Affordance-Guided Dual-Armed Disassembly Teleoperation for Mating Parts

Genichiro Matsuda; Gen Sako; Kensuke Harada; Naoya Miyaji; Takuya Kiyokawa; Tomoki Ishikura

REVIEW 4 major objections 5 minor 25 references

Affordance-Guided Dual-Armed Disassembly Teleoperation for Mating Parts

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Adding geometry-derived grasp and pull-direction cues plus a force-compliant wrist lifts dual-arm disassembly of snap-fit parts from 80% to 100% success.

desk verdict A plausible integrated teleoperation system, but the 10-trial data and an internal ablation contradiction don't support the central claims. read the letter →

arxiv 2508.05937 v1 pith:F65P4YER submitted 2025-08-08 cs.RO

Gen Sako , Takuya Kiyokawa , Kensuke Harada , Tomoki Ishikura , Naoya Miyaji , Genichiro Matsuda This is my paper · ORCID

classification cs.RO

keywords teleoperationdual-armmanipulationdisassemblymatingpartsaffordanceguidanceimpedancecontrolteachingbydemonstrationnon-destructive

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 argues that the hard part of robotic disassembly of snap-together parts is not the motion but knowing where to grip and which way to pull while the part is hidden inside a larger product. To solve that, the authors build a teleoperation system in which a human demonstrates the task while the robot's virtual view overlays two geometry-derived suggestions: stable grasp poses for both grippers and the disassembly direction for the pulling arm. A one-arm-holds, one-arm-pulls setup, plus an impedance-controlled wrist that yields instead of rigidly tracking the human's commands, is tested on the front cover of a real air-conditioning unit. In ten trials per method, the success rate rises from 80% with no guidance to 100% with the full system, and the held object drifts less in position and orientation during the operation.

What carries the argument

The working core is an affordance overlay plus a compliant wrist. The overlay consists of stable grasp poses and a disassembly direction, both computed from the object's mesh geometry: grasp candidates come from the contact-pair geometry method of [23], and the pull direction comes from modeling each snap-fit connection as a spring with in-plane, out-of-plane, and rotational stiffness, as in [24]. The second mechanism is the hybrid controller: the nominal position command $\hat{x}(t)$ from the demonstrator's hand is corrected as $x_{\mathrm{ref}}(t) = \hat{x}(t) + (M\ddot{x}(t) + D\dot{x}(t) + Kx(t))^{-1}F(t)$, so that contact forces bend the reference trajectory rather than being fully resi

What would settle it

Measure the end-effector pose error introduced by the hand-pose-to-robot calibration by comparing the commanded pose from the tracked hand with the actually achieved gripper pose under no load, and rerun the ten-trial test with the affordance overlay hidden from the operator. If the calibration error exceeds the gripper's tolerance, or if success stays at 100% when the overlay is hidden, then the geometry guidance is not what is carrying the result.

Watch

Extended reading notes

Core claim

The central claim is that the combination of three components—dual-arm fixation, geometry-derived grasp and disassembly affordances, and impedance-controlled position tracking—turns teleoperated disassembly of mating parts from failure-prone into reliable and non-destructive. The fixation arm holds the product while the disassembly arm follows a human demonstration; the operator sees overlay arrows for feasible grasp poses, produced by the geometry-based grasp candidate algorithm [23], and for the extraction direction, produced by a snap-fit stiffness model [24]. The disassembly command is run through a mass-spring-damper impedance filter so that when the operator pulls hard along the sugges

Load-bearing premise

The paper assumes the calibrated mapping from the operator's tracked hand pose to the robot end-effector pose is accurate enough to drive the gripper to the suggested grasp and pull poses, yet it reports no measurement of that hand-pose tracking or calibration error.

Editorial extensions

If this is right

The same geometry-derived affordance pipeline can transfer to other snap-fit products by swapping in their mesh models, so the demonstration setup need not be redesigned per appliance.
The impedance filter removes the dominant failure modes of the baseline—object slipping and gripper-mount breakage—so the method targets non-destructive disassembly, not just task completion.
Dual-arm fixation keeps the workpiece within a small pose-deviation envelope during the demonstration, making the recorded human motion a stable training signal for later imitation learning.
Success rate and object pose deviation provide two simple metrics that future disassembly-teleoperation systems can use for direct comparison on the same task.

Reading between the lines

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

Since the no-impedance comparison also reached 100% success in ten trials, the hybrid controller's measured contribution appears to be damage prevention and lower deviation rather than raising the raw success rate; a stricter evaluation would count damage-free outcomes, not just binary success.
A blinded operator study that toggles only the affordance overlay would isolate how much of the gain comes from the visual guidance versus from the operator's own adaptation; the paper does not run that ablation.
The paper does not report the calibration error between the tracked hand pose and the commanded robot end-effector pose; measuring it would identify whether the teaching channel or the affordance is the bottleneck when scaling to smaller parts.
Because the impedance parameters M, D, K are fixed, online estimation of the part's stiffness from contact forces could make the same system handle unfamiliar materials without manual retuning.

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.

Desk Editor's Note

read the letter

Colleague,

Read this one with a red pen. The system is a reasonable integration of known pieces—Wan's grasp planner, Suri's spring-based direction estimation, Hart-style affordance templates, and impedance control—wrapped in a MediaPipe hand-tracking teleoperation front end. The hardware setup is concrete, and the idea of showing a human operator feasible grasps and extraction directions for a dual-arm fix-and-pull disassembly is practical and well motivated. The paper deserves credit for building the whole thing and testing it on a real AC unit.

The problem is that the empirical evidence as reported doesn't support the headline claims. Ten trials per condition: 80% vs 100% is within chance (Fisher one-sided p≈0.24). The pose-deviation plots have no variance or statistics. More worrying, Table I lists every method as using dual arms, but Section IV.D.2 refers to "methods without dual arms" having higher deviation. That means either the table is wrong or the ablation never happened. Either way, the claim that the fixation arm reduces deviation is not established by the printed experiment. Similarly, Eq. 3 is dimensionally invalid as written; the impedance correction should be the output of a transfer function, not the inverse of the time-domain operator applied to F. This looks like a typo in presentation, but it's central to the method.

There's also no measurement of hand-pose calibration accuracy, so we can't judge the teleoperation mapping.

For a systems paper, the core idea is worth a serious referee. It's an incremental but sensible engineering contribution for disassembly teleoperation, and the authors cite the relevant prior work appropriately. But it needs more trials, proper statistics, a corrected table/ablation description, a fixed Eq. 3, and ideally code/data before the claims can be accepted.

I'd send it to review, but expect major revisions. Not a paper I'd cite for the empirical results as they stand.

Referee Report

4 major / 5 minor

Summary. The paper presents an affordance-guided teleoperation framework for dual-arm disassembly of mating parts in large home appliances. Hand motion is captured via MediaPipe and mapped to end-effector poses; geometric grasp candidates and disassembly directions are visualized in a virtual environment; a hybrid position/impedance controller is proposed to avoid excessive contact forces. The system is evaluated on a condenser-unit front-cover disassembly task using two KUKA LBR iiwa arms, comparing three conditions (Baseline, Comparison, Proposed) with 10 trials each. The reported results are 80%, 100%, and 100% task success rates and lower object pose deviation for the proposed method.

Significance. If the claims were supported, the contribution would be practically relevant for remanufacturing: combining dual-arm fixation, affordance visualization, and compliant teleoperation is a sensible systems integration. The paper also builds on established geometric grasp planning and stiffness-model-based disassembly-direction estimation, which is appropriate. However, the current experimental evidence is too weak to establish the central causal claims. The success-rate difference is not statistically significant at n=10, the pose-deviation curves lack variance or statistical comparison, and there is an internal inconsistency between Table I and the text regarding a 'without dual arms' condition. The controller formulation in Eq. (3) is also mathematically invalid as written. These issues directly affect the paper's main conclusions.

major comments (4)

[§IV.D.1, Table I] The central success-rate claim is not supported by the reported data. With 10 trials per condition, the difference between 8/10 (Baseline) and 10/10 (Comparison/Proposed) has a Fisher exact one-sided p≈0.24, so the result is fully compatible with chance. Furthermore, Comparison and Proposed both achieve 10/10, so the success-rate metric provides no evidence for the effect of the hybrid controller. The authors should report per-trial outcomes, include confidence intervals or a statistical test, and either increase the number of trials or weaken the causal claims accordingly.
[§IV.D.2 vs. Table I] There is a direct contradiction. Table I lists exactly three conditions, each marked as using dual arms: Baseline has only 'Dual arms', Comparison adds 'Affordance', and Proposed adds 'Hybrid controller'. No condition omits dual arms. Yet §IV.D.2 states that 'the deviations in the methods without dual arms are higher than the proposed method over time.' This means the claim that dual-arm fixation reduces object pose deviation is not supported by the reported experiment. Either Table I is incomplete, or the text refers to experiments not described. This must be corrected before the pose-deviation conclusions can be assessed.
[Fig. 8, §IV.B.2] The pose-deviation results are presented as time curves without error bars, per-trial variance, or any statistical comparison. Equations (4)–(6) define the metric, but the paper does not report means, medians, or hypothesis tests. The statement that the proposed method 'reduced object pose deviation' is therefore not quantitatively supported. Please provide summary statistics and significance testing, or explicitly characterize Fig. 8 as representative trials rather than aggregate evidence.
[§III.C.3, Eq. (3)] Equation (3) is mathematically invalid as written: x_ref(t) = x_hat(t) + (M x¨(t) + D x˙(t) + Kx(t))^{-1} F(t). The term in parentheses is a vector (or an operator expression), not an invertible matrix, and the notation is dimensionally inconsistent. If the inverse operator is intended, Eq. (3) reduces to Eq. (2) because Eq. (1) already gives x = (M s^2 + D s + K)^{-1} F in the frequency domain. Please provide the actual implemented control law (e.g., a discrete-time admittance filter or transfer-function form) and ensure the equations are self-consistent.

minor comments (5)

[Eq. (5)] The equation for ||∆p_t|| contains unmatched parentheses: it should be sqrt((x_t−x_0)^2 + (y_t−y_0)^2 + (z_t−z_0)^2). The printed form has an extra closing parenthesis.
[§IV.A] The overview says methods were evaluated 'both with and without the use of dual-arm operation, affordance presentation, and impedance control,' but Table I includes no condition without dual arms. Please align the overview with the actual experimental design.
[§IV.B.2, Eq. (4)] The pose deviation metric adds a position term and an orientation term without specifying units or relative weighting. Clarify whether the variables are normalized and how the combination is justified.
[§V] The sentence 'This result suggests that dual-armed fix-and-disassembly is possible to reduce the deviation of the target object during the operations' is awkward; recommend 'can reduce' and hedge only if the evidence supports it.
[Acknowledgment] Spelling error: 'ACKNOLEDGEMENT' should be 'ACKNOWLEDGMENT'.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the claimed contribution is an empirically evaluated teleoperation system, not a derivation that reduces to its inputs.

full rationale

I walked the paper's claimed derivation chain. The central claims are (1) geometry-based grasp and disassembly-direction affordances, (2) a hybrid position/impedance controller, and (3) experimental success-rate and pose-deviation improvements. None of these is a fitted parameter renamed as a prediction. The grasp candidates are computed by the externally cited method of Wan et al. [23], and the disassembly direction uses the spring-model abstraction of Suri and Luscher [24]; these are independent inputs, not outputs of the present paper. The impedance control equations (1)-(3) are a standard mass-spring-damper formulation with the reference position defined as nominal plus compliant correction; even though Eq. (3) is dimensionally suspicious, that is a correctness/consistency issue, not circularity. The hand-pose mapping in Section III.B is a calibrated transformation, not a prediction. The success-rate and pose-deviation results are empirical measurements, not derived from the method's own parameters in a way that would make them true by construction. The only self-citation is reference [4] (Kiyokawa et al.), used for the background statement that grasp poses and disassembly directions are critical in robotic disassembly; this is not load-bearing for the proposed method or its evaluation. The inconsistency between Table I and the claim in Section IV.D.2 about 'methods without dual arms' is an experimental identification problem — none of the three conditions actually omits dual arms — but it is not a circularity, because the paper does not claim to derive the effect from a definition or from a self-cited uniqueness theorem. No step in the paper reduces by construction to its own inputs, so the appropriate circularity score is 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The system relies on several unverified engineering assumptions about sensor accuracy, geometric model fidelity, and controller behavior. No new physical entities or mechanisms are introduced; the contribution is the integration and application of existing methods.

free parameters (4)

Impedance parameters M, D, K = Not reported
The mass, damping, and stiffness in Eq. (1) are chosen by the authors, no values are given, and they directly determine the compliant behavior that is central to the claimed safety improvement.
Grasp candidate similarity thresholds = Not reported
In Section III.C.1, predefined thresholds on positional and orientational differences are used to select grasp candidates; these are hand-set and no sensitivity analysis is provided.
User-defined vicinity for grasp snapping = Not reported
The distance around a candidate grasp pose that triggers automatic alignment is user-defined in Section III.A and not specified.
Temporal scaling to 16 seconds = 16 seconds
All disassembly videos are rescaled to a common duration in Section IV.B.2, which affects the shape of the pose deviation trajectories and may bias comparisons.

assumptions (4)

domain assumption The geometry-based grasp planner of Wan et al. produces realizable grasps for the target parts.
The paper adopts this planner without validating its predicted grasps on the actual air conditioner cover; if the planner's assumptions about mesh fidelity or friction fail, the affordance guidance is wrong.
domain assumption The snap-fit connections can be modeled by spring elements as in Suri et al.
Section III.C.2 assumes this model with in-plane, out-of-plane, and rotational stiffness, but no experimental identification or validation of the spring constants is reported.
domain assumption MediaPipe hand keypoints and the calibrated transformation matrix accurately estimate the operator's hand pose.
Section III.B assumes that the estimated hand pose maps directly to the robot end-effector, but no accuracy or repeatability measurement of the hand-pose estimation is given.
domain assumption The two-finger gripper is sufficient to grasp and hold the target parts during disassembly.
The Robotiq Hand-E is selected without a grasp stability analysis for the specific part geometry; the baseline failure of dropping the object suggests this assumption is not always satisfied.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Affordance-Guided Dual-Armed Disassembly Teleoperation for Mating Parts." pith.science (2026). https://pith.science/paper/F65P4YER

@misc{pith2026250805937,
  author       = {Pith},
  title        = {Pith review of: Affordance-Guided Dual-Armed Disassembly Teleoperation for Mating Parts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F65P4YER}},
  note         = {Machine review of arXiv:2508.05937}
}

read the original abstract

Robotic non-destructive disassembly of mating parts remains challenging due to the need for flexible manipulation and the limited visibility of internal structures. This study presents an affordance-guided teleoperation system that enables intuitive human demonstrations for dual-arm fix-and-disassemble tasks for mating parts. The system visualizes feasible grasp poses and disassembly directions in a virtual environment, both derived from the object's geometry, to address occlusions and structural complexity. To prevent excessive position tracking under load when following the affordance, we integrate a hybrid controller that combines position and impedance control into the teleoperated disassembly arm. Real-world experiments validate the effectiveness of the proposed system, showing improved task success rates and reduced object pose deviation.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 25 canonical work pages

[1]

In what way is remanufacturing good for the environment?

E. Sundin and H. M. Lee, “In what way is remanufacturing good for the environment?” in Proc. EcoDesign, 2012, pp. 552–557

work page 2012
[2]

Disassembly 4.0: A review on using robotics in disassembly tasks as a way of automation,

H. Poschmann, H. Br ¨uggemann, and D. Goldmann, “Disassembly 4.0: A review on using robotics in disassembly tasks as a way of automation,” Chem. Ing. Tech., vol. 92, 2020

work page 2020
[3]

Mechanics of a snap fit,

K. Yoshida and H. Wada, “Mechanics of a snap fit,” Phys. Rev. Lett., vol. 125, no. 19, p. 194301, 2020

work page 2020
[4]

Many-Objective-Optimized Semi-Automated Robotic Disassembly Sequences

T. Kiyokawa, K. Harada, W. Wan, T. Ishikura, N. Miyaji, and G. Mat- suda, “Many-objective-optimized semi-automated robotic disassembly sequences,” arXiv preprint arXiv:2401.01817 , 2024

work page Pith review arXiv 2024
[5]

A survey of imitation learning: Algorithms, recent developments, and challenges,

M. Zare, P. M. Kebria, A. Khosravi, and S. Nahavandi, “A survey of imitation learning: Algorithms, recent developments, and challenges,” IEEE Trans. Cybern. , vol. 54, no. 12, pp. 7173–7186, 2024

work page 2024
[6]

A system for automated disassembly of snap-fit covers,

P. Schumacher and M. Jouaneh, “A system for automated disassembly of snap-fit covers,” J. Adv. Manuf. Technol. , vol. 69, pp. 2055–2069, 2013

work page
[7]

Human-in-the-loop approach for teaching robot assembly tasks using impedance control interface,

L. Peternel, T. Petri ˇc, and J. Babi ˇc, “Human-in-the-loop approach for teaching robot assembly tasks using impedance control interface,” in Proc. IEEE ICRA , 2015, pp. 1497–1502

work page 2015
[8]

Peg–hole disassembly using active compliance,

Y . Zhang, H. Lu, D. T. Pham, Y . Wang, M. Qu, J. Lim, and S. Su, “Peg–hole disassembly using active compliance,” R. Soc. Open Sci. , vol. 6, no. 8, p. 190476, 2019

work page 2019

Show all 25 references

[9]

Disassembly of rectangle peg-hole structures through compliance strategy and reinforcement learning,

F. Goli, Y . Wang, and M. Saadat, “Disassembly of rectangle peg-hole structures through compliance strategy and reinforcement learning,” in Proc. IEEE CASE , 2024, pp. 3926–3931

2024
[10]

The central nervous system stabilizes unstable dynamics by learning optimal impedance,

E. Burdet, R. Osu, D. W. Franklin, T. E. Milner, and M. Kawato, “The central nervous system stabilizes unstable dynamics by learning optimal impedance,” Nature, vol. 414, no. 6862, pp. 446–449, 2001

2001
[11]

The role of direct and visual force feedback in suturing using a 7-dof dual-arm teleoperated system,

A. Talasaz, A. L. Trejos, and R. V . Patel, “The role of direct and visual force feedback in suturing using a 7-dof dual-arm teleoperated system,” IEEE Trans. Haptics. , vol. 10, no. 2, pp. 276–287, 2017

2017
[12]

Imitation learning for variable speed contact motion for operation up to control band- width,

S. Sakaino, K. Fujimoto, Y . Saigusa, and T. Tsuji, “Imitation learning for variable speed contact motion for operation up to control band- width,” IEEE open j. Ind. Electron. Soc. , vol. 3, pp. 116–127, 2022

2022
[13]

Learning fine-grained bimanual manipulation with low-cost hardware,

T. Z. Zhao, V . Kumar, S. Levine, and C. Finn, “Learning fine-grained bimanual manipulation with low-cost hardware,” in Proc. RSS, 2023

2023
[14]

Two-layer-based multiarms bilateral teleoperation architecture,

M. Minelli, N. Piccinelli, F. Falezza, F. Ferraguti, R. Muradore, and C. Secchi, “Two-layer-based multiarms bilateral teleoperation architecture,” IEEE Trans. Control Syst. Technol. , vol. 31, no. 3, pp. 1266–1279, 2023

2023
[15]

Semi-autonomous robotic disassembly enhanced by mixed reality,

A. Rastegarpanah, C. A. Contreras, and R. Stolkin, “Semi-autonomous robotic disassembly enhanced by mixed reality,” in Proc. IRCE, 2024, pp. 7–13

2024
[16]

Evaluation of force feedback for palpation and application of active constraints on a teleoperated system,

E. Psomopoulou, R. Persad, A. Koupparis, S. Abeywardena, M. F. Sani, C. Melhuish, and S. Dogramadzi, “Evaluation of force feedback for palpation and application of active constraints on a teleoperated system,” in Proc. MEDICON, 2020, pp. 1571–1580

2020
[17]

The affordance template ros package for robot task programming,

S. Hart, P. Dinh, and K. Hambuchen, “The affordance template ros package for robot task programming,” in Proc. IEEE ICRA, 2015, pp. 6227–6234

2015
[18]

A versatile affordance modeling framework using screw primitives to increase autonomy during manipulation contact tasks,

A. Pettinger, F. Alambeigi, and M. Pryor, “A versatile affordance modeling framework using screw primitives to increase autonomy during manipulation contact tasks,” IEEE RA-L , vol. 7, no. 3, pp. 7224–7231, 2022

2022
[19]

Affordance-based grasping and manipulation in real world applications,

C. Pohl, K. Hitzler, R. Grimm, A. Zea, U. D. Hanebeck, and T. Asfour, “Affordance-based grasping and manipulation in real world applications,” in Proc. IEEE/RSJ IROS, 2020, pp. 9569–9576

2020
[20]

Mixed reality teleoperation assistance for direct control of humanoids,

L. Penco, K. Momose, S. McCrory, D. Anderson, N. Kitchel, D. Calvert, and R. J. Griffin, “Mixed reality teleoperation assistance for direct control of humanoids,” IEEE RA-L, vol. 9, no. 2, pp. 1937– 1944, 2024

1937
[21]

Using single demonstrations to define autonomous manipulation contact tasks in unstructured environments via object affordances,

F. Regal, A. Pettinger, J. A. Duncan, F. Parra, E. Akita, A. Navarro, and M. Pryor, “Using single demonstrations to define autonomous manipulation contact tasks in unstructured environments via object affordances,” in Proc. IEEE/RSJ IROS, 2023, pp. 3664–3671

2023
[22]

An intuitive, affordances oriented telemanipulation framework for a dual robot arm hand system: On the execution of bimanual tasks,

G. Gorjup, A. Dwivedi, N. Elangovan, and M. Liarokapis, “An intuitive, affordances oriented telemanipulation framework for a dual robot arm hand system: On the execution of bimanual tasks,” in Proc. IEEE/RSJ IROS, 2019, pp. 3611–3616

2019
[23]

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 Trans. Robot. , vol. 37, no. 1, pp. 166–184, 2020

2020
[24]

Structural abstraction in snap-fit analysis,

G. Suri and A. F. Luscher, “Structural abstraction in snap-fit analysis,” J. Mech. Des. , vol. 122, no. 4, pp. 395–402, 2000

2000
[25]

MediaPipe: A framework for perceiving and processing reality,

C. Lugaresi, J. Tang, H. Nash, C. McClanahan, E. Uboweja, M. Hays, F. Zhang, C.-L. Chang, M. Yong, J. Lee, et al. , “MediaPipe: A framework for perceiving and processing reality,” in CV4ARVR at IEEE CVPR, 2019

2019

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

In what way is remanufacturing good for the environment?

E. Sundin and H. M. Lee, “In what way is remanufacturing good for the environment?” in Proc. EcoDesign, 2012, pp. 552–557

work page 2012

[2] [2]

Disassembly 4.0: A review on using robotics in disassembly tasks as a way of automation,

H. Poschmann, H. Br ¨uggemann, and D. Goldmann, “Disassembly 4.0: A review on using robotics in disassembly tasks as a way of automation,” Chem. Ing. Tech., vol. 92, 2020

work page 2020

[3] [3]

Mechanics of a snap fit,

K. Yoshida and H. Wada, “Mechanics of a snap fit,” Phys. Rev. Lett., vol. 125, no. 19, p. 194301, 2020

work page 2020

[4] [4]

Many-Objective-Optimized Semi-Automated Robotic Disassembly Sequences

T. Kiyokawa, K. Harada, W. Wan, T. Ishikura, N. Miyaji, and G. Mat- suda, “Many-objective-optimized semi-automated robotic disassembly sequences,” arXiv preprint arXiv:2401.01817 , 2024

work page Pith review arXiv 2024

[5] [5]

A survey of imitation learning: Algorithms, recent developments, and challenges,

M. Zare, P. M. Kebria, A. Khosravi, and S. Nahavandi, “A survey of imitation learning: Algorithms, recent developments, and challenges,” IEEE Trans. Cybern. , vol. 54, no. 12, pp. 7173–7186, 2024

work page 2024

[6] [6]

A system for automated disassembly of snap-fit covers,

P. Schumacher and M. Jouaneh, “A system for automated disassembly of snap-fit covers,” J. Adv. Manuf. Technol. , vol. 69, pp. 2055–2069, 2013

work page

[7] [7]

Human-in-the-loop approach for teaching robot assembly tasks using impedance control interface,

L. Peternel, T. Petri ˇc, and J. Babi ˇc, “Human-in-the-loop approach for teaching robot assembly tasks using impedance control interface,” in Proc. IEEE ICRA , 2015, pp. 1497–1502

work page 2015

[8] [8]

Peg–hole disassembly using active compliance,

Y . Zhang, H. Lu, D. T. Pham, Y . Wang, M. Qu, J. Lim, and S. Su, “Peg–hole disassembly using active compliance,” R. Soc. Open Sci. , vol. 6, no. 8, p. 190476, 2019

work page 2019