pith. sign in

arxiv: 2606.06569 · v1 · pith:KJGW5VQ6new · submitted 2026-06-04 · 💻 cs.RO

PhyRoGen: Synthetic Generation of Physical Robot Manipulation Puzzles Using Procedural Content Generation

Lennart Julian Dro{\ss} , Andreas Orthey , Marc Toussaint This is my paper

Pith reviewed 2026-06-28 01:18 UTC · model grok-4.3

classification 💻 cs.RO
keywords procedural content generationrobot manipulationphysical puzzlessynthetic datasetsinterlocking dependenciessampling-based planningKUKA LBR iiwamanipulation benchmarks
0
0 comments X

The pith

PhyRoGen uses procedural content generation to automatically create unique physical robot manipulation puzzles with interlocking object dependencies.

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

Creating large datasets for training robots to manipulate physical puzzles is time-consuming when done manually. The paper introduces PhyRoGen as a framework that applies procedural content generation rules to produce synthetic puzzles where one articulated object must be moved before another can be accessed. Six concrete generators built on this framework produced 24 distinct puzzles. Sampling-based planning algorithms solved every puzzle in simulation within 1 to 300 seconds, and a KUKA LBR iiwa robot arm was shown to manipulate each one successfully in the same simulation environment. This establishes an automated route to generating solvable instances for algorithm benchmarking and model training.

Core claim

PhyRoGen is a general-purpose framework that leverages procedural content generation to produce physical puzzles featuring interlocking object dependencies, where one articulated object must be manipulated before another can be moved. Six concrete generators defined within PhyRoGen yielded 24 puzzles. All 24 puzzles were solved by sampling-based planning algorithms in 1 to 300 seconds, and each was demonstrated to be manipulatable by a KUKA LBR iiwa robot in physical simulation. The framework thereby produces unique, solvable robot manipulation puzzles suitable for benchmarking manipulation algorithms and developing foundation models.

What carries the argument

The PhyRoGen framework, which encodes procedural rules into six concrete generators that create puzzles defined by sequential manipulation requirements arising from physical interlocking dependencies.

If this is right

  • Large collections of distinct, solvable puzzles can be produced without manual design for each instance.
  • Manipulation algorithms gain standardized test cases that vary in their dependency structures.
  • Training datasets for learning-based manipulation methods become available at lower human effort.
  • The generated puzzles support repeated benchmarking runs because each instance is known to be solvable in simulation.

Where Pith is reading between the lines

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

  • The same procedural approach could be extended to generate puzzles with higher numbers of interlocks to stress-test more sophisticated planners.
  • If the simulation-to-real transfer holds, the puzzles could function as standardized physical test artifacts for hardware validation.
  • The method opens a path to procedural generation of related tasks such as automated assembly or disassembly sequences.

Load-bearing premise

The puzzles generated by the six concrete generators have interlocking dependencies that remain physically valid and whose simulated solutions match actions a real robot can perform.

What would settle it

A physical robot executing the planned sequences on the generated puzzles fails to complete the manipulations because of unmodeled contact forces, friction, or dynamics present in the real world but absent from simulation.

Figures

Figures reproduced from arXiv: 2606.06569 by Andreas Orthey, Lennart Julian Dro{\ss}, Marc Toussaint.

Figure 1
Figure 1. Figure 1: A KUKA mobile manipulator solves a procedurally gen￾erated manipulation puzzle with interlocking object dependencies. Left: The grid world environment, where a green slider has to be moved, but is blocked by multiple red objects. Right: The Move N Times environment, where a green cube has to be moved out of a maze which is blocked by a sliding prismatic door, requiring multiple re-grasping sequences. which… view at source ↗
Figure 2
Figure 2. Figure 2: Systems overview about this paper, including the puzzle generation (left), and the puzzle benchmarking and manipulation modules (right). Algorithm 1 PhyRoGen Maker Input: (Objects, Joints, Transforms, N, Seed) Output: Kinematic Chain K 1: K ← INITIALIZEKINEMATICCHAIN() 2: for n ∈ {1, . . . , N} do 3: while True do 4: obj ← GETITEM(Objects, n, Seed) 5: jnt ← GETITEM(Joints, n, Seed) 6: tf ← GETITEM(Transfor… view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of puzzle environments for the first four seeds using the input from Table I. way. Eventually, Frame 4 shows the last action where the robot moves the green object (the only one with a dedicated goal configuration) towards its goal position. The robot used is a KUKA LBR iiwa on a mobile manipulator base, having three degrees of freedom (dof) for the base and 7-dof for the manipulator arm. Ful… view at source ↗
Figure 4
Figure 4. Figure 4: Visualization of benchmark results across six generators, each with four variations, and with the legend below [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Visualization of manipulation tasks across all six environments. The manipulation task for the mobile manipulator robot is to move the green object to its goal position (transparent green), while red objects block the green object and have to be moved out of the way (interlocking dependencies). The Rooms environment is different in that the robot has to move out of the room itself, but is blocked by all ob… view at source ↗
read the original abstract

Robot manipulation of physical puzzles is important for automatic assembly and disassembly tasks. However, to enable robots to solve physical puzzles, manipulation skills need to be learned, which requires large training datasets, the generation of which is often time consuming and tedious. To overcome this problem, we propose the Physical Robot Manipulation Puzzle Generation framework (PhyRoGen), which leverages procedural content generation (PCG) for automated generation of synthetic datasets of manipulation puzzles. PhyRoGen is a general-purpose puzzle generator, which can generate physical puzzles with interlocking object dependencies, where one articulated object must be manipulated before another can be moved. Based upon PhyRoGen, we define six concrete generators which we use to generate 24 physical puzzles. By using a benchmarking framework, we are able to solve all puzzles in 1 to 300 seconds using sampling-based planning algorithms. Finally, we demonstrate that every generated puzzle is manipulatable by using a KUKA LBR iiwa robot in a physical simulation. This shows that our framework is able to procedurally generate unique, solvable robot manipulation puzzles, which is a crucial ingredient to benchmark manipulation algorithms and to develop robust foundation models.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The paper introduces PhyRoGen, a procedural content generation (PCG) framework for automatically synthesizing datasets of physical robot manipulation puzzles that feature interlocking object dependencies (one articulated object must be manipulated before another can move). Six concrete generators are defined and used to produce 24 puzzles; all are reported solvable via sampling-based planning in 1–300 s and demonstrated manipulable by a KUKA LBR iiwa in simulation. The central claim is that this framework supplies unique, solvable puzzles suitable for benchmarking manipulation algorithms and training foundation models.

Significance. If the interlocking dependencies are shown to be physically enforced rather than artifacts of planner constraints or initial conditions, the framework could supply scalable synthetic data for manipulation research. The constructive nature of the generators and the explicit demonstration of simulation solvability are positive features, but the absence of quantitative validation metrics, error analysis, or physical-validity tests limits the immediate impact.

major comments (3)
  1. [Abstract / benchmarking section] Abstract and § on benchmarking/results: solvability of all 24 puzzles in 1–300 s via sampling-based planning is reported, yet no quantitative test (e.g., collision counts, forced-order violation rates, or comparison against a non-interlocking baseline) is supplied to confirm that the claimed interlocking dependencies arise from the physics engine rather than planner heuristics or initial pose constraints.
  2. [Generator definitions] Description of the six concrete generators: the procedural rules are stated to produce “interlocking object dependencies,” but the manuscript provides no explicit mechanism (e.g., joint limits, collision geometry, or stability checks) that would make bypassing an interlock physically impossible inside the simulator; solvability alone does not establish this.
  3. [Simulation experiments] KUKA LBR iiwa simulation demonstration: the claim that “every generated puzzle is manipulatable” is supported only by successful execution; no metrics on grasp success rate, trajectory deviation, or failure modes under perturbed initial conditions are given, leaving open whether the generated puzzles test sequential manipulation or merely planner reachability.
minor comments (2)
  1. [Abstract] The abstract states “24 physical puzzles” but the text does not clarify whether these are distinct up to rigid transformation or merely 24 instances; a table enumerating generator-to-puzzle mapping would improve clarity.
  2. [Methods] No mention of the underlying physics engine parameters (friction, contact solver tolerance) or whether the same parameters are used for both generation and planning validation.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive feedback. The comments correctly identify areas where additional evidence would strengthen the paper's claims regarding the physical enforcement of interlocking dependencies. We address each point below and commit to revisions that incorporate the suggested validations.

read point-by-point responses

  1. Referee: [Abstract / benchmarking section] Abstract and § on benchmarking/results: solvability of all 24 puzzles in 1–300 s via sampling-based planning is reported, yet no quantitative test (e.g., collision counts, forced-order violation rates, or comparison against a non-interlocking baseline) is supplied to confirm that the claimed interlocking dependencies arise from the physics engine rather than planner heuristics or initial pose constraints.

    Authors: We agree with this observation. The current results focus on solvability times but do not include quantitative tests to isolate the effect of physical interlocks. In the revised manuscript, we will add quantitative metrics including collision counts during planning, rates of forced-order violations, and comparisons to non-interlocking baseline puzzles. This will be presented in an expanded benchmarking section to demonstrate that the dependencies are enforced by the physics engine. revision: yes

  2. Referee: [Generator definitions] Description of the six concrete generators: the procedural rules are stated to produce “interlocking object dependencies,” but the manuscript provides no explicit mechanism (e.g., joint limits, collision geometry, or stability checks) that would make bypassing an interlock physically impossible inside the simulator; solvability alone does not establish this.

    Authors: The manuscript outlines the procedural rules for generating puzzles with interlocking dependencies but does not provide detailed descriptions of the simulator-level mechanisms. We will revise the section on generator definitions to explicitly specify the physical mechanisms, including joint limits, collision geometries, and stability checks that prevent bypassing the interlocks. This will clarify how the dependencies are physically enforced. revision: yes

  3. Referee: [Simulation experiments] KUKA LBR iiwa simulation demonstration: the claim that “every generated puzzle is manipulatable” is supported only by successful execution; no metrics on grasp success rate, trajectory deviation, or failure modes under perturbed initial conditions are given, leaving open whether the generated puzzles test sequential manipulation or merely planner reachability.

    Authors: We recognize that the simulation results are presented only as successful demonstrations without supporting metrics. To address this, the revised version will include quantitative metrics such as grasp success rates over multiple trials, average trajectory deviations, and performance under perturbed initial conditions. This will better illustrate that the puzzles challenge sequential manipulation capabilities. revision: yes

Circularity Check

0 steps flagged

No circularity: constructive PCG framework with direct simulation demonstration

full rationale

The paper describes a procedural content generation framework (PhyRoGen) that defines six concrete generators to produce 24 puzzles, then reports that sampling-based planners solve all of them in simulation and that a KUKA LBR iiwa model can manipulate them. No equations, parameter fitting, or derivation chain exists. The central claim is that the generators produce solvable interlocking puzzles; this is shown by explicit construction and planner output rather than by reducing any quantity to a fitted input or self-citation. The work is self-contained as an engineering artifact whose outputs are directly exhibited.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that procedurally generated puzzles with stated interlocking dependencies are physically realizable and solvable; no free parameters or invented entities are described in the abstract.

axioms (1)
  • domain assumption Procedural generators can produce puzzles whose interlocking dependencies require sequential manipulation and remain physically valid.
    Invoked when claiming the generated puzzles are solvable and manipulatable (abstract, final sentences).

pith-pipeline@v0.9.1-grok · 5737 in / 1140 out tokens · 22454 ms · 2026-06-28T01:18:50.056980+00:00 · methodology

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

59 extracted references

  1. [1]

    Motionbenchmaker: A tool to generate and benchmark motion planning datasets,

    C. Chamzas, C. Quintero-Peña, Z. Kingston, A. Orthey, D. Rakita, M. Gleicher, M. Toussaint, and L. E. Kavraki, “Motionbenchmaker: A tool to generate and benchmark motion planning datasets,”IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 882–889, apr 2022

    2022

  2. [2]

    Benchmarking motion planning algorithms: An extensible infrastructure for analysis and visualization,

    M. Moll, I. A. Sucan, and L. E. Kavraki, “Benchmarking motion planning algorithms: An extensible infrastructure for analysis and visualization,”IEEE Robotics and Automation Magazine, vol. 22, no. 3, pp. 96–102, 2015

    2015

  3. [3]

    A review of robot learning for manipulation: Challenges, representations, and algorithms,

    O. Kroemer, S. Niekum, and G. Konidaris, “A review of robot learning for manipulation: Challenges, representations, and algorithms,”Jour- nal of Machine Learning Research, vol. 22, no. 30, pp. 1–82, 2021

    2021

  4. [4]

    Foundation models in robotics: Applications, challenges, and the future,

    R. Firoozi, J. Tucker, S. Tianet al., “Foundation models in robotics: Applications, challenges, and the future,”International Journal of Robotics Research, vol. 44, no. 5, pp. 701–739, 2025

    2025

  5. [5]

    Generative artificial intelligence in robotic manipulation: A survey,

    K. Zhang, P. Yun, J. Cenet al., “Generative artificial intelligence in robotic manipulation: A survey,”CoRR, vol. abs/2503.03464, 2025

    arXiv 2025

  6. [6]

    Procthor: Large-scale embodied ai using procedural generation,

    M. Deitke, E. VanderBilt, A. Herrastiet al., “Procthor: Large-scale embodied ai using procedural generation,”Advances in Neural Infor- mation Processing Systems, vol. 35, pp. 5982–5994, 2022

    2022

  7. [7]

    Opening a lockbox through physical exploration,

    M. Baum, M. Bernstein, R. Martin-Martinet al., “Opening a lockbox through physical exploration,” inIEEE International Conference on Humanoid Robots. IEEE, 2017, pp. 461–467

    2017

  8. [8]

    Solving rearrangement puzzles using path defragmentation in factored state spaces,

    S. B. Bayraktar, A. Orthey, Z. Kingston, M. Toussaint, and L. E. Kavraki, “Solving rearrangement puzzles using path defragmentation in factored state spaces,”IEEE Robotics and Automation Letters, vol. 8, no. 8, pp. 4529–4536, 2023

    2023

  9. [9]

    Manipula- tion planning with probabilistic roadmaps,

    T. Siméon, J.-P. Laumond, J. Cortés, and A. Sahbani, “Manipula- tion planning with probabilistic roadmaps,”International Journal of Robotics Research, vol. 23, no. 7–8, pp. 729–746, 2004

    2004

  10. [10]

    Blender,

    B. O. Community, “Blender,” Blender Foundation, Stichting Blender Foundation, Amsterdam, 2021

    2021

  11. [11]

    Phobos: A tool for creating complex robot models,

    K. von Szadkowski and S. Reichel, “Phobos: A tool for creating complex robot models,”Journal of Open Source Software, vol. 5, no. 45, p. 1326, 2020

    2020

  12. [12]

    Robowflex: Robot motion planning with moveit made easy,

    Z. Kingston and L. E. Kavraki, “Robowflex: Robot motion planning with moveit made easy,” inIEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2022, pp. 3108–3114

    2022

  13. [13]

    Dart: Dynamic animation and robotics toolkit,

    J. Lee, M. X. Grey, S. Haet al., “Dart: Dynamic animation and robotics toolkit,”Journal of Open Source Software, vol. 3, no. 22, p. 500, 2018

    2018

  14. [14]

    The open motion planning library,

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

    2012

  15. [15]

    Pybullet, a python module for physics simulation for games, robotics and machine learning,

    E. Coumans and Y . Bai, “Pybullet, a python module for physics simulation for games, robotics and machine learning,” 2021

    2021

  16. [16]

    What is procedural content generation? mario on the borderline,

    J. Togelius, E. Kastbjerg, D. Schedl, and G. N. Yannakakis, “What is procedural content generation? mario on the borderline,” inInterna- tional Workshop on Procedural Content Generation in Games, 2011, pp. 1–6

    2011

  17. [17]

    Shaker, J

    N. Shaker, J. Togelius, and M. J. Nelson,Procedural Content Gener- ation in Games: A Textbook and an Overview of Current Research. Springer, 2016

    2016

  18. [18]

    Harris,Exploring Roguelike Games

    J. Harris,Exploring Roguelike Games. CRC Press, 2020

    2020

  19. [19]

    Procedural dungeon generation: A survey,

    B. M. F. Viana and S. R. dos Santos, “Procedural dungeon generation: A survey,”Journal on Interactive Systems, vol. 12, no. 1, pp. 83–101, 2021

    2021

  20. [20]

    Procedural content generation for games: A survey,

    M. Hendrikx, S. Meijer, J. Van Der Velden, and A. Iosup, “Procedural content generation for games: A survey,”ACM Transactions on Multimedia Computing, Communications, and Applications, vol. 9, no. 1, pp. 1–22, 2013

    2013

  21. [21]

    Comparing pcg metrics with human evaluation in minecraft settlement generation,

    J.-B. Hervé and C. Salge, “Comparing pcg metrics with human evaluation in minecraft settlement generation,” inProceedings of the International Conference on the Foundations of Digital Games, 2021, pp. 1–15

    2021

  22. [22]

    Wor(l)d-gan: To- ward natural-language-based pcg in minecraft,

    M. Awiszus, F. Schubert, and B. Rosenhahn, “Wor(l)d-gan: To- ward natural-language-based pcg in minecraft,”IEEE Transactions on Games, vol. 15, no. 2, pp. 182–192, 2022

    2022

  23. [23]

    Procedural puzzle generation: A survey,

    B. De Kegel and M. Haahr, “Procedural puzzle generation: A survey,” IEEE Transactions on Games, vol. 12, no. 1, pp. 21–40, 2019

    2019

  24. [24]

    Procedural level generation for sokoban via deep learning: An experimental study,

    Y . Zakaria, M. Fayek, and M. Hadhoud, “Procedural level generation for sokoban via deep learning: An experimental study,”IEEE Trans- actions on Games, vol. 15, no. 1, pp. 108–120, 2022

    2022

  25. [25]

    Sokoban is pspace-complete,

    J. Culberson, “Sokoban is pspace-complete,”IEICE Technical Report, 1997

    1997

  26. [26]

    Data-driven sokoban puzzle generation with monte carlo tree search,

    B. Kartal, N. Sohre, and S. J. Guy, “Data-driven sokoban puzzle generation with monte carlo tree search,” inProceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, vol. 12, 2016, pp. 58–64

    2016

  27. [27]

    Taylor,The Procedural Generation of Interesting Sokoban Levels

    J. Taylor,The Procedural Generation of Interesting Sokoban Levels. University of North Texas, 2015

    2015

  28. [28]

    Generating sokoban levels that are interesting to play using simulation,

    S. Karmanet al., “Generating sokoban levels that are interesting to play using simulation,” Master’s thesis, Utrecht University, 2018

    2018

  29. [29]

    R. S. Sutton and A. G. Barto,Reinforcement Learning: An Introduc- tion. MIT Press, 2018

    2018

  30. [30]

    The arcade learning environment: An evaluation platform for general agents,

    M. G. Bellemare, Y . Naddaf, J. Veness, and M. Bowling, “The arcade learning environment: An evaluation platform for general agents,” Journal of Artificial Intelligence Research, vol. 47, pp. 253–279, 2013

    2013

  31. [31]

    Human-level control through deep reinforcement learning,

    V . Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski et al., “Human-level control through deep reinforcement learning,” Nature, vol. 518, no. 7540, pp. 529–533, 2015

    2015

  32. [32]

    Grandmaster level in starcraft ii using multi-agent reinforcement learning,

    O. Vinyals, I. Babuschkin, W. M. Czarnecki, M. Mathieu, A. Dudzik, J. Chung, D. H. Choi, R. Powell, T. Ewalds, P. Georgievet al., “Grandmaster level in starcraft ii using multi-agent reinforcement learning,”Nature, vol. 575, no. 7782, pp. 350–354, 2019

    2019

  33. [33]

    Leveraging procedu- ral generation to benchmark reinforcement learning,

    K. Cobbe, C. Hesse, J. Hilton, and J. Schulman, “Leveraging procedu- ral generation to benchmark reinforcement learning,” inProceedings of the International Conference on Machine Learning. PMLR, 2020, pp. 2048–2056

    2020

  34. [34]

    Minihack the planet: A sandbox for open-ended reinforcement learning research,

    M. Samvelyan, R. Kirk, V . Kurinet al., “Minihack the planet: A sandbox for open-ended reinforcement learning research,” inNeural Information Processing Systems, 2021

    2021

  35. [35]

    Robot learning in the era of foundation models: A survey,

    X. Xiao, J. Liu, Z. Wanget al., “Robot learning in the era of foundation models: A survey,”Neurocomputing, p. 129963, 2025

    2025

  36. [36]

    Reinforcement learning in robotics: A survey,

    J. Kober, J. A. Bagnell, and J. Peters, “Reinforcement learning in robotics: A survey,”International Journal of Robotics Research, vol. 32, no. 11, pp. 1238–1274, 2013

    2013

  37. [37]

    Solving rubik’s cube with a robot hand,

    OpenAI, I. Akkaya, M. Andrychowiczet al., “Solving rubik’s cube with a robot hand,”ArXiv Preprint, 2019

    2019

  38. [38]

    Cluttergen: A cluttered scene generator for robot learning,

    Y . Jia and B. Chen, “Cluttergen: A cluttered scene generator for robot learning,” inProceedings of the Conference on Robot Learning, 2024

    2024

  39. [39]

    Meta-world: A benchmark and evalu- ation for multi-task and meta reinforcement learning,

    T. Yu, D. Quillen, Z. Heet al., “Meta-world: A benchmark and evalu- ation for multi-task and meta reinforcement learning,” inProceedings of the Conference on Robot Learning. PMLR, 2020, pp. 1094–1100

    2020

  40. [40]

    Rlbench: The robot learning benchmark and learning environment,

    S. James, Z. Ma, D. R. Arrojo, and A. J. Davison, “Rlbench: The robot learning benchmark and learning environment,”IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 3019–3026, 2020

    2020

  41. [41]

    Hierarchical task and motion planning in the now,

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

    2011

  42. [42]

    Integrated task and mo- tion planning,

    C. R. Garrett, R. Chitnis, R. Holladayet al., “Integrated task and mo- tion planning,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, no. 1, pp. 265–293, 2021

    2021

  43. [43]

    PaLM-e: An embodied multimodal language model,

    D. Driess, F. Xia, M. S. M. Sajjadiet al., “PaLM-e: An embodied multimodal language model,” inProceedings of the International Con- ference on Machine Learning, ser. Proceedings of Machine Learning Research, A. e. a. Krause, Ed., vol. 202. PMLR, 23–29 Jul 2023, pp. 8469–8488

    2023

  44. [44]

    Differentiable physics and stable modes for tool-use and manipulation planning,

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

    2018

  45. [45]

    Exploring implicit spaces for constrained sampling-based planning,

    Z. Kingston, M. Moll, and L. E. Kavraki, “Exploring implicit spaces for constrained sampling-based planning,”International Journal of Robotics Research, vol. 38, no. 10-11, pp. 1151–1178, 2019

    2019

  46. [46]

    Fast high-quality tabletop rearrangement in bounded workspace,

    K. Gao, D. Lau, B. Huang, K. E. Bekris, and J. Yu, “Fast high-quality tabletop rearrangement in bounded workspace,” inIEEE International Conference on Robotics and Automation. IEEE, 2022, pp. 1961–1967

    2022

  47. [47]

    Re-assembling the past: The repair dataset and benchmark for real world 2d and 3d puzzle solving,

    T. Tsesmelis, L. Palmieri, M. Khoroshiltsevaet al., “Re-assembling the past: The repair dataset and benchmark for real world 2d and 3d puzzle solving,”Advances in Neural Information Processing Systems, vol. 37, pp. 30 076–30 105, 2024

    2024

  48. [48]

    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,” inIEEE International Conference on Robotics and Automation, vol. 2. IEEE, 2000, pp. 995–1001

    2000

  49. [49]

    Conlan,The Blender Python API: Precision 3D Modeling and Add- on Development

    C. Conlan,The Blender Python API: Precision 3D Modeling and Add- on Development. Apress, 2017

    2017

  50. [50]

    Sampling-based motion planning: A comparative review,

    A. Orthey, C. Chamzas, and L. E. Kavraki, “Sampling-based motion planning: A comparative review,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 7, no. 1, pp. 285–310, 2024

    2024

  51. [51]

    J. J. Craig,Introduction to Robotics: Mechanics and Control, 4th ed. Upper Saddle River, NJ: Pearson, 2017

    2017

  52. [52]

    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,”Inter- national Journal of Robotics Research, vol. 39, no. 5, pp. 543–567, 2020

    2020

  53. [53]

    Sampling-based algorithms for optimal motion planning,

    S. Karaman and E. Frazzoli, “Sampling-based algorithms for optimal motion planning,”International Journal of Robotics Research, vol. 30, no. 7, pp. 846–894, 2011

    2011

  54. [54]

    Asymptotically near-optimal rrt for fast, high-quality motion planning,

    O. Salzman and D. Halperin, “Asymptotically near-optimal rrt for fast, high-quality motion planning,”IEEE Transactions on Robotics, vol. 32, no. 3, pp. 473–483, 2016

    2016

  55. [55]

    Asymptotically optimal sampling- based motion planning methods,

    J. D. Gammell and M. P. Strub, “Asymptotically optimal sampling- based motion planning methods,”Annual Review of Control, Robotics, and Autonomous Systems, vol. 4, no. 1, pp. 295–318, 2021

    2021

  56. [56]

    A survey of optimization-based task and motion planning: From classical to learning approaches,

    Z. Zhao, S. Cheng, Y . Dinget al., “A survey of optimization-based task and motion planning: From classical to learning approaches,” IEEE/ASME Transactions on Mechatronics, 2024

    2024

  57. [57]

    Section patterns: Efficiently solving narrow passage problems in multilevel motion planning,

    A. Orthey and M. Toussaint, “Section patterns: Efficiently solving narrow passage problems in multilevel motion planning,”IEEE Trans- actions on Robotics, vol. 37, pp. 1891–1905, 2021

    1905

  58. [58]

    Bitkomo: Combining sampling and optimization for fast convergence in optimal motion planning,

    J. Kamat, J. Ortiz-Haro, M. Toussaint, F. T. Pokorny, and A. Orthey, “Bitkomo: Combining sampling and optimization for fast convergence in optimal motion planning,” inInternational Conference on Intelligent Robots and Systems, 2022, pp. 4492–4497

    2022

  59. [59]

    Ad- versarial reinforcement learning for procedural content generation,

    L. Gisslén, A. Eakins, C. Gordillo, J. Bergdahl, and K. Tollmar, “Ad- versarial reinforcement learning for procedural content generation,” in Proceedings of the IEEE Conference on Games. IEEE, 2021, pp. 1–8

    2021