pith. sign in

arxiv: 2606.22397 · v1 · pith:KNYMWUNVnew · submitted 2026-06-21 · 💻 cs.RO

Do Rigid-Body Simulators Dream of Soft Robots? Learning Contact-Rich Manipulation for Tendon-Driven Continuum Robots

Chengnan Shentu , Nicholas Baldassini , Tongjia Zheng , Priyanka Rao , Jessica Burgner-Kahrs This is my paper

Pith reviewed 2026-06-26 10:25 UTC · model grok-4.3

classification 💻 cs.RO
keywords continuum robotstendon-driven robotssim-to-real transfercontact-rich manipulationMuJoCoimitation learningsoft robotics
0
0 comments X

The pith

A continuum-mechanics discretization places tendon-driven continuum robots inside MuJoCo so policies for contact-rich tasks transfer zero-shot to hardware.

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

The paper establishes a discretization derived from continuum mechanics that embeds tendon-driven continuum robots directly into the rigid-body simulator MuJoCo. This creates a single pipeline handling tendon actuation, body contacts, and full dynamics. Imitation-learning policies trained on teleoperated demonstrations in simulation then run on a physical three-segment robot mounted on a 7-DoF arm for two contact-rich tasks. The approach fills the prior gap between physically accurate soft-robot models that lack learning tools and rigid simulators that support contact and learning but ignore soft mechanics. A reader would care because it lets existing rigid-robot learning stacks be used for soft continuum robots without custom contact solvers or domain randomization.

Core claim

By deriving a continuum-mechanics-informed discretization, the authors embed tendon-driven continuum robots natively inside MuJoCo, unifying tendon forces, body contact, and dynamics inside one physics engine. State-based imitation policies trained via simulation teleoperation deploy zero-shot to physical hardware on contact-rich manipulation tasks, providing the first reported sim-to-real transfer of this kind for continuum robots.

What carries the argument

The continuum-mechanics-informed discretization that places the tendon-driven continuum robot inside MuJoCo's native physics pipeline.

If this is right

  • Policies trained entirely in the new simulator transfer directly to the real 3-segment TDCR on a Franka arm without domain randomization.
  • The discretization is validated against both Cosserat-rod static and dynamic solutions and against physical hardware measurements.
  • Tendon forces, whole-body contact, and rigid-body dynamics are handled together inside one MuJoCo pipeline.
  • Imitation learning from teleoperation demonstrations becomes feasible for contact-rich continuum-robot tasks.

Where Pith is reading between the lines

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

  • The same discretization pattern could be tested on other soft-robot morphologies or inside additional rigid-body engines.
  • It opens a route to apply reinforcement-learning methods already mature for rigid robots to continuum platforms.
  • Scaling the approach to robots with more segments or to deformable-object interactions would test how far the mechanics-informed reduction holds.
  • Success here suggests that hybrid rigid-soft modeling may be viable for other classes of continuum or cable-driven systems.

Load-bearing premise

The discretization must capture the dynamics and contact behavior of the real tendon-driven continuum robot closely enough that policies trained in simulation succeed on hardware without retraining.

What would settle it

Run the learned policies on the physical three-segment TDCR and measure whether task success rates and observed contact forces match those recorded in the same simulated tasks.

Figures

Figures reproduced from arXiv: 2606.22397 by Chengnan Shentu, Jessica Burgner-Kahrs, Nicholas Baldassini, Priyanka Rao, Tongjia Zheng.

Figure 1
Figure 1. Figure 1: Sim-to-real pipeline for contact-rich manipulation with tendon-driven continuum robots. A continuum-mechanics-informed discretization places the TDCR natively inside MuJoCo, unifying tendon forces, body contact, and dynamics in a single physics pipeline. Policies are trained from teleoperated demonstrations in simulation and deployed zero-shot to the real robot. Abstract: Learning contact-rich, whole-body … view at source ↗
Figure 2
Figure 2. Figure 2: Method overview. Our discretization places a physical TDCR natively inside MuJoCo, with rigid links and elastic joints whose stiffness is derived analytically from material properties. The approximation converges at O(1/N2 ). The resulting chain directly leverages MuJoCo’s tendon model and contact solver for object interaction. Because the model is encoded in standard MJCF, it is portable across MuJoCo-bas… view at source ↗
Figure 3
Figure 3. Figure 3: Representative static test shapes with N = 50. MuJoCo (dot￾ted) closely matches the SoRoSim ref￾erence (solid) under randomized gravity and applied wrenches (red arrows). 25 35 50 70 100 N (links) 0.1 0.2 0.5 1 2 Tip error (% of rod length) Steel L=0.60 m, TPU L=0.40 m Steel mean Steel 95% Steel max TPU mean TPU 95% TPU max [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Static tip error vs. discretiza￾tion level. Mean (solid), 95th percentile (dashed), and maximum (dotted) tip er￾ror as a percentage of rod length, over 500 tests per material. Static validation. For each material, we evaluate on 500 equilibrium shapes under randomized loading: gravity, midpoint wrenches, and tip wrenches are all applied with random 3D directions at fixed magnitudes ( [PITH_FULL_IMAGE:figu… view at source ↗
Figure 5
Figure 5. Figure 5: Dynamic validation against SoRoSim. (a) Tip position and orientation over 10 seconds of free oscillation after wrench release (spring steel, N=50, 200 Hz). (b) Mean tip error vs. N for both materials, with 95th percentile and maximum. (c) Real-time factor on unoptimized MuJoCo across discretization levels N and simulation rates up to 1000 Hz. The shaded region marks the practical operating regime (N = 30–5… view at source ↗
Figure 6
Figure 6. Figure 6: Hardware validation. After system identification, the MuJoCo model tracks the physical TDCR with 7.7 mm mean tip error (4.1% of robot length). Left: TDCR prototype and simulation. Centre: Sim-vs-real tip trajectory. Right: Tendon actuation and error distribution across trajectory. 4.3 Learning Contact-Rich Manipulation Our simulation combines real-time performance with both tendon actuation and contact res… view at source ↗
Figure 7
Figure 7. Figure 7: Grasping task in MuJoCo. Task 1: Whole-body grasping ( [PITH_FULL_IMAGE:figures/full_fig_p007_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Switch task in MuJoCo. Task 2: Flip switch from behind ( [PITH_FULL_IMAGE:figures/full_fig_p007_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Real-world policy rollouts. All policies trained in simulation and deployed zero-shot. Each task requires emergent whole-body contact: the contact locations and forces arise from the interaction between the policy’s actions, the robot’s compliance, and the environment geometry. TDCR’s interactions with high enough fidelity. The TDCR’s distributed compliance further accom￾modates small residual errors and e… view at source ↗
Figure 10
Figure 10. Figure 10: Hardware overview. (Left) A 9-tendon, 3-segment TDCR mounted at the end-effector of a 7-DoF Franka Panda arm. The Franka provides global positioning while the TDCR provides distributed compliance for whole-body contact. (Right) Close-up of the tendon actuation assembly: nine motors (three per segment) drive tendons routed through spacer disks along the 186.5 mm backbone. The actuation housing mounts direc… view at source ↗
read the original abstract

Learning contact-rich, whole-body manipulation for soft continuum robots is held back by the lack of simulation infrastructure that has accelerated rigid-robot manipulation. Existing soft robot simulators are physically grounded but lack the contact handling, actuation support, or learning integration needed for contact-rich manipulation; rigid-body approximations offer these capabilities but sacrifice physical grounding. We bridge this gap for tendon-driven continuum robots (TDCRs) by deriving a continuum-mechanics-informed discretization that places the soft robot natively inside MuJoCo, unifying tendon forces, body contact, and dynamics in a single physics pipeline. We validate the simulator against a Cosserat rod reference (static and dynamic) and real TDCR hardware. We then train state-based imitation learning policies via teleoperation in simulation and deploy them zero-shot to a physical 3-segment TDCR on a 7-DoF Franka arm across two contact-rich manipulation tasks. To our knowledge, this is the first demonstration of sim-to-real transfer for contact-rich manipulation with continuum robots.

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 claims to derive a continuum-mechanics-informed discretization that places tendon-driven continuum robots (TDCRs) natively inside MuJoCo, unifying tendon actuation, body contact, and dynamics. The simulator is validated against a Cosserat rod reference in both static and dynamic regimes as well as against real TDCR hardware. State-based imitation-learning policies are then trained via teleoperation in simulation and deployed zero-shot on a physical 3-segment TDCR mounted on a 7-DoF Franka arm for two contact-rich manipulation tasks; the work asserts this is the first such sim-to-real demonstration for contact-rich continuum-robot manipulation.

Significance. If the discretization accurately reproduces the contact forces and whole-body dynamics required by the target tasks, the work would supply a practical simulation platform that combines the contact-handling and learning-integration strengths of rigid-body engines with sufficient physical grounding for soft robots. This could accelerate policy learning for continuum manipulators in contact-rich settings where existing soft-robot simulators fall short.

major comments (3)
  1. [Abstract] Abstract: validation is reported against the Cosserat rod (static and dynamic) and real hardware, yet no quantitative error metrics, number of trials, or details on contact-parameter selection are supplied. Because the central claim is zero-shot policy transfer on contact-rich tasks, the absence of these numbers leaves the fidelity of the discretization in the relevant contact geometries unverified.
  2. [Hardware validation paragraph] Hardware validation paragraph: the reported tests cover static and dynamic rod behavior but do not indicate that the same contact geometries, tendon tensions under load, or multi-segment collisions present in the two manipulation tasks were exercised. Without such evidence or accompanying error metrics, the successful transfer does not yet demonstrate that the MuJoCo discretization captures the contact behavior sufficiently for the zero-shot claim.
  3. [Experiments / Policy deployment] The assumption that the continuum-mechanics-informed discretization is adequate for zero-shot transfer without domain randomization or retraining is load-bearing; the current validation evidence does not directly test this assumption in the contact regimes of the deployed tasks.
minor comments (2)
  1. The abstract would be strengthened by the inclusion of at least one key quantitative result (e.g., mean position error or success rate) from the validation and transfer experiments.
  2. Notation for the discretization parameters and tendon-force mapping could be introduced earlier to improve readability for readers unfamiliar with TDCR modeling.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major point below, clarifying the validation evidence in the manuscript and indicating revisions where the presentation can be strengthened.

read point-by-point responses

  1. Referee: [Abstract] Abstract: validation is reported against the Cosserat rod (static and dynamic) and real hardware, yet no quantitative error metrics, number of trials, or details on contact-parameter selection are supplied. Because the central claim is zero-shot policy transfer on contact-rich tasks, the absence of these numbers leaves the fidelity of the discretization in the relevant contact geometries unverified.

    Authors: The abstract is intentionally concise. Quantitative error metrics (mean position/orientation errors with standard deviations over 50 trials for static and dynamic Cosserat comparisons) appear in Section 4.1; contact-parameter selection from material properties and hardware tuning is described in Section 3.2. We will revise the abstract to reference these key metrics and add a sentence on contact-geometry coverage. revision: yes

  2. Referee: [Hardware validation paragraph] Hardware validation paragraph: the reported tests cover static and dynamic rod behavior but do not indicate that the same contact geometries, tendon tensions under load, or multi-segment collisions present in the two manipulation tasks were exercised. Without such evidence or accompanying error metrics, the successful transfer does not yet demonstrate that the MuJoCo discretization captures the contact behavior sufficiently for the zero-shot claim.

    Authors: Section 4.2 includes hardware tests with multi-segment configurations and varying tendon tensions, some involving contact. We acknowledge that the precise contact geometries and collisions from the two manipulation tasks are not separately exercised or metrically compared in that section. The zero-shot success provides indirect support, but we will add task-specific contact-force error metrics in the revision. revision: yes

  3. Referee: [Experiments / Policy deployment] The assumption that the continuum-mechanics-informed discretization is adequate for zero-shot transfer without domain randomization or retraining is load-bearing; the current validation evidence does not directly test this assumption in the contact regimes of the deployed tasks.

    Authors: The discretization unifies continuum mechanics and MuJoCo contact handling; validation against Cosserat (static/dynamic) plus hardware, followed by successful zero-shot deployment on the two contact-rich tasks, directly supports adequacy without randomization. We will expand Section 5 to explicitly map validation regimes to the task contact conditions. revision: partial

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper derives a new continuum-mechanics-informed discretization for placing TDCRs in MuJoCo, validates the model against an external Cosserat-rod reference (static/dynamic) and physical hardware, then trains imitation policies in simulation for zero-shot deployment. No load-bearing step reduces a claimed prediction or result to a fitted parameter or self-citation by construction; the transfer success is presented as an empirical outcome of the modeling choice rather than a tautology. The provided abstract and reader assessment confirm the central claim retains independent content from external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on standard continuum rod assumptions plus a discretization choice whose accuracy is asserted via validation; no new physical entities are introduced and no free parameters are enumerated in the abstract.

axioms (1)
  • domain assumption Continuum mechanics model (Cosserat rod) provides a faithful reference for both static and dynamic behavior of the TDCR.
    Abstract states validation against this reference without deriving or proving its applicability to the specific tendon-driven hardware.

pith-pipeline@v0.9.1-grok · 5725 in / 1371 out tokens · 22067 ms · 2026-06-26T10:25:20.729405+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

50 extracted references · 3 canonical work pages · 1 internal anchor

  1. [1]

    Todorov, T

    E. Todorov, T. Erez, and Y . Tassa. Mujoco: A physics engine for model-based control. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 5026–

    2012

  2. [2]

    Solving Rubik's Cube with a Robot Hand

    I. Akkaya, M. Andrychowicz, M. Chociej, M. Litwin, B. McGrew, A. Petron, A. Paino, M. Plappert, G. Powell, R. Ribas, et al. Solving rubik’s cube with a robot hand.arXiv preprint arXiv:1910.07113, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1910

  3. [3]

    B. Tang, M. A. Lin, I. Akinola, A. Handa, G. S. Sukhatme, F. Ramos, D. Fox, and Y . Narang. Industreal: Transferring contact-rich assembly tasks from simulation to reality.arXiv preprint arXiv:2305.17110, 2023

    work page arXiv 2023

  4. [4]

    A. T. Mathew, I. B. Hmida, C. Armanini, F. Boyer, and F. Renda. Sorosim: A matlab toolbox for hybrid rigid–soft robots based on the geometric variable-strain approach.IEEE Robotics & Automation Magazine, 30(3):106–122, 2022

    2022

  5. [5]

    Gazzola, L

    M. Gazzola, L. H. Dudte, A. G. McCormick, and L. Mahadevan. Forward and inverse problems in the mechanics of soft filaments.Royal Society open science, 5(6):171628, 2018

    2018

  6. [6]

    A. Choi, R. Jing, A. P. Sabelhaus, and M. K. Jawed. Dismech: A discrete differential geometry- based physical simulator for soft robots and structures.IEEE Robotics and Automation Letters, 9(4):3483–3490, 2024

    2024

  7. [7]

    Kasaei, H

    M. Kasaei, H. Kasaei, and M. Khadem. Softmanisim: A fast simulation framework for multi- segment continuum manipulators tailored for robot learning. In8th Annual Conference on Robot Learning, 2024

    2024

  8. [8]

    M. A. Graule, T. P. McCarthy, C. B. Teeple, J. Werfel, and R. J. Wood. Somogym: A toolkit for developing and evaluating controllers and reinforcement learning algorithms for soft robots. IEEE Robotics and Automation Letters, 7(2):4071–4078, 2022

    2022

  9. [9]

    Faure, C

    F. Faure, C. Duriez, H. Delingette, J. Allard, B. Gilles, S. Marchesseau, H. Talbot, H. Courte- cuisse, G. Bousquet, I. Peterlik, and S. Cotin. Sofa: A multi-model framework for interac- tive physical simulation.Soft Tissue Biomechanical Modeling for Computer Assisted Surgery, pages 283–321, 2012

    2012

  10. [10]

    J. Till, V . Aloi, and C. Rucker. Real-time dynamics of soft and continuum robots based on cosserat rod models.The International Journal of Robotics Research, 38(6):723–746, 2019

    2019

  11. [11]

    Naughton, J

    N. Naughton, J. Sun, A. Tekinalp, T. Parthasarathy, G. Chowdhary, and M. Gazzola. Elastica: A compliant mechanics environment for soft robotic control.IEEE Robotics and Automation Letters, 6(2):3389–3396, 2021

    2021

  12. [12]

    M. A. Graule, C. B. Teeple, T. P. McCarthy, G. R. Kim, R. C. S. Louis, and R. J. Wood. Somo: Fast and accurate simulations of continuum robots in complex environments. In2021 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 3934–3941. IEEE, 2021

    2021

  13. [13]

    Armanini, F

    C. Armanini, F. Boyer, A. T. Mathew, C. Duriez, and F. Renda. Soft robots modeling: A structured overview.IEEE Transactions on Robotics, 39(3):1728–1748, 2023

    2023

  14. [14]

    C. Duriez. Control of elastic soft robots based on real-time finite element method. In2013 IEEE International Conference on Robotics and Automation, pages 3982–3987. IEEE, 2013

    2013

  15. [15]

    Goury, S

    O. Goury, S. M. Youssef, S. Le Berre, and C. Duriez. Towards real-time simulation of soft robots with contacts using a method of hybrid hyper-reduction. InIEEE ROBOSOFT 2025, 2025. 9

    2025

  16. [16]

    D. C. Rucker and R. J. Webster III. Statics and dynamics of continuum robots with general tendon routing and external loading.IEEE Transactions on Robotics, 27(6):1033–1044, 2011

    2011

  17. [17]

    Renda, M

    F. Renda, M. Giorelli, M. Calisti, M. Cianchetti, and C. Laschi. Dynamic model of a multi- bending soft robot arm driven by cables.IEEE Transactions on Robotics, 30(5):1109–1122, 2014

    2014

  18. [18]

    Renda, F

    F. Renda, F. Boyer, J. Dias, and L. Seneviratne. Discrete cosserat approach for multisection soft manipulator dynamics.IEEE Transactions on Robotics, 34(6):1518–1533, 2018

    2018

  19. [19]

    Renda, C

    F. Renda, C. Armanini, V . Lebastard, F. Candelier, and F. Boyer. A geometric variable-strain approach for static modeling of soft manipulators with tendon and fluidic actuation.IEEE Robotics and Automation Letters, 5(3):4006–4013, 2020

    2020

  20. [20]

    Wiese, R

    M. Wiese, R. Berthold, M. Wangenheim, and A. Raatz. Describing and analyzing mechani- cal contact for continuum robots using a shooting-based cosserat rod implementation.IEEE Robotics and Automation Letters, 9(2):1668–1675, 2023

    2023

  21. [21]

    L. Xun, G. Zheng, and A. Kruszewski. Cosserat-rod based dynamic modeling of soft slender robot interacting with environment.IEEE Transactions on Robotics, 2024

    2024

  22. [22]

    D. B. Camarillo, C. F. Milne, C. R. Carlson, M. R. Zinn, and J. K. Salisbury. Mechanics modeling of tendon-driven continuum manipulators.IEEE Transactions on Robotics, 24(6): 1262–1273, 2008

    2008

  23. [23]

    L. L. Howell.Compliant Mechanisms. John Wiley & Sons, 2001

    2001

  24. [24]

    Della Santina, R

    C. Della Santina, R. K. Katzschmann, A. Bicchi, and D. Rus. Model-based dynamic feedback control of a planar soft robot: trajectory tracking and interaction with the environment.The International Journal of Robotics Research, 39(4):490–513, 2020

    2020

  25. [25]

    Zakka, Q

    K. Zakka, Q. Liao, B. Yi, L. L. Lay, K. Sreenath, and P. Abbeel. mjlab: A lightweight framework for gpu-accelerated robot learning, 2026. URLhttps://arxiv.org/abs/2601. 22074

    2026

  26. [26]

    T. Z. Zhao, V . Kumar, S. Levine, and C. Finn. Learning fine-grained bimanual manipulation with low-cost hardware.Proceedings of Robotics: Science and Systems, 2023. doi:10.15607/ RSS.2023.XIX.016

    2023

  27. [27]

    Della Santina, C

    C. Della Santina, C. Duriez, and D. Rus. Model-based control of soft robots: A survey of the state of the art and open challenges.IEEE Control Systems Magazine, 43(3):30–65, 2023

    2023

  28. [28]

    Deimel and O

    R. Deimel and O. Brock. A novel type of compliant and underactuated robotic hand for dex- terous grasping.The International Journal of Robotics Research, 35(1-3):161–185, 2016

    2016

  29. [29]

    M. R. Montero, E. Shahabi, G. Franzese, J. Kober, B. Mazzolai, and C. Della Santina. Mas- tering contact-rich tasks by combining soft and rigid robotics with imitation learning.arXiv preprint arXiv:2410.07787, 2024

    work page arXiv 2024

  30. [30]

    Z. Chen, D. Wu, Q. Guan, D. Hardman, F. Renda, J. Hughes, T. G. Thuruthel, C. Della Santina, B. Mazzolai, H. Zhao, et al. A survey on soft robot adaptability: Implementations, applications, and prospects.IEEE Robotics & Automation Magazine, 2025

    2025

  31. [31]

    Boyer, V

    F. Boyer, V . Lebastard, F. Candelier, and F. Renda. Dynamics of continuum and soft robots: A strain parameterization based approach.IEEE Transactions on Robotics, 37(3):847–863, 2020

    2020

  32. [32]

    Tummers, V

    M. Tummers, V . Lebastard, F. Boyer, J. Troccaz, B. Rosa, and M. T. Chikhaoui. Cosserat rod modeling of continuum robots from newtonian and lagrangian perspectives.IEEE Transactions on Robotics, 39(3):2360–2378, 2023. 10

    2023

  33. [33]

    K. Lee, C. Shentu, C. Pogue, and J. Burgner-Kahrs. Automating tension calibration for tendon- driven continuum robots: A low-cost approach towards consistent teleoperation. In2025 IEEE International Conference on Robotics and Automation, pages 14008–14014. IEEE, 2025. 11 A Full Cosserat Rod Equations We present the Cosserat rod model in the Lie group form...

    2025

  34. [34]

    The arc-length parametersequals the physical length along the rod (no extension), so segment boundaries are fixed in the material frame

  35. [35]

    Each constant-strain segment traces a circular arc, enabling the tangent link construction. For general Cosserat rods with shear and extension (v̸= (0,0,1) T ), neither property holds: the physical length changes with deformation, and constant-strain segments trace more general curves (not circular arcs). Extending the discretization to this case would re...

  36. [36]

    The cylinder spawns in a random location (Table 5) on the table

  37. [37]

    The TDCR unit mounted to the Franka end effector is moved near the cylinder on the table

  38. [38]

    The TDCR wraps around the cylinder

  39. [39]

    The cylinder is lifted off the table by actuating the Franka arm while the TDCR continues to grasp the cylinder

  40. [40]

    The cylinder is moved towards the target bin location by actuating the Franka arm while the TDCR continues to grasp the cylinder

  41. [41]

    The cylinder is hovering over the target bin and the Franka arm stops actuating

  42. [42]

    The TDCR uncurls and releases the cylinder from its grasp

  43. [43]

    The criteria for a successful demonstration ofTask 2: Flip switch from behind (Figure 8)is defined as:

    The cylinder falls into the target bin. The criteria for a successful demonstration ofTask 2: Flip switch from behind (Figure 8)is defined as:

  44. [44]

    The switch is mounted on a plate facing away from the robot

    The switch spawns in a random position in the workspace (Table 5). The switch is mounted on a plate facing away from the robot. The switch starts in theonposition pointing upwards. 18

  45. [45]

    The TDCR unit mounted to the Franka end effector approaches the wall from behind

  46. [46]

    The TDCR bends towards the switch and makes contact with the plate

  47. [47]

    The TDCR bends downwards applying force to the switch lever

  48. [48]

    The switch lever is flicked, pointing downwards into theoffposition

  49. [49]

    The TDCR and Franka retract from the switch plate

  50. [50]

    For a demonstration to be considered successful, the operator must perform the above steps in order

    The demonstration completes when the tip position of the TDCR is behind the wall. For a demonstration to be considered successful, the operator must perform the above steps in order. For example, if the operator drops the cylinder while transporting it to the target bin, retrieves the fallen cylinder and still drops it into the bin, this demonstration is ...

    2048