Master Rules from Chaos: Learning to Reason, Plan, and Interact from Chaos for Tangram Assembly

Chao Zhao; Chunli Jiang; Guanlan Zhang; Hongyu Yu; Lifan Luo; Michael Yu Wang; Qifeng Chen

REVIEW 3 major objections 6 minor 27 references

Master Rules from Chaos: Learning to Reason, Plan, and Interact from Chaos for Tangram Assembly

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read A robot policy trained only on random tangram shapes can assemble novel, denser silhouettes it has never seen, using a single image prompt per target.

desk verdict New task and a clean self-supervised training scheme, but the central generalization claim is undercut by a permissive coverage metric and missing statistics. read the letter →

arxiv 2505.11818 v1 pith:6L6XU5EM submitted 2025-05-17 cs.RO

Chao Zhao , Chunli Jiang , Lifan Luo , Guanlan Zhang , Hongyu Yu , Michael Yu Wang , Qifeng Chen This is my paper · ORCID

classification cs.RO

keywords tangramassemblyroboticreinforcementlearninggeneralizationself-supervisedsilhouettepromptsim-to-realtransferrandomobjectgeneration

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

MRChaos claims that a robot can learn general assembly skills by practicing only on randomly generated tangram objects, with no demonstrations, no pre-modeled target objects, and no hand-written assembly rules. Trained in simulation with a reward that simply measures how many pixels of each placed piece fall inside the target silhouette, the policy transfers to human-created tangram silhouettes it has never seen, using only the silhouette image as a prompt. If true, this means the complexity of object assembly can be tamed by learning in a much simpler random domain, avoiding brittle modeling and costly real-world trial-and-error.

What carries the argument

The load-bearing mechanism is the pairing of a random-object training distribution with a visual-coverage reward. Random object generation constructs a fresh target each episode by placing the seven pieces randomly with no stacking, optionally then attracting them with a gravitational force at one point to create denser, more bonded configurations; no semantics and no human assembly rules are injected. The reward r_t is the ratio of overlapping pixels between the placed piece and its corresponding region in the silhouette, so the agent is pushed purely by visual change. A two-stage curriculum — first placing one piece with a few pieces pre-assembled, then assembling all pieces from empty — carries the argument that local placement competency plus global error-correcting feedback is what makes the policy robust to novel dense silhouettes.

What would settle it

Train the policy on randomly generated objects deliberately restricted to low-aggregation configurations, for example by removing the gravitational attractor and capping contact between pieces, and then test on the H-Fiendish family; if final coverage remains near 73%, the generalization does not depend on the training distribution covering dense configurations. Alternatively, compute silhouette perimeter and contact-graph statistics for the random training set and the 104 human-created objects; if the H-Fiendish family lies outside the convex hull of the training statistics, the transfer claim would need another explanation.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a single policy, trained entirely by self-exploration on randomly assembled tangram objects in simulation, can assemble novel human-created tangram silhouettes at test time. The robot sequentially picks the seven pieces in fixed size order and, at each step, the policy outputs a gripper displacement from the current depth image and the silhouette prompt; the reward is the pixel-overlap between the placed piece and its region in the target silhouette. With proximal policy optimization and a two-stage curriculum — first learning local placement on partially assembled objects, then full assembly — MRChaos achieves an 80.3% final coverage on the H-Normal family and 73.6% on the H-Fiendish family in simulation, and 62.4% on real-world H-Hard objects, while a behavior-cloning baseline trained on the 104 human-created objects reaches only around 60-62% final coverage and fails after the first few pieces. The same approach, retrained with cutlery and soda cans, reaches 89.7% and 80.6% final coverage on new goal configurations.

Load-bearing premise

The random object generator used for training must produce the same range of structural complexity and piece aggregation as the human-created test silhouettes; if that coverage is missing, the reported generalization is incidental similarity rather than learned assembly rules.

Editorial extensions

If this is right

A robot can assemble previously unseen tangram silhouettes from a single image prompt without target-specific modeling or demonstrations.
Training on simple random objects is enough to transfer to denser, semantically meaningful human-created assemblies, contradicting the idea that training complexity must match task complexity.
The visual-difference reward removes the need for manual reward engineering or task-specific annotations.
The approach extends to other multi-piece planar arrangements, such as cutlery and soda combinations, with only substituted object meshes and prompts.
Global assembly learning, the second curriculum stage, is critical for dense objects; omitting it drops final coverage by up to 10 percentage points on the H-Fiendish family.

Reading between the lines

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

A direct test of the paper's mechanism would be to measure the distributional overlap between randomly generated training silhouettes and the human-created test families; if the random generator's contact graphs and perimeter statistics already cover the H-Fiendish region, the 'generalization' may be interpolation rather than rule discovery.
The fixed piece-ordering assumption leaves open whether the learned policy would generalize to arbitrary assembly orders; allowing the policy to choose the next piece would test whether the reasoning is compositional.
The same self-supervised recipe may transfer to other planar assembly problems with silhouette-like goal prompts, such as arranging irregular pantry items, provided the random generator can be made to produce sufficiently varied goal configurations.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Colleague,

The thing you should know: this is a genuinely new task formulation and an honest attempt at a hard problem, but the headline 'radical generalization' rests on a metric that looks too forgiving. The Rela vs Final gap on H-Fiendish (34.6% vs 73.6%) tells me the policy is filling the silhouette outline without consistently placing pieces in their correct positions. That doesn't kill the paper, but it means the central claim needs a sharper metric before I'd trust it.

What's good: tangram assembly is a nice benchmark for robotic assembly; training only on randomly generated objects and testing on held-out human-created ones is a real generalization test. The coverage reward from visual observation is clean and requires no annotations. The two-stage curriculum makes sense, and the ablations directionally support it. The real-world demos show sim-to-real transfer is feasible. The code availability promise helps.

Soft spots, in order of severity. First, the metric. Final coverage is a pixel-overlap rate that doesn't penalize protruding or overlapping pieces. With seven pieces, you could score high by scattering them inside the silhouette. The large gap between Rela and Final on the human-created families is consistent with that. The paper needs a per-piece precision metric, something like IoU per piece or a penalty for excess area. Second, the numbers are single points with no standard deviations, trial counts, or per-object breakdowns. The ablation deltas are small—up to 10%—so without variance I can't tell if they're real. Third, the BC baseline is trained on disassembling the same 104 human-created objects, which is a different distribution from the RL training data; a fairer baseline would be BC on the random objects or another RL method. Fourth, the real-world evaluation is under-specified: how many trials per family? What's the success criterion? Table II reports only final coverage, inheriting the same metric problem. Finally, the assumption that random objects cover the space of human-created difficulties is plausible but not measured; a simple distributional analysis (e.g., silhouette perimeter/entropy) would help.

Bottom line: the core idea is worth testing, and the task itself is a useful benchmark. But as written, the paper overclaims. I'd send it to peer review, not desk-reject, with a request for major revisions: fix the metric, add proper statistics, strengthen the baseline, and give the real-world protocol. If the authors do that, the generalization claim could stand. For me personally, it's a maybe for the reading group; I'd cite it as a task formulation, but not yet as evidence for radical generalization.

Referee Report

3 major / 6 minor

Summary. The paper introduces MRChaos, a reinforcement-learning pipeline for robotic tangram assembly from silhouette prompts. A policy is trained in simulation with PPO on randomly generated tangram objects, receiving a reward based on the pixel-wise coverage of each placed piece against the target silhouette; a two-stage curriculum first learns single-piece placement and then full assembly. The resulting policy is evaluated in simulation on randomly generated objects and on three families of human-created objects unseen during training (H-Normal, H-Hard, H-Fiendish), and transferred to a real robot without retraining. The authors report relative and final coverage rates, compare against behavior cloning and ablated versions, and demonstrate extensions to cutlery and soda-can combination tasks. The central claim is that a policy trained only on simple random objects can generalize to novel objects with only silhouette prompts, indicating that radical generalization can emerge from learning in simpler domains.

Significance. The paper addresses a genuinely hard and underexplored problem: assembly of novel objects from an abstract silhouette prompt, without demonstrations, object-specific models, or hand-coded assembly rules. The experimental design has real strengths: the training distribution is randomized, the test objects are human-created and held out from training, the reward is derived purely from visual differences, and the real-world transfer section provides a concrete instantiation of the claim. If the generalization results are as strong as reported, this would be a meaningful step toward flexible robotic assembly. At the same time, the quantitative evidence for the central claim is currently weakened by the choice of metrics and the absence of statistical detail; the paper's own numbers suggest that high final coverage can coexist with low per-piece placement accuracy, which leaves the 'assembly' claim under-supported.

major comments (3)

[Sec. V-B, Table I] The divergence between Rela and Final on the novel-object families undermines the central generalization claim as currently quantified. On H-Fiendish, MRChaos achieves Rela=34.6% but Final=73.6%, a gap of 39 percentage points. Because Final measures only the fraction of silhouette pixels covered by the union of placed pieces, it does not penalize pieces that protrude outside the silhouette, overlap one another, or sit in the wrong part of the outline. The per-piece reward of Eq. (1) is likewise pure coverage with no precision or overlap penalty. A policy that merely distributes the seven pieces inside the prompt silhouette can therefore accumulate a high Final score while failing to reproduce the target configuration, exactly the pattern visible in Table I. The paper should report a precision term (e.g., the fraction of the union of placed pieces that lies inside the silhouette), an overlap/duplication penalty, and per-object success rates based on a tolerance on final configuration error. Without such metrics, the claim that MRChaos assembles novel objects 'robustly' is not established; the real-world results in Table II, which report only Final, inherit the same ambiguity.
[Sec. V, Tables I and II] All reported results are single point estimates with no standard deviations, no number of evaluation episodes per object, no number of objects per family, and no information about how many random seeds were used during training. Several head-to-head comparisons are small (e.g., 80.3% vs. 79.4% Final on H-Normal for MRChaos vs. MRChaos w/o global), and without variance or per-object counts these differences cannot be assessed. The central quantitative claims of the paper depend on these comparisons, so the evaluation should be reported with means, standard deviations, and trial counts, or with per-object scatter plots. This is a load-bearing reporting issue, not a stylistic one.
[Sec. IV-D, Eq. (1)] The reward is defined solely as per-piece coverage of the target region, with no penalty for placing a piece partially or fully outside its target region as long as some overlap is achieved, and no penalty for overlapping previously placed pieces. This reward design is consistent with the observed Rela/Final gap: the policy can maximize reward by covering silhouette area with individually misplaced pieces. The paper should either add a precision or non-overlap term to the reward, or explicitly analyze the extent to which the learned policy exploits this gap. The current formulation leaves open the possibility that the reported generalization is an artifact of the metric rather than of true assembly competence.

minor comments (6)

[Sec. IV-A and Sec. V-A] The random object generation described in Sec. IV-A is the sole training distribution, but the paper does not measure how well this distribution covers the feature space of the human-created test families. Reporting a quantitative coverage analysis (e.g., silhouette perimeter, piece adjacency statistics, or goal-conditioned observation distances) would clarify whether the reported generalization reflects learned assembly rules or incidental similarity to the training distribution.
[Sec. IV-E] The curriculum description states that in the first stage 'a predetermined number of pieces, randomly ranging from 0 to j-1, are pre-assembled.' If zero pieces are pre-assembled, the first stage is identical to the full task; please clarify whether the range is 1 to j-1 or whether the zero case is intentionally included.
[Sec. IV-B and Sec. IV-C] The action space is described only as a gripper displacement (x_t, y_t, theta_t) with no bounds, discretization, or coordinate frame specification. The observation resolution and camera placement are also unspecified beyond a 120x120 image. Adding these details is necessary for reproducibility.
[Sec. V-C, Table II] The real-world task families H-Simple, H-Middle, and H-Hard are not defined or related to the simulation families H-Normal, H-Hard, and H-Fiendish. Please clarify the correspondence, the number of real-world objects per family, and the number of trials per object.
[Fig. 3] The caption says the second curriculum stage begins at the orange line, but the curve is rendered in grayscale and no orange line is visible. Please make the stage boundary explicit and legible.
[Sec. V-A] The paper states that 104 human-created tangram objects are used but does not state how many fall into each of H-Normal, H-Hard, and H-Fiendish. This distribution is relevant for interpreting the aggregate scores in Table I.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: novelty claim is supported by human-created unseen test objects.

full rationale

The paper's central claim is generalization to novel, human-created tangram objects (H-Normal, H-Hard, H-Fiendish), while training uses randomly generated objects (Sec. IV-A). The test families are explicitly declared unseen in Sec. V-A: "objects in H-Normal, H-Hard, and H-Fiendish are previously unseen." The reward in Eq. (1) is a per-piece overlap objective used for RL training; it is not a fitted parameter and is applied to unseen test objects without being tuned to them. The relative coverage metric is described as "similar to the reward calculation" (Sec. V-A), but the policy output is a gripper displacement learned from observations, not constructed from the test metric. The only distribution overlap is the Random family, which the paper explicitly labels as "similar to training objects" rather than using it to argue for novel-object generalization. There are no load-bearing self-citations, no imported uniqueness theorems, and no ansatz smuggled in via citation. Any concerns about the strictness of the coverage metric are validity/correctness issues, not circularity.

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

The central claim rests on a set of design choices and domain assumptions: the reward is a pixel-overlap coverage signal, the random target distribution is representative of novel objects, the observation is Markovian, and the simulator transfers to the real world. None of these are independently established beyond the paper's own experiments. No new physical entities are introduced.

free parameters (5)

PPO learning rate = 2e-4
Chosen in Sec. IV-E; a standard value but hand-selected and not swept, affecting training stability and final policy quality.
MLP hidden units = 512
Hidden layer size in Sec. IV-C; chosen by the authors without reported sensitivity analysis.
Observation resolution = 120x120
Image size for Is and Itc in Sec. IV-B and IV-C; a hand-picked resolution for the Conv encoder.
Curriculum stage switch criterion = when rewards of the first stage cease to increase
Sec. IV-E; an ad hoc trigger not precisely quantified, affecting which skills the policy learns in stage two.
Number of parallel environments = 256
Sec. IV-E; a computational choice made by the authors that influences data diversity and training throughput.

assumptions (5)

domain assumption The tangram assembly task can be modeled as a discrete-time MDP whose state is fully captured by the current depth observation and silhouette prompt, with no memory of past steps.
Sec. III-B formulates the problem as an MDP and Sec. IV-B defines the observation ot = (Is, Itc) without temporal stacking or an explicit step index, assuming this observation carries enough information for placing the next piece.
domain assumption Randomly generated tangram objects, created by random placement with or without a gravitational bonding step, span the structure space needed to learn assembly of human-created complex objects.
Sec. IV-A introduces the random generation, and the generalization claim depends on this distribution covering sufficient structural complexity. The H-Fiendish family in Sec. V-A has denser aggregation, so this assumption is only partially validated.
domain assumption The 2D silhouette coverage reward rt = |pt intersect p't| / |p't| is a sufficient learning signal for correct physical assembly.
Sec. IV-D defines the reward from pixel overlap only; it does not explicitly reward physical stability or contact correctness, yet successful assembly in the real world requires both.
domain assumption Pybullet simulation is a faithful proxy for real-world physics for suction-based planar placement, and the policy transfers without explicit domain randomization.
Training is conducted in Pybullet (Sec. IV-A), and Sec. V-C deploys the same policy directly on a real robot. Sim-to-real fidelity is implicitly assumed rather than measured.
standard math Standard mathematical and algorithmic background, including PPO, Adam, and neural network approximation, is assumed.
Used throughout Secs. IV and V without proof; standard material in the reinforcement learning literature.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Master Rules from Chaos: Learning to Reason, Plan, and Interact from Chaos for Tangram Assembly." pith.science (2026). https://pith.science/paper/6L6XU5EM

@misc{pith2026250511818,
  author       = {Pith},
  title        = {Pith review of: Master Rules from Chaos: Learning to Reason, Plan, and Interact from Chaos for Tangram Assembly},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6L6XU5EM}},
  note         = {Machine review of arXiv:2505.11818}
}

read the original abstract

Tangram assembly, the art of human intelligence and manipulation dexterity, is a new challenge for robotics and reveals the limitations of state-of-the-arts. Here, we describe our initial exploration and highlight key problems in reasoning, planning, and manipulation for robotic tangram assembly. We present MRChaos (Master Rules from Chaos), a robust and general solution for learning assembly policies that can generalize to novel objects. In contrast to conventional methods based on prior geometric and kinematic models, MRChaos learns to assemble randomly generated objects through self-exploration in simulation without prior experience in assembling target objects. The reward signal is obtained from the visual observation change without manually designed models or annotations. MRChaos retains its robustness in assembling various novel tangram objects that have never been encountered during training, with only silhouette prompts. We show the potential of MRChaos in wider applications such as cutlery combinations. The presented work indicates that radical generalization in robotic assembly can be achieved by learning in much simpler domains.

Figures

Figures reproduced from arXiv: 2505.11818 by the authors.

**Figure 2.** System Overview. A: During training, a random target object, along with a silhouette image Is , is generated. At time step t, the Itc is captured from a top-down camera and concatenated with Is as ot . The agent receives the ot and outputs the action at for the robot to place the nt piece. The agent finally receives rewards rt according to the visual difference and updates the policy with PPO; B: Deployment of MRCha… view at source ↗

**Figure 3.** (a) Silhouette examples of randomly generated objects. The top two [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: Qualitative results of tangram assembly. A: Sequences show MRChaos assembling objects with silhouette prompts. From top to bottom, the objects belong to the H-Normal, H-Hard, and H-Fiendish task families, respectively. B: A sequence shows the robot assembling objects t…

**Figure 6.** Figure 6: (a) Silhouettes and corresponding goal configurations of cutlery [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 24 canonical work pages

[1]

Learning insertion primitives with discrete-continuous hybrid action space for robotic assembly tasks,

X. Zhang, S. Jin, C. Wang, X. Zhu, and M. Tomizuka, “Learning insertion primitives with discrete-continuous hybrid action space for robotic assembly tasks,” in 2022 International Conference on Robotics and Automation (ICRA) , pp. 9881–9887, IEEE, 2022

work page 2022
[2]

Form2fit: Learning shape priors for generalizable assembly from disassembly,

K. Zakka, A. Zeng, J. Lee, and S. Song, “Form2fit: Learning shape priors for generalizable assembly from disassembly,” in 2020 IEEE International Conference on Robotics and Automation (ICRA) , pp. 9404– 9410, IEEE, 2020

work page 2020
[3]

Factory: Fast contact for robotic assembly,

Y . Narang, K. Storey, I. Akinola, M. Macklin, P. Reist, L. Wawrzyniak, Y . Guo, A. Moravanszky, G. State, M. Lu, et al., “Factory: Fast contact for robotic assembly,” arXiv preprint arXiv:2205.03532 , 2022

arXiv 2022
[4]

Compliant peg-in-hole assembly using partial spiral force trajectory with tilted peg posture,

H. Park, J. Park, D.-H. Lee, J.-H. Park, and J.-H. Bae, “Compliant peg-in-hole assembly using partial spiral force trajectory with tilted peg posture,” IEEE Robotics and Automation Letters , vol. 5, no. 3, pp. 4447–4454, 2020

work page 2020
[5]

Planning grasps with suction cups and parallel grippers using superimposed segmentation of object meshes,

W. Wan, K. Harada, and F. Kanehiro, “Planning grasps with suction cups and parallel grippers using superimposed segmentation of object meshes,” IEEE Transactions on Robotics , vol. 37, no. 1, pp. 166–184, 2020

work page 2020
[6]

Planning to build block structures with unstable intermediate states using two manipulators,

H. Chen, W. Wan, K. Koyama, and K. Harada, “Planning to build block structures with unstable intermediate states using two manipulators,” IEEE Transactions on Automation Science and Engineering , 2021

work page 2021
[7]

A framework for fine robotic assembly,

F. Suárez-Ruiz and Q.-C. Pham, “A framework for fine robotic assembly,” in 2016 IEEE international conference on robotics and automation (ICRA) , pp. 421–426, IEEE, 2016

work page 2016
[8]

A survey of methods and strategies for high- precision robotic grasping and assembly tasks—some new trends,

R. Li and H. Qiao, “A survey of methods and strategies for high- precision robotic grasping and assembly tasks—some new trends,” IEEE/ASME Transactions on Mechatronics , vol. 24, no. 6, pp. 2718– 2732, 2019

work page 2019

Show all 27 references

[9]

The moped framework: Object recognition and pose estimation for manipulation,

A. Collet, M. Martinez, and S. S. Srinivasa, “The moped framework: Object recognition and pose estimation for manipulation,” The inter- national journal of robotics research , vol. 30, no. 10, pp. 1284–1306, 2011

2011
[10]

Variable compliance control for robotic peg-in-hole assembly: A deep-reinforcement-learning approach,

C. C. Beltran-Hernandez, D. Petit, I. G. Ramirez-Alpizar, and K. Harada, “Variable compliance control for robotic peg-in-hole assembly: A deep-reinforcement-learning approach,” Applied Sciences , vol. 10, no. 19, 2020

2020
[11]

Reinforcement learning on variable impedance controller for high-precision robotic assembly,

J. Luo, E. Solowjow, C. Wen, J. A. Ojea, A. M. Agogino, A. Tamar, and P. Abbeel, “Reinforcement learning on variable impedance controller for high-precision robotic assembly,” in 2019 International Conference on Robotics and Automation (ICRA) , pp. 3080–3087, 2019

2019
[12]

Deep reinforcement learning for high precision assembly tasks,

T. Inoue, G. De Magistris, A. Munawar, T. Yokoya, and R. Tachibana, “Deep reinforcement learning for high precision assembly tasks,” in 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 819–825, IEEE, 2017

2017
[13]

Learning sequences of manip- ulation primitives for robotic assembly,

N. Vuong, H. Pham, and Q.-C. Pham, “Learning sequences of manip- ulation primitives for robotic assembly,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) , pp. 4086–4092, IEEE, 2021

2021
[14]

Benchmarking protocols for evaluating small parts robotic assembly systems,

K. Kimble, K. Van Wyk, J. Falco, E. Messina, Y . Sun, M. Shibata, W. Uemura, and Y . Yokokohji, “Benchmarking protocols for evaluating small parts robotic assembly systems,” IEEE Robotics and Automation Letters, vol. 5, no. 2, pp. 883–889, 2020

2020
[15]

Experimental study of ordering of hard cubes by shearing,

K. Asencio, M. Acevedo, I. Zuriguel, and D. Maza, “Experimental study of ordering of hard cubes by shearing,” Phys. Rev. Lett., vol. 119, p. 228002, Dec 2017

2017
[16]

Multi-object rearrangement with monte carlo tree search: A case study on planar nonprehensile sorting,

H. Song, J. A. Haustein, W. Yuan, K. Hang, M. Y . Wang, D. Kragic, and J. A. Stork, “Multi-object rearrangement with monte carlo tree search: A case study on planar nonprehensile sorting,” in 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp....

2020
[17]

Monte-carlo tree search for efficient visually guided rearrangement planning,

Y . Labbé, S. Zagoruyko, I. Kalevatykh, I. Laptev, J. Carpentier, M. Aubry, and J. Sivic, “Monte-carlo tree search for efficient visually guided rearrangement planning,” IEEE Robotics and Automation Letters , vol. 5, no. 2, pp. 3715–3722, 2020

2020
[18]

Transporter networks: Rearranging the visual world for robotic manipulation,

A. Zeng, P. Florence, J. Tompson, S. Welker, J. Chien, M. Attarian, T. Armstrong, I. Krasin, D. Duong, V . Sindhwani, et al. , “Transporter networks: Rearranging the visual world for robotic manipulation,” in Conference on Robot Learning , pp. 726–747, PMLR, 2021

2021
[19]

Dall-e-bot: Introducing web- scale diffusion models to robotics,

I. Kapelyukh, V . V osylius, and E. Johns, “Dall-e-bot: Introducing web- scale diffusion models to robotics,” arXiv preprint arXiv:2210.02438 , 2022

2022 arXiv
[20]

Structformer: Learning spatial structure for language-guided semantic rearrangement of novel objects,

W. Liu, C. Paxton, T. Hermans, and D. Fox, “Structformer: Learning spatial structure for language-guided semantic rearrangement of novel objects,” in 2022 International Conference on Robotics and Automation (ICRA), pp. 6322–6329, IEEE, 2022

2022
[21]

Playing tangram with a humanoid robot,

J. Hirth, N. Schmitz, and K. Berns, “Playing tangram with a humanoid robot,” in ROBOTIK 2012; 7th German Conference on Robotics , pp. 1– 6, 2012

2012
[22]

Growing growth mindset with a social robot peer,

H. W. Park, R. Rosenberg-Kima, M. Rosenberg, G. Gordon, and C. Breazeal, “Growing growth mindset with a social robot peer,” in Proceedings of the 2017 ACM/IEEE international conference on human- robot interaction, pp. 137–145, 2017

2017
[23]

Expressive cognitive architecture for a curious social robot,

M. Rosenberg, H. W. Park, R. Rosenberg-Kima, S. Ali, A. K. Ostrowski, C. Breazeal, and G. Gordon, “Expressive cognitive architecture for a curious social robot,” ACM Transactions on Interactive Intelligent Systems (TiiS), vol. 11, no. 2, pp. 1–25, 2021

2021
[24]

Vision-guided hand–eye coordination for robotic grasping and its application in tangram puzzles,

H. Wei, S. Pan, G. Ma, and X. Duan, “Vision-guided hand–eye coordination for robotic grasping and its application in tangram puzzles,” AI, vol. 2, no. 2, pp. 209–228, 2021

2021
[25]

Robot tangram assembly line based on vision,

S. Qin, “Robot tangram assembly line based on vision,” in Journal of Physics: Conference Series , vol. 2229, p. 012017, IOP Publishing, 2022

2022
[26]

Prox- imal policy optimization algorithms,

J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Prox- imal policy optimization algorithms,” arXiv preprint arXiv:1707.06347 , 2017

2017 arXiv
[27]

Adam: A method for stochastic optimization,

D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” in 3rd International Conference on Learning Representations, ICLR 2015, San Diego, CA, USA, May 7-9, 2015, Conference Track Proceedings (Y . Bengio and Y . LeCun, eds.), 2015

2015

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

Learning insertion primitives with discrete-continuous hybrid action space for robotic assembly tasks,

X. Zhang, S. Jin, C. Wang, X. Zhu, and M. Tomizuka, “Learning insertion primitives with discrete-continuous hybrid action space for robotic assembly tasks,” in 2022 International Conference on Robotics and Automation (ICRA) , pp. 9881–9887, IEEE, 2022

work page 2022

[2] [2]

Form2fit: Learning shape priors for generalizable assembly from disassembly,

K. Zakka, A. Zeng, J. Lee, and S. Song, “Form2fit: Learning shape priors for generalizable assembly from disassembly,” in 2020 IEEE International Conference on Robotics and Automation (ICRA) , pp. 9404– 9410, IEEE, 2020

work page 2020

[3] [3]

Factory: Fast contact for robotic assembly,

Y . Narang, K. Storey, I. Akinola, M. Macklin, P. Reist, L. Wawrzyniak, Y . Guo, A. Moravanszky, G. State, M. Lu, et al., “Factory: Fast contact for robotic assembly,” arXiv preprint arXiv:2205.03532 , 2022

arXiv 2022

[4] [4]

Compliant peg-in-hole assembly using partial spiral force trajectory with tilted peg posture,

H. Park, J. Park, D.-H. Lee, J.-H. Park, and J.-H. Bae, “Compliant peg-in-hole assembly using partial spiral force trajectory with tilted peg posture,” IEEE Robotics and Automation Letters , vol. 5, no. 3, pp. 4447–4454, 2020

work page 2020

[5] [5]

Planning grasps with suction cups and parallel grippers using superimposed segmentation of object meshes,

W. Wan, K. Harada, and F. Kanehiro, “Planning grasps with suction cups and parallel grippers using superimposed segmentation of object meshes,” IEEE Transactions on Robotics , vol. 37, no. 1, pp. 166–184, 2020

work page 2020

[6] [6]

Planning to build block structures with unstable intermediate states using two manipulators,

H. Chen, W. Wan, K. Koyama, and K. Harada, “Planning to build block structures with unstable intermediate states using two manipulators,” IEEE Transactions on Automation Science and Engineering , 2021

work page 2021

[7] [7]

A framework for fine robotic assembly,

F. Suárez-Ruiz and Q.-C. Pham, “A framework for fine robotic assembly,” in 2016 IEEE international conference on robotics and automation (ICRA) , pp. 421–426, IEEE, 2016

work page 2016

[8] [8]

A survey of methods and strategies for high- precision robotic grasping and assembly tasks—some new trends,

R. Li and H. Qiao, “A survey of methods and strategies for high- precision robotic grasping and assembly tasks—some new trends,” IEEE/ASME Transactions on Mechatronics , vol. 24, no. 6, pp. 2718– 2732, 2019

work page 2019