Pith. sign in

REVIEW 5 major objections 7 minor 55 references

Sensor-Space Based Robust Kinematic Control of Redundant Soft Manipulator by Learning

T0 review · 5 major / 7 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read This paper shows that a redundant soft manipulator can be controlled reliably by learning policies in sensor space, so that unknown loads, actuator saturation, and confined-space obstacles can all be handled with zero-shot…

desk verdict A credible integration of sensor-space RL, GAIL, and S2R calibration with honest hardware demos; the zero-shot/load-robustness claims run ahead of the evidence because the S2R correction is load-free and statically calibrated. read the letter →

arxiv 2507.16842 v1 pith:VMVONUQJ submitted 2025-07-19 cs.RO

Yinan Meng , Kun Qian , Jiong Yang , Renbo Su , Zhenhong Li , Charlie C.L. Wang This is my paper · ORCID
classification cs.RO
keywords softmanipulatorcontrolsensor-spacereinforcementlearninggenerativeadversarialimitationsim-to-realtransferactuatorsaturationconfined-spacemanipulationkinematic
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

SS-ILKC is a learning-based kinematic controller for redundant soft manipulators that aims to solve three problems at once: unknown external loads, actuator saturation, and operation in confined spaces. The key idea is to control the robot in sensor space, using conductive springs that measure each chamber's length, so the control objective is load-independent geometry rather than pressure or actuation. An RL-derived multi-goal policy trained in a high-fidelity simulator handles workspace-wide reaching and avoids saturation, while a GAIL-based component learns rewards from sparse human demonstrations for confined-space tasks. A pre-calibrated sim-to-real correction lets the resulting policy deploy zero-shot on real hardware. The authors report path-following errors around 9 to 11 mm and successful pick-and-place inside a cross-shaped pipe, including on a pipe the policy was never trained on.

What carries the argument

The mechanism that carries the argument is the sensor-space representation: conductive springs integrated along each of the nine chambers act as proprioceptive sensors whose inductance maps to chamber length, so the controller observes geometry instead of pressure. Around this, the framework builds an MDP whose state includes the sim-to-real corrected pose, the goal, and scaled pose error, and whose action is the nine sensor setpoints; a PID loop tracks these setpoints through the pumps. The learning consists of a multi-goal RL stage using Truncated Quantile Critics with hindsight-experience-replay relabeling, and a GAIL stage whose discriminator, trained with a gradient penalty, supplies reward for demonstrations collected by remote control in confined spaces. The pre-processed sim-to-real transfer, a linear correction of spring lengths plus a lightweight network mapping simulated poses to real poses, is what makes the simulator an accurate enough teacher for zero-shot deployment.

What would settle it

Run a dense grid of goal-reaching and circular-path trials near the workspace boundary while loading the gripper with 1 kg and compare the simulated saturation predictions to real chamber pressures. If the real robot saturates in regions where the sim-to-real corrected simulator predicts no saturation, the zero-shot transfer claim is falsified; the paper's own Fig. 16 already shows one such mismatch.

Watch

Extended reading notes

Core claim

The paper's central claim is that the SS-ILKC framework provides reliable kinematic control for a pneumatically actuated, three-section soft manipulator by replacing actuator-space control with sensor-space control. The state is built from spring-derived sensor feedback and the sim-to-real corrected end-effector pose, the action is the nine chamber sensor signals, and the reward penalizes both reaching error and actuator saturation. Multi-goal RL with hindsight relabeling trains a policy that generalizes across the workspace, and a GAIL discriminator with gradient penalty converts sparse expert demonstrations into reward signals for confined-space behavior. Experiments on the physical robot show average translation errors of about 11 mm on a circular path without loads, about 9 mm under object loads between 35 and 500 g, and successful pick-and-place in a cross-shaped pipe including a generalization run with a narrower 250 mm pipe. The authors state the framework operates without post-training fine-tuning thanks to the pre-processed sim-to-real correction, and they acknowledge saturation near the workspace boundary under a 1 kg load as a residual limitation.

Load-bearing premise

The whole method rests on the assumption that a simulator corrected with a small set of measured poses and a simple spring-length adjustment accurately predicts the real robot's motion and its actuator limits everywhere in the workspace; the 1 kg boundary test shows this can fail where the calibration data are sparse.

Editorial extensions

If this is right

  • If the central claim holds, soft-manipulator kinematic control no longer needs an explicit actuator model or load model: the same policy trained without loads tracks paths while carrying objects up to 500 g.
  • The pre-calibrated simulator means a trained policy can be deployed zero-shot on real hardware, skipping the usual post-training fine-tuning or online adaptation loop.
  • Because rewards and demonstrations are expressed in sensor space, the learned behaviors transfer to a more confined environment, with pipe diameter reduced from 300 mm to 250 mm, without retraining the reward network.
  • Saturation avoidance is encoded during training, so null-space regulation emerges from the learned policy rather than from an optimization step that can push actuators to their limits.
  • The residual 1 kg-load failure marks the boundary of the method: near the workspace limit, saturation can still defeat the policy.

Reading between the lines

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

  • Beyond the paper, the same sensor-space plus GAIL plus sim-to-real recipe could be applied to other continuum or soft robots that have embedded proprioceptive length sensing, not just pneumatic three-section arms.
  • Beyond the paper, a denser, load-aware calibration that samples poses near the workspace boundary under several payloads would likely extend the zero-shot envelope beyond the demonstrated 500 g range.
  • Beyond the paper, combining the learned reward with online uncertainty estimation about the sim-to-real correction could turn the current open-loop transfer into a safety-aware deployment that knows when it is outside the calibrated region.
  • Beyond the paper, the GAIL-from-demonstrations component suggests a general way to inject human knowledge into RL controllers for soft robots in environments too complex to simulate faithfully, at the cost of collecting physical demonstrations.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 7 minor

Summary. The paper presents SS-ILKC, a learning-based kinematic control framework for a three-module pneumatically actuated soft manipulator. The control operates in sensor space: the policy outputs reference spring-length/frequency commands that a low-level PID controller tracks, with the manipulator state represented by conductive-spring inductance measurements. Training is twofold: a multi-goal reinforcement-learning (TQC) policy is trained in MuJoCo for open-space reaching, and a GAIL-based reward network learns from human demonstrations to handle confined-space tasks such as pick-and-place inside a cross-shaped pipe. A pre-processed sim-to-real (S2R) component, consisting of a linear mapping (Eq. 4) and a lightweight neural network NS2R, corrects simulated end-effector poses to match reality before policy training. Experiments demonstrate circular path following with mean translation error around 11 mm, successful path following with loads up to 500 g using a policy trained without loads, and pick-and-place in a 300 mm pipe that generalizes to a 250 mm pipe. The authors disclose a failure case: a 1 kg load causes chamber 2 to saturate and tracking error to reach 23.76 mm.

Significance. If the claims are substantiated, the framework would be a valuable contribution to soft-robot control: it addresses three hard problems (load-dependent deformation, actuator saturation, and confined-space operation) in a single sensor-space architecture, and the S2R pre-calibration idea is a useful alternative to post-hoc fine-tuning. The paper is honest about its limitations, provides ablation comparisons, and includes a supplemental video. However, the strengths are tempered by the absence of repeated trials and error bars, an S2R calibration whose coverage and error are not quantified, and claims of "zero-shot," "load-independent," and "saturation-free" behavior that are stronger than the evidence in Sec. 6 and Sec. 7 supports. The central idea is sound and the experimental scope is appropriate for a robotics venue, but several load-bearing validation points need to be strengthened.

major comments (5)
  1. [§3.3 and §6.2/§7] The S2R correction has no load input, and the simulator's saturation boundary is calibrated in unloaded conditions. The 1 kg experiment in Fig. 16 shows chamber 2 saturating near the workspace boundary, with tracking error spiking to 23.76 mm; this is the failure mode one would expect if the S2R mapping does not account for load-dependent changes in the feasible actuation range. The paper attributes the failure to "calibration errors near the workspace boundary," but the more direct interpretation is that the policy commands spring-length references that are infeasible under load. Please either add a load-conditioned term to the S2R mapping or the simulator, or explicitly bound the "load-independent" claim to loads below 1 kg and discuss how the saturation boundary scales with load. This is load-bearing because the RL policy is trained entirely inside the S2R-corrected simulation.
  2. [§6 (all experiments)] No experiment is repeated; all reported path errors, success rates, and load-test results appear to come from single runs. The claims of robustness, generalization, and the relative performance of the ablations (e.g., Fig. 14(c)) cannot be statistically assessed without error bars or confidence intervals. Please report at least three to five repetitions per condition, including the path-following, load, and pipe-transfer experiments, and provide mean and standard deviation for each reported metric.
  3. [§4.2.1 and §6.1] The goal-reaching threshold θ = 0.03, combined with the scaling vector w = diag([0.0056×I3, 0.001×I3]), implies a translation-equivalent tolerance of about 5.36 mm (0.03 / 0.0056). The reported average translation error of 11.29 mm for SS-ILKC is more than twice this tolerance, making it unclear whether the sparse reward Rg = 100 was ever awarded during the path-following evaluation. Please clarify the relationship between the threshold, the scaling vector, and the reported path-following error, and report the fraction of evaluation steps that satisfy the goal condition. This is important for interpreting the learning signal and the meaning of the final error.
  4. [§5.2 and §6.1/§6.3] The GAIL demonstrations are collected from the same task and environment used for evaluation: in Sec. 6.1 the demonstrations are 16 points along the very circular path later used for evaluation, and in Sec. 6.3 the demonstrations are recorded in the same cross-shaped pipe used for testing. This introduces a circularity risk because the reward network may memorize the demonstrated trajectories rather than represent a generalizable task distribution. The 16-to-40 point densification and the 250 mm pipe transfer provide some evidence of generalization, but the paper should either add an evaluation on a path that was never demonstrated or quantify how far the evaluated trajectories lie from the demonstration set. In addition, the number of demonstrations and the number of human demonstrators are not reported; please provide these details.
  5. [§3.3] The S2R network NS2R is trained on an unquantified set of real-world pose correspondences. The text states only the optimizer, learning rate, and batch size; it does not report the number of calibration poses, their spatial distribution, the train/validation/test split, or the network's prediction error on held-out poses. Without this information, the zero-shot deployment claim cannot be evaluated, and the risk of the network being accurate only in well-sampled regions remains unaddressed. Please provide a dataset description and a validation error map over the workspace.
minor comments (7)
  1. [Eq. (5)] The numerator appears to contain a typo: combining Eq. (3) with Lspring = β·l + δ yields a factor of 4π^3, not 4π^4, in front of μ0 N^2 C r^2 f^2. Please verify the derivation and correct the equation.
  2. [Eq. (4)] The values β = 0.93 and δ = 2.45 are given as fixed constants, but the estimation procedure is not described. State how these constants were obtained (e.g., least-squares fit on which dataset) and give their units.
  3. [Algorithm 1] The line "Sample goal uniformly from state space g ∼ Uniform(st)" is ambiguous: st is an element of the state space, not a distribution. If the goal is sampled uniformly from a set of possible goals, write this explicitly with the correct notation.
  4. [Algorithm 1] The replay buffer is reset to empty before each discriminator update and then repopulated with new transitions. This is unusual and could destabilize the RL training; please clarify whether this is intentional and discuss the effect on sample efficiency.
  5. [§6.1] The sentence "the learning framework exhibited a training error of 3.89 mm in the simulation environment" does not specify which variant of the framework produced this error and under what conditions. Please clarify the context.
  6. [Figure 12(c)] The caption mentions that chamber 2 is close to saturation, with a dashed red line, but the saturation pressure level is not defined. Add the saturation threshold value to the figure or caption.
  7. [References] Della Santina et al. 2020b and 2020c appear to be the same paper; please deduplicate and correct the in-text citations.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the learned policy, GAIL reward, and S2R calibration are distinct fitted components, and the central claims are supported by generalization tests that are not forced by the training data.

full rationale

The paper's derivation chain is not circular. The S2R network (Sec. 3.3) is trained on real-world pose correspondences, but it acts as a pre-calibration of the simulator rather than as a predictor of control performance; the RL policy (Sec. 4) is trained in the corrected simulator against externally defined goals and evaluated with dense-path generalization (16 to 40 points, Sec. 6.1), load tests up to 500 g (Sec. 6.2), and a 250 mm pipe with 37 new goals (Sec. 6.3) that are not statistically forced by the demonstration data. The fitted constants beta and delta in Eq. (4) and the NS2R weights are calibration parameters, not renamed predictions. The paper explicitly discloses the 1 kg saturation failure (Fig. 16) and attributes it to calibration error near the workspace boundary, which is a genuine robustness risk about S2R coverage but does not reduce any claim to its inputs by construction. Self-citations to Meng et al. (2024) are background for the hardware and sensor-space concept and are not load-bearing for the new learning results.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The approach rests on calibrated simulation and hand-set rewards rather than new physical entities. The free parameters above are fitted to real-world data or chosen by hand; none are derived from first principles, and no new physical objects or forces are postulated.

free parameters (6)
  • beta, delta linear mapping coefficients = beta=0.93, delta=2.45
    Eq (4): L_spring = beta*l + delta, fit to compensate manufacturing variability and sensor alignment errors.
  • S2R network NS2R weights = trained
    Trained with MSE loss on real-world (P, Lspring) to pose pairs; dataset size not reported.
  • simulated chamber stiffness and damping = adjusted
    Adjusted in Sec 3.2 so simulated poses match motion capture data.
  • reward function hyperparameters = Rg=100, eps=10, zeta=0.1, Rs=100, theta=0.03, w=diag([0.0056, 0.001]), lambda=20
    Set by hand in Sec 6.1; these values influence training and the goal success threshold.
  • fourth-order polynomial marker fits = coefficients not listed
    Used in Sec 3.2 to impose tendon length constraints from motion capture marker positions.
  • GAIL replay buffer mixing proportions = adjusted heuristically
    Sec 6.3: proportions of each demonstration in replay buffer adjusted every 1e5 steps based on training performance.
assumptions (5)
  • domain assumption Quasi-static operation
    Sec 4.1 formulates control as a sequence of quasi-static states; dynamics are not modeled.
  • domain assumption Sensor-space sufficiency
    Sec 1.3 and 2.2: spring lengths are assumed to determine manipulator configuration and end-effector pose, enabling load-independent control.
  • domain assumption MuJoCo tendon model fidelity
    Sec 3.1: pneumatic chambers modeled as nine tendons; fidelity relies on calibration and may not capture hysteresis and friction.
  • domain assumption Expert demonstration validity
    Sec 5.2: human teleoperated demonstrations are assumed to encode constraint-aware behavior suitable for reward learning.
  • standard math MDP and RL convergence
    Assumes TQC and GAIL converge to a valid policy; no convergence proof is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Sensor-Space Based Robust Kinematic Control of Redundant Soft Manipulator by Learning." pith.science (2026). https://pith.science/paper/VMVONUQJ

@misc{pith2026250716842,
  author       = {Pith},
  title        = {Pith review of: Sensor-Space Based Robust Kinematic Control of Redundant Soft Manipulator by Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VMVONUQJ}},
  note         = {Machine review of arXiv:2507.16842}
}
read the original abstract

The intrinsic compliance and high degree of freedom (DoF) of redundant soft manipulators facilitate safe interaction and flexible task execution. However, effective kinematic control remains highly challenging, as it must handle deformations caused by unknown external loads and avoid actuator saturation due to improper null-space regulation - particularly in confined environments. In this paper, we propose a Sensor-Space Imitation Learning Kinematic Control (SS-ILKC) framework to enable robust kinematic control under actuator saturation and restrictive environmental constraints. We employ a dual-learning strategy: a multi-goal sensor-space control framework based on reinforcement learning principle is trained in simulation to develop robust control policies for open spaces, while a generative adversarial imitation learning approach enables effective policy learning from sparse expert demonstrations for confined spaces. To enable zero-shot real-world deployment, a pre-processed sim-to-real transfer mechanism is proposed to mitigate the simulation-to-reality gap and accurately characterize actuator saturation limits. Experimental results demonstrate that our method can effectively control a pneumatically actuated soft manipulator, achieving precise path-following and object manipulation in confined environments under unknown loading conditions.

Figures

Figures reproduced from arXiv: 2507.16842 by the authors.

Figure 1
Figure 1. Challenges encountered in the kinematic control of the redundant soft manipulator include a) unknown loads, b) actuator saturation and c) operation within confined spaces. load-induced deformations, resulting in limited generaliz￾ability. Actuator saturation also significantly impacts the kine￾matic control of soft manipulators (see [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Schematic diagrams of (a) the sensor-space based control loop and (b) our SS-ILKC framework learning the control policies from both simulation (for open space) and physical demonstration (for confined space) – i.e., imitation learning. two control strategies for successfully transferring a trained RL policy to manage various task-space obstacles, even in the presence of actuation failure. However, operating redundan… view at source ↗
Figure 3
Figure 3. Workspace of the redundant soft manipulator. (a) A 3D representation of the workspace, with the color gradient indicating the distance from the origin. (b) A 2D projection of the workspace on the YZ-plane, with the red curve representing the theoretical boundary of the workspace determined by the manipulator’s actuation limits. The insets illustrate the stabilizing rings and the integrated conductive springs, which … view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: Soft manipulator calibration procedure and results: (a) The markers on soft manipulator are employed to obtain its configuration by the motion capture system, (b) polynomial fitting of X, Y, and Z coordinates for markers under maximum bending, (c) comparison of simulat…
Figure 5
Figure 5. Figure 5: The proposed Sim-to-Real (S2R) transfer employs a neural network bridges the gap between simulation and reality. parameters of the chambers are adjusted accordingly so that the simulated end-effector poses closely align with real￾world counterparts. The calibration res…
Figure 7
Figure 7. Figure 7: Schematic flow of the sensor-space multi-goal RL algorithm for obtaining the soft manipulator’s kinematic control policy in simulation. the state allows the policy to be trained under diverse goal conditions, resulting in a robust policy capable of reliable reaching ac…
Figure 8
Figure 8. Figure 8: Schematic flow of the SS-ILKC framework, which employs a GAIL-based approach to generate rewards from expert demonstrations. The newly added reward learning workflow, built upon our multi-goal RL algorithm in Sec. 4, is highlighted in orange to enrich the learning from…
Figure 9
Figure 9. Figure 9: Comparison of the losses with and without gradient penalty for: (a) critic loss, (b) actor loss, and (c) discriminator loss. functions, the reward is learned via an adversarial training process in SS-ILKC, where a discriminator Dψ(s, a, g) is trained to distinguish bet…
Figure 10
Figure 10. Figure 10: The schematic diagram of deploying sensor-space policy to control the soft manipulator in reality. Given a goal g, the trained policy outputs the optimal control action, i.e., at+1 based on the current state-action set (st, at, st+1). at+1 will then serve as the contr…
Figure 11
Figure 11. Figure 11: Control performance of path-following using our SS-ILKC framework under three different conditions: (1) w/o S2R transfer and physical demonstration, (2) with S2R transfer but w/o demonstration, and (3) full deployment. (a) 2D view of the reference and ground-truth pat…
Figure 12
Figure 12. Figure 12: Performance of SS-ILKC in the path following task in the presence of different loads. (a) The results of holding objects with weights as 35 g, 330 g and 500 g, where blue and red circles represent reference and actual manipulator’s paths respectively. (b) The referenc…
Figure 13
Figure 13. Figure 13: Performance verification of our SS-ILKC framework in executing a series of pick-and-place tasks within the confined space of a cross-shaped pipe. (a) Experimental setup with the specified dimensions and the weights of objects. (b) Demonstrated path, SS-ILKC path, and …
Figure 14
Figure 14. Figure 14: (a), we first evaluate the soft manipulator’s behavior after 3 × 105 steps. At this stage, the soft manipulator failed to precisely reach the target position and grasp the ball from the right container. As training progressed to 6 × 105 steps—the same number of steps …
Figure 15
Figure 15. Figure 15: Generalizability of the trained reward network – in these pick-and-place tests, we apply the policies trained on a setup as shown in Fig.13 on a new setup as shown in (a) with the pipe diameter reduced from 300 mm to 250 mm. (b) Completion of two pick-and-place tasks …
Figure 16
Figure 16. Figure 16: A limitation of our approach. (a) Chamber saturation occurs at the top section of the soft manipulator when a very large load is applied (e.g., 1 kg in this case). (b) The top view of path following results illustrate where the saturation happens. (c) Reference and ac…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 49 canonical work pages

  1. [1]

    Alessi C, Bianchi D, Stano G, Cianchetti M and Falotico E (2024) Pushing with soft robotic arms via deep reinforcement learning. Adv. Intell. Syst. 6(8): 2300899

    work page 2024

  2. [2]

    IEEE Trans

    Almanzor E, Ye F, Shi J, Thuruthel TG, Wurdemann HA and Iida F (2023) Static shape control of soft continuum robots using deep visual inverse kinematic models. IEEE Trans. Robot. 39(4): 2973--2988

    work page 2023

  3. [3]

    IEEE Robot

    Amehri W, Zheng G and Kruszewski A (2022) Fem-based exterior workspace boundary estimation for soft robots via optimization. IEEE Robot. Autom. Lett. 7(2): 3672--3678

    work page 2022

  4. [4]

    Andrychowicz M, Wolski F, Ray A, Schneider J, Fong R, Welinder P, McGrew B, Tobin J, Pieter Abbeel O and Zaremba W (2017) Hindsight experience replay. Adv. Neural Inf. Process. Syst. 30

    work page 2017

  5. [5]

    IEEE Trans

    Armanini C, Boyer F, Mathew AT, Duriez C and Renda F (2023) Soft robots modeling: A structured overview. IEEE Trans. Robot. 39(3): 1728--1748

    work page 2023

  6. [6]

    IEEE Robot

    Bruder D, Fu X, Gillespie RB, Remy CD and Vasudevan R (2021) Koopman-based control of a soft continuum manipulator under variable loading conditions. IEEE Robot. Autom. Lett. 6(4): 6852--6859

    work page 2021

  7. [7]

    Bruder D, Graule MA, Teeple CB and Wood RJ (2023) Increasing the payload capacity of soft robot arms by localized stiffening. Sci. Robot. 8(81): eadf9001

    work page 2023

  8. [8]

    IEEE Trans

    Cao S, Sun L, Jiang J and Zuo Z (2021) Reinforcement learning-based fixed-time trajectory tracking control for uncertain robotic manipulators with input saturation. IEEE Trans. Neural Netw. Learn. Syst. 34(8): 4584--4595

    work page 2021

Show all 55 references
  1. [9]

    In: DDCLS

    Cao Z, Huang T, Bao Z, Xie Y and Shi H (2022) Obstacle avoidance path planning and motion control for a multi-joint soft manipulator. In: DDCLS. pp. 1419--1426

    2022

  2. [10]

    IEEE Robot

    Centurelli A, Arleo L, Rizzo A, Tolu S, Laschi C and Falotico E (2022) Closed-loop dynamic control of a soft manipulator using deep reinforcement learning. IEEE Robot. Autom. Lett. 7(2): 4741--4748

    2022

  3. [11]

    IEEE Trans

    Chen J, Mu D, Hua C, Luo X, Zhang Y and Sun F (2024) Adaptive tracking control for uncertain unmanned fire fighting robot with input saturation and full-state constraints. IEEE Trans. Intell. Transp. Syst

    2024

  4. [12]

    IEEE Robot

    Della Santina C, Bicchi A and Rus D (2020 a ) On an improved state parametrization for soft robots with piecewise constant curvature and its use in model based control. IEEE Robot. Autom. Lett. 5(2): 1001--1008

    2020

  5. [13]

    IEEE Contr

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

    2023

  6. [14]

    Della Santina C, Katzschmann RK, Bicchi A and Rus D (2020 b ) Model-based dynamic feedback control of a planar soft robot: trajectory tracking and interaction with the environment. Int. J. Robot. Res. 39(4): 490--513

    2020

  7. [15]

    Della Santina C, Katzschmann RK, Bicchi A and Rus D (2020 c ) Model-based dynamic feedback control of a planar soft robot: trajectory tracking and interaction with the environment. Int. J. Robot. Res. 39(4): 490--513

    2020

  8. [16]

    IEEE Trans

    Fang G, Matte CD, Scharff RB, Kwok TH and Wang CC (2020) Kinematics of soft robots by geometric computing. IEEE Trans. Robot. 36(4): 1272--1286

    2020

  9. [17]

    IEEE Robot

    Fang G, Wang X, Wang K, Lee KH, Ho JD, Fu HC, Fu DKC and Kwok KW (2019) Vision-based online learning kinematic control for soft robots using local gaussian process regression. IEEE Robot. Autom. Lett. 4(2): 1194--1201

    2019

  10. [18]

    IEEE Trans

    Goury O and Duriez C (2018) Fast, generic, and reliable control and simulation of soft robots using model order reduction. IEEE Trans. Robot. 34(6): 1565--1576

    2018

  11. [19]

    Gulrajani I, Ahmed F, Arjovsky M, Dumoulin V and Courville AC (2017) Improved training of wasserstein gans. Adv. Neural Inf. Process. Syst. 30

    2017

  12. [20]

    Ho J and Ermon S (2016) Generative adversarial imitation learning. Adv. Neural Inf. Process. Syst. 29

    2016

  13. [21]

    In: ICRA

    Hu Y, Liu J, Spielberg A, Tenenbaum JB, Freeman WT, Wu J, Rus D and Matusik W (2019) Chainqueen: A real-time differentiable physical simulator for soft robotics. In: ICRA. pp. 6265--6271

    2019

  14. [22]

    IEEE ASME Trans

    Huang X, Zou J and Gu G (2021) Kinematic modeling and control of variable curvature soft continuum robots. IEEE ASME Trans. Mechatron. 26(6): 3175--3185

    2021

  15. [23]

    Science Robotics 4(26): eaau5872

    Hwangbo J, Lee J, Dosovitskiy A, Bellicoso D, Tsounis V, Koltun V and Hutter M (2019) Learning agile and dynamic motor skills for legged robots. Science Robotics 4(26): eaau5872

    2019

  16. [24]

    Izadbakhsh A and Khorashadizadeh S (2020) Robust adaptive control of robot manipulators using bernstein polynomials as universal approximator. Int. J. Robust Nonlinear Control 30(7): 2719--2735

    2020

  17. [25]

    In: CoRL

    Jitosho R, Lum TGW, Okamura A and Liu K (2023) Reinforcement learning enables real-time planning and control of agile maneuvers for soft robot arms. In: CoRL. pp. 1131--1153

    2023

  18. [26]

    Soft Robot

    Joshi S and Paik J (2021) Pneumatic supply system parameter optimization for soft actuators. Soft Robot. 8(2): 152--163

    2021

  19. [27]

    In: IJCAI, volume 2

    Kaelbling LP (1993) Learning to achieve goals. In: IJCAI, volume 2. pp. 1094--8

    1993

  20. [28]

    In: CoRL

    Kuznetsov A, Shvechikov P, Grishin A and Vetrov D (2020) Controlling overestimation bias with truncated mixture of continuous distributional quantile critics. In: CoRL. pp. 5556--5566

    2020

  21. [29]

    IEEE Robot

    Lai J, Lu B, Zhao Q and Chu HK (2022) Constrained motion planning of a cable-driven soft robot with compressible curvature modeling. IEEE Robot. Autom. Lett. 7(2): 4813--4820

    2022

  22. [30]

    arXiv preprint arXiv:1901.07517

    Lee J, Hwangbo J and Hutter M (2019) Robust recovery controller for a quadrupedal robot using deep reinforcement learning. arXiv preprint arXiv:1901.07517

    2019 arXiv

  23. [31]

    Soft Robot

    Martin A, Barrientos A and Del Cerro J (2018) The natural-ccd algorithm, a novel method to solve the inverse kinematics of hyper-redundant and soft robots. Soft Robot. 5(3): 242--257

    2018

  24. [32]

    IEEE ASME Trans

    Meng Y, Fang G, Yang J, Guo Y and Wang CCL (2024) Spring-imu fusion-based proprioception for feedback control of soft manipulators. IEEE ASME Trans. Mechatron. 29(2): 832--842

    2024

  25. [33]

    IEEE Trans

    Nazeer MS, Laschi C and Falotico E (2024) Rl-based adaptive controller for high precision reaching in a soft robot arm. IEEE Trans. Robot. 40: 2498--2512

    2024

  26. [34]

    (2018) Multi-goal reinforcement learning: Challenging robotics environments and request for research

    Plappert M, Andrychowicz M, Ray A, McGrew B, Baker B, Powell G, Schneider J, Tobin J, Chociej M, Welinder P et al. (2018) Multi-goal reinforcement learning: Challenging robotics environments and request for research. arXiv preprint arXiv:1802.09464

    2018 arXiv

  27. [35]

    IEEE Trans

    Renda F, Giorelli M, Calisti M, Cianchetti M and Laschi C (2014) Dynamic model of a multibending soft robot arm driven by cables. IEEE Trans. Robot. 30(5): 1109--1122

    2014

  28. [36]

    Nature 521(7553): 467--475

    Rus D and Tolley MT (2015) Design, fabrication and control of soft robots. Nature 521(7553): 467--475

    2015

  29. [37]

    In: International conference on machine learning

    Schaul T, Horgan D, Gregor K and Silver D (2015) Universal value function approximators. In: International conference on machine learning. PMLR, pp. 1312--1320

    2015

  30. [38]

    IEEE Trans

    Shao X, Pustina P, St \"o lzle M, Sun G, De Luca A, Wu L and Della Santina C (2023 a ) Model-based control for soft robots with system uncertainties and input saturation. IEEE Trans. Ind. Electron

    2023

  31. [39]

    IEEE Trans

    Shao X, Pustina P, St \"o lzle M, Sun G, De Luca A, Wu L and Della Santina C (2023 b ) Model-based control for soft robots with system uncertainties and input saturation. IEEE Trans. Ind. Electron

    2023

  32. [40]

    Shih B, Shah D, Li J, Thuruthel TG, Park YL, Iida F, Bao Z, Kramer-Bottiglio R and Tolley MT (2020) Electronic skins and machine learning for intelligent soft robots. Sci. Robot. 5(41): eaaz9239

    2020

  33. [41]

    IEEE Robot

    Spielberg A, Amini A, Chin L, Matusik W and Rus D (2021) Co-learning of task and sensor placement for soft robotics. IEEE Robot. Autom. Lett. 6(2): 1208--1215

    2021

  34. [42]

    Soft Robot

    Tang Z, Xin W, Wang P and Laschi C (2024) Learning-based control for soft robot--environment interaction with force/position tracking capability. Soft Robot

    2024

  35. [43]

    IEEE Trans

    Thuruthel TG, Falotico E, Renda F and Laschi C (2018) Model-based reinforcement learning for closed-loop dynamic control of soft robotic manipulators. IEEE Trans. Robot. 35(1): 124--134

    2018

  36. [44]

    Till J, Aloi V and Rucker C (2019) Real-time dynamics of soft and continuum robots based on cosserat rod models. Int. J. Robot. Res. 38(6): 723--746

    2019

  37. [45]

    In: IROS

    Todorov E, Erez T and Tassa Y (2012) Mujoco: A physics engine for model-based control. In: IROS. pp. 5026--5033

    2012

  38. [46]

    Van Diepen M and Shea K (2019) A spatial grammar method for the computational design synthesis of virtual soft locomotion robots. J. Mech. Des. 141(10): 101402

    2019

  39. [47]

    IEEE ASME Trans

    Wang Y, Wang H, Liu Z and Chen W (2020) Visual servo-collision avoidance hybrid task by considering detection and localization of contact for a soft manipulator. IEEE ASME Trans. Mechatron. 25(3): 1310--1321

    2020

  40. [48]

    IEEE Robot

    Xavier MS, Fleming AJ and Yong YK (2021) Design and control of pneumatic systems for soft robotics: A simulation approach. IEEE Robot. Autom. Lett. 6(3): 5800--5807

    2021

  41. [49]

    IEEE ASME Trans

    Xu F, Kang X and Wang H (2024) Hybrid visual servoing control of a soft robot with compliant obstacle avoidance. IEEE ASME Trans. Mechatron. 29(6): 4446--4455

    2024

  42. [50]

    IEEE Trans

    Yang C, Huang D, He W and Cheng L (2020) Neural control of robot manipulators with trajectory tracking constraints and input saturation. IEEE Trans. Neural Netw. Learn. Syst. 32(9): 4231--4242

    2020

  43. [51]

    IEEE ASME Trans

    Zhao Q, Lai J, Huang K, Hu X and Chu HK (2021) Shape estimation and control of a soft continuum robot under external payloads. IEEE ASME Trans. Mechatron. 27(5): 2511--2522

    2021

  44. [52]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution isbn journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence...

  45. [53]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

  46. [54]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter doi edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type url volume year label INTEGERS output.state before.all mid.sentence after.sentence aft...

  47. [55]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.