Storage representation of attachment data a time window, consisting of the attach and detach time (Line 10). The relevant attachData is then finally returned (Line 11). 2) Setting the Simulation State: To execute collision check￾ing at a specific time point, every obje…
Figure 6
Figure 6. Figure 6: Timeline illustration for future collision check Algorithm 10: Check Future Collision Input: robot, q, tat Parameters: ∆t Output: validConfiguration 1. t ← tat 2. tmax ← getMaxTrajectoryTime() 3. while t ≤ tmax do 4. t ← t + ∆t 5. setSimTime(t) 6. setConfiguration(robo…
Figure 7
Figure 7. Figure 7: Scenarios used in the disassembly experiments. D. Multi-Robot Disassembly Benchmark For each scenario, we report on computation time, makespan (total time elapsed from start to finish), Pareto front (makespan versus computation time), scheduling timeline, and analyze t…
Figure 8
Figure 8. Figure 8: Scaling behavior for the two metrics of makespan and computation time for different robot counts in each disassembly scenario. while varying the number of robots from one to nine. Each run has a timeout of 1000 seconds, 10 runs per robot count, and an individual planne…
Figure 9
Figure 9. Figure 9: Pareto front for each disassembly scenario. and computation time, thereby allowing users of the system to carefully weigh the number of robots needed to accomplish a certain disassembly task. The second result concerns the comparison between ST￾RRT* and RRT*. This comp…
Figure 10
Figure 10. Figure 10: Scheduling timelines of the disassembly process for each scenario. Tasks are color coded as Pick, Place, Pull, and Exit. objects could be grouped so that the IK solver can try to fit as many manipulators as possible into the workspace simultaneously. • Global Optimali…
Figure 11
Figure 11. Figure 11: Failure rates across the different disassembly scenarios. Success Exit fail Pull fail Plan to object fail Plan to goal fail 1 2 3 4 5 6 7 8 9 Number of robots 20 40 60 80 100 120 140 160 180 Makespan (s) [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: Makespan comparison between ST-RRT* and RRT*, show￾ing ST-RRT* , RRT* 5s, and RRT* 10s. 1 2 3 4 5 6 7 8 9 Number of robots 0 20 40 60 80 100 Success rate (%) [PITH_FULL_IMAGE:figures/full_fig_p015_12.png]
Figure 13
Figure 13. Figure 13: Success rate comparison of ST-RRT* and RRT*, showing ST-RRT*, RRT* 5s, and RRT* 10s. [4] M. Cˇ ap, P. Nov ´ ak, A. Kleiner, and M. Seleck ´ y, “Prioritized planning ´ algorithms for trajectory coordination of multiple mobile robots,” IEEE Transactions on Automation Sc…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 37 canonical work pages

  1. [1]

    Multi-agent pathfinding with continuous time,

    A. Andreychuk, K. Yakovlev, P. Surynek, D. Atzmon, and R. Stern, “Multi-agent pathfinding with continuous time,”Artificial Intelligence, vol. 305, p. 103662, 2022

    2022

  2. [2]

    Scale-invariant sampling in multi-arm bandit motion planning for object extraction,

    S. B. Bayraktar, A. Orthey, and M. Toussaint, “Scale-invariant sampling in multi-arm bandit motion planning for object extraction,” inWorld Symposium on the Algorithmic Foundations of Robotics, 2026

    work page 2026

  3. [3]

    Finding and optimizing solvable priority schemes for decoupled path planning techniques for teams of mobile robots,

    M. Bennewitz, W. Burgard, and S. Thrun, “Finding and optimizing solvable priority schemes for decoupled path planning techniques for teams of mobile robots,”Robotics and Autonomous Systems, vol. 41, no. 2-3, pp. 89–99, 2002. 15 1 2 3 4 5 6 7 8 9 Number of robots 0 20 40 60 80 100Percentage79% 86% 83% 79% 76% 78% 70% 80% 70% 21% 14% 17% 21% 23% 19% 30% 19%...

    work page 2002

  4. [4]

    Prioritized planning algorithms for trajectory coordination of multiple mobile robots,

    M. ˇC´ap, P. Nov ´ak, A. Kleiner, and M. Seleck ´y, “Prioritized planning algorithms for trajectory coordination of multiple mobile robots,”IEEE Transactions on Automation Science and Engineering, vol. 12, no. 3, pp. 835–849, 2015

    work page 2015

  5. [5]

    Cooperative task and motion planning for multi-arm assembly systems,

    J. Chen, J. Li, Y . Huang, C. Garrett, D. Sun, C. Fan, A. Hofmann, C. Mueller, S. Koenig, and B. C. Williams, “Cooperative task and motion planning for multi-arm assembly systems,”arXiv preprint arXiv:2203.02475, 2022

    arXiv 2022

  6. [6]

    On multiple moving objects,

    M. Erdmann and T. Lozano-P ´erez, “On multiple moving objects,” Algorithmica, vol. 2, no. 1, pp. 477–521, 1987

    work page 1987

  7. [7]

    Pddlstream: Integrating symbolic planners and blackbox samplers via optimistic adaptive planning,

    C. R. Garrett, T. Lozano-P ´erez, and L. P. Kaelbling, “Pddlstream: Integrating symbolic planners and blackbox samplers via optimistic adaptive planning,” inProceedings of the international conference on automated planning and scheduling, vol. 30, 2020, pp. 440–448

    work page 2020

  8. [8]

    Integrated task and motion planning,

    C. R. Garrett, R. Chitnis, R. Holladay, B. Kim, T. Silver, L. P. Kaelbling, and T. Lozano-P ´erez, “Integrated task and motion planning,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, pp. 265– 293, 2021

    work page 2021

Show all 43 references
  1. [9]

    Pddl - the planning domain definition language,

    M. Ghallab, C. Knoblock, D. Wilkins, A. Barrett, D. Christianson, M. Friedman, C. Kwok, K. Golden, S. Penberthy, D. Smith, Y . Sun, and D. Weld, “Pddl - the planning domain definition language,” Yale Center for Computational Vision and Control, Tech. Rep. CVC TR-98- 003/DCS TR...

    1998

  2. [10]

    ST-RRT*: Asymptotically-optimal bidirectional motion planning through space- time,

    F. Grothe, V . N. Hartmann, A. Orthey, and M. Toussaint, “ST-RRT*: Asymptotically-optimal bidirectional motion planning through space- time,” in2022 International Conference on Robotics and Automation (ICRA), 2022, pp. 3314–3320

    2022

  3. [11]

    Towards computing low-makespan solutions for multi-arm multi-task planning problems,

    V . N. Hartmann and M. Toussaint, “Towards computing low-makespan solutions for multi-arm multi-task planning problems,” inICAPS Work- shop on Planning and Robotics (PlanRob), 2023

    2023

  4. [12]

    Long-horizon multi-robot rearrangement planning for construction as- sembly,

    V . N. Hartmann, A. Orthey, D. Driess, O. S. Oguz, and M. Toussaint, “Long-horizon multi-robot rearrangement planning for construction as- sembly,”IEEE Transactions on Robotics, vol. 39, no. 1, pp. 239–252, 2023

    2023

  5. [13]

    Multi-modal motion planning in non- expansive spaces,

    K. Hauser and J.-C. Latombe, “Multi-modal motion planning in non- expansive spaces,”The International Journal of Robotics Research, vol. 29, no. 7, pp. 897–915, 2010

    2010

  6. [14]

    Randomized kin- odynamic motion planning with moving obstacles,

    D. Hsu, R. Kindel, J.-C. Latombe, and S. Rock, “Randomized kin- odynamic motion planning with moving obstacles,”The International Journal of Robotics Research, vol. 21, no. 3, pp. 233–255, 2002

    2002

  7. [15]

    Hierarchical task and motion planning in the now,

    L. P. Kaelbling and T. Lozano-Perez, “Hierarchical task and motion planning in the now,” in2011 IEEE International Conference on Robotics and Automation, 2011, pp. 1470–1477

    2011

  8. [16]

    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

  9. [17]

    Disassembly sequence planning based on a genetic algorithm,

    M. Kheder, M. Trigui, and N. Aifaoui, “Disassembly sequence planning based on a genetic algorithm,”Proceedings of the Institution of Mechan- ical Engineers, Part C: Journal of Mechanical Engineering Science, vol. 16 229, 11 2014

    2014

  10. [18]

    Disassembly sequencing using genetic algorithm,

    E. Kongar and S. M. Gupta, “Disassembly sequencing using genetic algorithm,”The International Journal of Advanced Manufacturing Tech- nology, vol. 30, no. 5-6, pp. 497–506, 2006

    2006

  11. [19]

    Dealing with difficult instances of object rearrangement,

    A. Krontiris and K. E. Bekris, “Dealing with difficult instances of object rearrangement,” inProceedings of Robotics: Science and Systems (RSS), 2015

    2015

  12. [20]

    Disassembly sequencing: A survey,

    A. Lambert, “Disassembly sequencing: A survey,”International Journal of Production Research, vol. 41, pp. 3721–3759, 11 2003

    2003

  13. [21]

    A genetic algorithm for generating optimal assembly plans,

    B. Lazzerini and F. Marcelloni, “A genetic algorithm for generating optimal assembly plans,”Artificial Intelligence in Engineering, vol. 14, no. 4, pp. 319–329, 2000

    2000

  14. [22]

    Searching with consistent prioritization for multi-agent path finding,

    H. Ma, D. Harabor, P. J. Stuckey, J. Li, and S. Koenig, “Searching with consistent prioritization for multi-agent path finding,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 33, no. 1, 2019, pp. 7643–7650

    2019

  15. [23]

    A general task and motion planning framework for multiple manipulators,

    T. Pan, A. M. Wells, R. Shome, and L. E. Kavraki, “A general task and motion planning framework for multiple manipulators,” in2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021, pp. 3168–3174

    2021

  16. [24]

    SIPP: Safe interval path planning for dynamic environments,

    M. Phillips and M. Likhachev, “SIPP: Safe interval path planning for dynamic environments,” inProceedings of the IEEE International Conference on Robotics and Automation (ICRA), 2011, pp. 5628–5635

    2011

  17. [25]

    Fostering end- of-life utilization by information-driven robotic disassembly,

    H. Poschmann, H. Br ¨uggemann, and D. Goldmann, “Fostering end- of-life utilization by information-driven robotic disassembly,”Procedia CIRP, vol. 98, pp. 282–287, 2021

    2021

  18. [26]

    Conflict-based search for optimal multi-agent pathfinding,

    G. Sharon, R. Stern, A. Felner, and N. R. Sturtevant, “Conflict-based search for optimal multi-agent pathfinding,”Artificial Intelligence, vol. 219, pp. 40–66, 2015

    2015

  19. [27]

    dRRT*: Scalable and informed asymptotically-optimal multi-robot mo- tion planning,

    R. Shome, K. Solovey, A. Dobson, D. Halperin, and K. E. Bekris, “dRRT*: Scalable and informed asymptotically-optimal multi-robot mo- tion planning,”Autonomous Robots, vol. 44, no. 3–4, pp. 443–467, 2020

    2020

  20. [28]

    Manipulation planning with probabilistic roadmaps,

    T. Sim ´eon, J.-P. Laumond, J. Cort ´es, and A. Sahbani, “Manipulation planning with probabilistic roadmaps,”The International Journal of Robotics Research, vol. 23, no. 7-8, pp. 729–746, 2004

    2004

  21. [29]

    Time-based RRT algorithm for rendezvous planning of two dynamic systems,

    A. Sintov and A. Shapiro, “Time-based RRT algorithm for rendezvous planning of two dynamic systems,” inProceedings of the IEEE Interna- tional Conference on Robotics and Automation (ICRA), 2014, pp. 6745– 6750

    2014

  22. [30]

    Disassembly sequence structure graphs: An optimal approach for multiple-target selective disassembly sequence planning,

    S. Smith, G. Smith, and W.-H. Chen, “Disassembly sequence structure graphs: An optimal approach for multiple-target selective disassembly sequence planning,”Advanced Engineering Informatics, vol. 26, no. 2, pp. 306–316, 2012

    2012

  23. [31]

    Finding a needle in an exponential haystack: Discrete RRT for exploration of implicit roadmaps in multi-robot motion planning,

    K. Solovey, O. Salzman, and D. Halperin, “Finding a needle in an exponential haystack: Discrete RRT for exploration of implicit roadmaps in multi-robot motion planning,”The International Journal of Robotics Research, vol. 35, no. 5, pp. 501–513, 2016

    2016

  24. [32]

    Combined task and motion planning through an extensible planner- independent interface layer,

    S. Srivastava, E. Fang, L. Riano, R. Chitnis, S. Russell, and P. Abbeel, “Combined task and motion planning through an extensible planner- independent interface layer,” in2014 IEEE International Conference on Robotics and Automation (ICRA), 2014, pp. 639–646

    2014

  25. [33]

    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

    2012

  26. [34]

    Assemble them all: Physics-based planning for generaliz- able assembly by disassembly,

    Y . Tian, J. Xu, Y . Li, J. Luo, S. Sueda, H. Li, K. D. D. Willis, and W. Matusik, “Assemble them all: Physics-based planning for generaliz- able assembly by disassembly,”ACM Transactions on Graphics, vol. 41, no. 6, 2022

    2022

  27. [35]

    Logic-geometric programming: An optimization-based approach to combined task and motion planning,

    M. Toussaint, “Logic-geometric programming: An optimization-based approach to combined task and motion planning,” inProceedings of the 24th International Conference on Artificial Intelligence, ser. IJCAI’15. AAAI Press, 2015, pp. 1930–1936

    2015

  28. [36]

    A tutorial on Newton methods for constrained trajectory opti- mization and relations to SLAM, Gaussian process smoothing, opti- mal control, and probabilistic inference,

    ——, “A tutorial on Newton methods for constrained trajectory opti- mization and relations to SLAM, Gaussian process smoothing, opti- mal control, and probabilistic inference,” inGeometric and Numerical Foundations of Movements, ser. Springer Tracts in Advanced Robotics. Spring...

    2017

  29. [37]

    Multi-bound tree search for logic- geometric programming in cooperative manipulation domains,

    M. Toussaint and M. Lopes, “Multi-bound tree search for logic- geometric programming in cooperative manipulation domains,” in2017 IEEE International Conference on Robotics and Automation (ICRA), 2017, pp. 4044–4051

    2017

  30. [38]

    Differen- tiable physics and stable modes for tool-use and manipulation planning,

    M. Toussaint, K. R. Allen, K. A. Smith, and J. B. Tenenbaum, “Differen- tiable physics and stable modes for tool-use and manipulation planning,” inProceedings of Robotics: Science and Systems (RSS), 2018

    2018

  31. [39]

    Roadmap-based motion planning in dynamic environments,

    J. van den Berg and M. H. Overmars, “Roadmap-based motion planning in dynamic environments,”IEEE Transactions on Robotics, vol. 21, no. 5, pp. 885–897, 2005

    2005

  32. [40]

    Planning time-minimal safe paths amidst unpredictably moving obstacles,

    ——, “Planning time-minimal safe paths amidst unpredictably moving obstacles,”The International Journal of Robotics Research, vol. 27, no. 11-12, pp. 1274–1294, 2008

    2008

  33. [41]

    Asymptotically optimal planning under piecewise-analytic constraints,

    W. Vega-Brown and N. Roy, “Asymptotically optimal planning under piecewise-analytic constraints,” inAlgorithmic Foundations of Robotics XII (WAFR 2016), ser. Springer Proceedings in Advanced Robotics. Springer, 2020, vol. 13, pp. 528–543

    2016

  34. [42]

    M*: A complete multirobot path planning algorithm with performance bounds,

    G. Wagner and H. Choset, “M*: A complete multirobot path planning algorithm with performance bounds,” in2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2011, pp. 3260–3267

    2011

  35. [43]

    Structure and intractability of optimal multi- robot path planning on graphs,

    J. Yu and S. M. LaValle, “Structure and intractability of optimal multi- robot path planning on graphs,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 27, no. 1, 2013, pp. 1443–1449

    2013

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.