Trajectory Planning of a Curtain Wall Installation Robot Based on Biomimetic Mechanisms

Tianlun Huang; Wei Feng; Weijun Wang; Xiao Liu; Zhiyong Wang

REVIEW 4 major objections 6 minor 20 references

Trajectory Planning of a Curtain Wall Installation Robot Based on Biomimetic Mechanisms

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A curtain wall installation robot can cut simulated peak energy use by 48.4 percent by shaping its trajectory like a human dumbbell curl.

desk verdict A genuinely bio-inspired trajectory constraint applied to a curtain wall robot, but the headline energy saving conflates peak and total reduction. read the letter →

arxiv 2507.16305 v1 pith:F5D2JPF2 submitted 2025-07-22 cs.RO

Xiao Liu , Weijun Wang , Tianlun Huang , Zhiyong Wang , Wei Feng This is my paper · ORCID

classification cs.RO

keywords biomimetictrajectoryplanningcurtainwallinstallationrobotenergyconsumptionoptimizationparticleswarmEMG-basedmotionanalysiskinetic-potentialconversionquinticpolynomialconstructionrobotics

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

The paper tries to show that a robot arm can save energy by borrowing the way humans time their muscle effort during a heavy dumbbell curl. Using motion and surface EMG data from a single lifting task, the authors derive a three-phase movement pattern and encode it as a trajectory constraint: the velocity peak should occur near the elbow angle where the arm is weakest, about 62 degrees. They optimize a quintic polynomial trajectory with particle swarm optimization under that constraint and test it in a dynamic simulation of a six-degree-of-freedom curtain wall installation robot. The simulation shows a 48.4 percent drop in the peak of the energy consumption curve compared with a conventional trajectory. A sympathetic reader would care because this suggests a cheap, parameter-light way to reduce energy use in construction robots, though the result currently rests on simulation only.

What carries the argument

The load-bearing mechanism is the kinetic-potential energy conversion principle, expressed as a three-phase segmentation of the lift: a high-load-capacity phase where short bursts of acceleration build kinetic energy, a weakest-load phase around 40 to 90 degrees of elbow flexion where kinetic energy is converted to potential energy to reduce joint torque, and a deceleration phase beyond 90 degrees where gravity does the braking. This biomechanical feature is implanted into a quintic polynomial trajectory—a fifth-degree polynomial in time used for smooth start-stop motion—through a particle swarm optimization constraint that places the velocity peak near 62 degrees of elbow flexion. The comparison is drawn between the conventional fifth-degree curve and the bio-constrained optimized curve.

What would settle it

Run the same conventional and bio-inspired trajectories on a physical six-degree-of-freedom curtain wall installation robot with a power meter measuring joint energy consumption; if the measured peak energy reduction is not close to 48.4 percent, or if the bio-inspired trajectory uses more energy than the conventional one, the simulated claim fails. A cheaper falsifier is to re-run the dynamic simulation with the velocity peak constraint moved away from 62 degrees and check whether the energy peak rises monotonically.

Watch

Extended reading notes

Core claim

The central claim is that human-like energy management—accelerating quickly in the strong posture, converting kinetic energy to potential energy while passing through the weak posture, and letting gravity decelerate at the end—can be transferred to robot trajectory planning as a single feature-point constraint. The paper maps the human shoulder and elbow to two key joints of a six-degree-of-freedom curtain wall installation robot, collects joint-angle and EMG data during dumbbell curls, and uses particle swarm optimization to shape a quintic polynomial trajectory so that peak elbow velocity is placed near 62 degrees of flexion. In the simulation, this optimized trajectory lowers the peak of the energy consumption curve by 48.4 percent relative to the standard quintic trajectory. The authors present this as evidence that kinetic-potential energy conversion, rather than trajectory smoothness alone, is a useful design principle for load-carrying robot motion.

Load-bearing premise

The paper's result rests on the assumption that one trajectory-shaping constraint from a single human dumbbell-curl experiment, peak elbow velocity placed near 62 degrees, transfers to a different six-degree-of-freedom robot and appears in a dynamic simulation without any physical robot measurement.

Editorial extensions

If this is right

A single biomechanical feature point, peak velocity at the weakest-load posture, can be used as a constraint in trajectory optimization without modeling full human muscle dynamics.
In the simulated curtain wall installation scenario, the bio-inspired trajectory lowers the peak energy consumption by 48.4 percent compared with the conventional quintic trajectory.
The three-phase movement template of accelerate, convert kinetic to potential, and decelerate is presented as reusable for other load-carrying robot tasks.
The paper's own conclusion states that the method is validated only in simulation and that hardware experiments with energy consumption analyzers are the next step.

Reading between the lines

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

If the 48.4 percent figure holds on hardware, the same one-constraint recipe might transfer to other heavy-lift robot tasks, but only if the robot's weakest-load posture is identified the same way; nothing in the paper guarantees the 62-degree value is universal.
The paper's logic suggests an alternative to minimizing jerk or time: deliberately shaping the velocity profile around the torque-capability envelope, which could be tested directly against minimum-jerk and minimum-time trajectories in the same simulation.
Because the constraint was extracted from a 12 kg dumbbell curl, varying payload mass in the simulation is an immediate testable extension; if the optimal elbow angle shifts with load, the single 62-degree value is a property of the human trial, not a general constant.

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

Quick take: the paper's real contribution is small but genuine. They use EMG-derived elbow angle at peak velocity as a constraint in PSO to shape a quintic trajectory, then test it in ADAMS on a curtain wall robot. That's a reasonable extension of known techniques. What is not genuine is the headline: the 48.4% reduction in the abstract is a peak energy curve reduction, while the actual total energy reduction reported in Section IV.C is 12%. Those are different metrics, and the paper never reconciles them.

What it does well: the human experiment (dumbbell curl with Mediapipe and Datalog EMG) is described clearly enough to follow. The idea of dividing motion into high-load, weakest-load, and deceleration phases is intuitive, and mapping that to a robotic trajectory is a sensible use of biomimetic data. They also honestly state in Section VI that validation is simulation-only, which is more than many such papers do.

Soft spots, in order of severity: - The central claim is undercut by the metric ambiguity. Peak power reduction is not energy savings; you can lower a peak while increasing integrated consumption. The abstract and conclusion use the 48.4% number as if it were total energy. - The PSO formulation is underspecified. We only see that a constraint was set at 62 degrees peak elbow velocity. No objective function, no design variables, no bounds, no algorithm parameters. That makes the optimization irreproducible. - No error bars, no sensitivity analysis, single ADAMS run. The simulation's energy metric is undefined, so we don't know what is being measured. - Minor typos in Eq. (9) and (10): the last coefficient should be d5, not d3. That suggests a lack of proofreading but doesn't break the method. - The 2-DOF human model is developed and then not used; the robot is 6-DOF. The mapping section is mostly verbal.

The self-referential concern (the constraint comes from your own data) is not itself fatal, but the authors don't test whether the 62-degree peak is robust to subject variability. That should be part of the sensitivity analysis.

Who is this for? Robotics practitioners looking for bio-inspired trajectory heuristics. It's not a theory paper; it's an application note with a plausible pipeline. With a rewritten section reconciling the metrics, a proper PSO description, and at least a sensitivity study, it could be a decent conference paper.

My recommendation: send it to peer review, but the referee should treat it as a major-revision case, not a clear reject. The core idea is testable; the paper just hasn't yet presented the evidence cleanly.

Referee Report

4 major / 6 minor

Summary. The paper proposes a biomimetic trajectory-planning framework for a curtain-wall installation robot. The authors collect kinematic and surface-EMG data from human dumbbell curls, divide the motion into three load phases, and use particle swarm optimization to shape a quintic-polynomial trajectory with the velocity peak placed at roughly 62 degrees of elbow flexion. The planned trajectory is then evaluated in an ADAMS simulation against a conventional quintic trajectory. The abstract claims a 48.4% reduction in energy consumption, while Section IV.C reports about 12% and Section V reports a 48.4% reduction in the peak of the energy-consumption curve. The paper also presents a Lagrangian dynamic model of a simplified two-link arm and a 6-DOF curtain-wall robot simulation, with no hardware experiments.

Significance. If the result were substantiated, the transfer of a human kinetic-potential energy conversion principle to trajectory planning for a construction robot would be a useful contribution to energy-efficient robotics. The idea of using human movement data to set a temporal constraint in trajectory optimization is plausible and worth exploring. However, the manuscript currently lacks a reproducible optimization formulation, a clear definition of the simulated energy metric, and any statistical or experimental validation, so the headline claim is not yet established.

major comments (4)

[Abstract, Section IV.C, Section V] The central energy-saving claim is inconsistently stated. The abstract says a '48.4% reduction in energy consumption'; Section V says 'The peak of the energy consumption curve decreased by 48.4%'; and Section IV.C says 'the optimized trajectory significantly reduces energy consumption by about 12%.' Peak energy-consumption rate and total integrated energy are different quantities, and the text never defines the ADAMS energy-consumption metric. As written, the reader cannot determine which quantity is actually smaller, and the headline figure is therefore not supported.
[Section IV.C] The PSO-based trajectory optimization is underspecified. No objective function, design variables, bounds, or algorithm parameters are given; the only stated constraint is a peak elbow velocity of approximately 62 degrees. Equations (9)-(11) define only the baseline quintic polynomial. Without the PSO formulation, the optimized trajectory cannot be reproduced, and the claimed causal link between the biomimetic constraint and the simulated energy reduction cannot be assessed.
[Section V and Section VI] The ADAMS simulation is the sole evidence for the energy claims, but the energy-consumption quantity is not defined, no multiple simulation runs or error bars are reported, and no sensitivity analysis is provided. Section VI concedes that the current research is 'validated solely through simulation.' Consequently, the 48.4% figure is a single point estimate from an unspecified simulator output, which is insufficient to support the paper's headline claim.
[Section IV.B and Section IV.C] The biomimetic constraint is self-referential as evidence for anthropomorphism: the phase boundaries (0-40, 40-90, 90-150 degrees) and the 62-degree peak-velocity constraint are extracted from the authors' own human-subject data and then imposed in PSO. The energy comparison against the conventional baseline is independent, so the energy claim is not circular, but the claim that the savings arise specifically from 'human energy conversion principles' is not supported without a mechanism analysis or a sensitivity study showing that the savings depend on the bio-inspired constraint rather than on another feature of the optimized trajectory.

minor comments (6)

[Equation (9)] Equation (9) contains a typo: the coefficient of t^5 is written as d3, but it should be d5 in a quintic polynomial.
[Equation (10)] Equation (10) contains a typo: the last term should be 20*d5*t^3, not 20*d3*t^3.
[Equation (8)] Equation (8) is dimensionally unclear: it shows the same 2x2 matrix multiplied by the acceleration vector and then by a vector of squared angular velocities. If the second term is intended to represent Coriolis and centrifugal effects, it should be written with a distinct matrix and explained, rather than using the same symbol M.
[Section II.B] The Lagrangian dynamic model in Section II.B is not used in the PSO optimization or the ADAMS simulation. The paper should either connect this model to the optimization objective or clearly state that the simulation uses the multibody dynamics of ADAMS.
[Section IV.C] The phrase 'significantly reduces energy consumption by about 12%' uses 'significant' without any statistical test; since only a single simulation is presented, the word should be replaced with a quantitative comparison.
[Section V] Figure 10 is referenced only indirectly; the text should describe what is plotted in the simulation result figure, including the units of the energy-consumption axis.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the human-derived trajectory constraint is an input, and the energy reduction is independently simulated against a conventional baseline.

full rationale

The paper's derivation chain is: (1) collect human EMG and motion data during dumbbell curls; (2) identify characteristic phases and a peak-velocity location at roughly 62 degrees of elbow flexion; (3) impose this as a constraint in PSO on quintic polynomial trajectories; (4) compare the resulting trajectory against a conventional quintic trajectory in an ADAMS simulation of a 6-DOF curtain wall robot, reporting a 48.4% decrease in the peak energy-consumption curve and about 12% reduction in energy. The human-derived 62-degree constraint is an input to the optimization, not a quantity re-derived from the output energy numbers; the ADAMS energy comparison is computed on a robot model and against a conventional baseline, so the headline improvement is not forced by construction. There are no load-bearing self-citations: the references are external, and no uniqueness theorem is invoked. The paper's own limitation that the research is 'validated solely through simulation' and the discrepancy between peak (48.4%) and total (~12%) energy reduction are concerns about correctness and interpretation, not circularity. No circular step is exhibited, so the appropriate finding is no significant circularity.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim relies on a small number of parameters taken from one human experiment and on the unvalidated fidelity of the ADAMS simulation. No new physical entities are introduced.

free parameters (2)

Peak-velocity elbow angle constraint = ~62 degrees
Selected from human EMG and motion data as the characteristic PSO constraint in Section IV.C; no sensitivity analysis is reported.
Elbow phase boundaries = 0-40, 40-90, 90-150 degrees
Hand-divided movement phases in Section IV.B based on EMG observations; treated as hard boundaries without justification.

assumptions (4)

domain assumption The 2-DOF planar rigid-link model accurately captures the energy conversion behavior of the human upper limb during a dumbbell curl.
Introduced in Section II.B to interpret EMG and motion data; no validation against measured joint torques or metabolic energy.
domain assumption Quintic polynomial interpolation provides a sufficiently representative trajectory family for comparison.
Adopted in Section IV.A after a qualitative comparison of curve types; no metric for the selection is given.
ad hoc to paper The human shoulder/elbow mapping to the robot base/intermediate joints preserves the energy-saving mechanism.
Stated in Section V with a mechanism mapping diagram but no dynamic analysis of the 6-DOF robot.
domain assumption ADAMS simulation faithfully represents the curtain wall robot's energy consumption.
Used in Section V as the only evidence for the 48.4 percent reduction; no calibration, validation, or experimental comparison is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Trajectory Planning of a Curtain Wall Installation Robot Based on Biomimetic Mechanisms." pith.science (2026). https://pith.science/paper/F5D2JPF2

@misc{pith2026250716305,
  author       = {Pith},
  title        = {Pith review of: Trajectory Planning of a Curtain Wall Installation Robot Based on Biomimetic Mechanisms},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/F5D2JPF2}},
  note         = {Machine review of arXiv:2507.16305}
}

read the original abstract

As the robotics market rapidly evolves, energy consumption has become a critical issue, particularly restricting the application of construction robots. To tackle this challenge, our study innovatively draws inspiration from the mechanics of human upper limb movements during weight lifting, proposing a bio-inspired trajectory planning framework that incorporates human energy conversion principles. By collecting motion trajectories and electromyography (EMG) signals during dumbbell curls, we construct an anthropomorphic trajectory planning that integrates human force exertion patterns and energy consumption patterns. Utilizing the Particle Swarm Optimization (PSO) algorithm, we achieve dynamic load distribution for robotic arm trajectory planning based on human-like movement features. In practical application, these bio-inspired movement characteristics are applied to curtain wall installation tasks, validating the correctness and superiority of our trajectory planning method. Simulation results demonstrate a 48.4% reduction in energy consumption through intelligent conversion between kinetic and potential energy. This approach provides new insights and theoretical support for optimizing energy use in curtain wall installation robots during actual handling tasks.

Figures

Figures reproduced from arXiv: 2507.16305 by the authors.

**Figure 1.** Overview of the Proposed Method strategies of coordinated human upper limb movements is of tremendous value for robot motion planning and intelligent control. Commonly used parameters to describe human upper limb movement patterns include joint angles, angular velocity, electromyography (EMG) signals, mobility indices, and performance indices [14]. [15] investigated the relationship between average task completion t… view at source ↗

**Figure 2.** Simplified upper limb model Establish a basic dynamic model, where τ represents the joint torque. The dynamics model is constructed using [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Motion posture detection based on mediapipe [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (7 more)

**Figure 5.** Figure 5: Comparison of velocity between different classical curves [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

**Figure 4.** Figure 4: Collection of EMG signals for upper limb movements [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 6.** Figure 6: Fifth degree polynomial curve B. Bionic Mechanisms Based on Dataset Real-time EMG data from various muscle groups were collected using the Datalog EMG system. This enabled the identification of the activation sequence and intensity variations of key muscle groups durin…

**Figure 7.** Figure 7: Comparison between Conventional Trajectory and Humanoid [PITH_FULL_IMAGE:figures/full_fig_p005_7.png]

**Figure 8.** Figure 8: Mechanism mapping between human upper limbs and robots [PITH_FULL_IMAGE:figures/full_fig_p006_8.png]

**Figure 10.** Figure 10: Simulation result diagram of curtain wall installation robot [PITH_FULL_IMAGE:figures/full_fig_p006_10.png]

**Figure 9.** Figure 9: Simulation process diagram of curtain wall installation robot [PITH_FULL_IMAGE:figures/full_fig_p006_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

[1]

From biomimicry to robotic co-creation: rethinking the boundaries between nature and technology[J]

Tamborini M. From biomimicry to robotic co-creation: rethinking the boundaries between nature and technology[J]. Bioinspiration & Biomimetics, 2024, 19(2): 023001

work page 2024
[2]

Bioinspiration and biomimetic art in robotic grippers[J]

Nguyen V P, Dhyan S B, Mai V , et al. Bioinspiration and biomimetic art in robotic grippers[J]. Micromachines, 2023, 14(9): 1772

work page 2023
[3]

Recent advances in sources of Bio- Inspiration and materials for robotics and actuators[J]

Yang Y , Ai C, Chen W, et al. Recent advances in sources of Bio- Inspiration and materials for robotics and actuators[J]. Small Methods, 2023, 7(9): 2300338

work page 2023
[4]

Design method for a bionic wrist based on tensegrity structures[J]

Sun J, Cao X, Song G. Design method for a bionic wrist based on tensegrity structures[J]. Biosurface and Biotribology, 2020, 6(2): 31-36

work page 2020
[5]

The Latest Research Progress on Bionic Artificial Hands: A Systematic Review[J]

Guo K, Lu J, Wu Y , et al. The Latest Research Progress on Bionic Artificial Hands: A Systematic Review[J]. Micromachines, 2024, 15(7): 891

work page 2024
[6]

Mechatronic design of a robot for upper limb rehabilitation at home[J]

Curcio E M, Carbone G. Mechatronic design of a robot for upper limb rehabilitation at home[J]. Journal of Bionic Engineering, 2021, 18(4): 857-871

work page 2021
[7]

A review on design of upper limb exoskeletons[J]

Gull M A, Bai S, Bak T. A review on design of upper limb exoskeletons[J]. Robotics, 2020, 9(1): 16

work page 2020
[8]

Mechanical design and friction modelling of a cable-driven upper-limb exoskeleton[J]

Deˇzman M, Asfour T, Ude A, et al. Mechanical design and friction modelling of a cable-driven upper-limb exoskeleton[J]. Mechanism and Machine Theory, 2022, 171: 104746

work page 2022

Show all 20 references

[9]

Design and evaluation of a four-DoF upper limb exoskeleton with gravity compensation[J]

Shi K, Yang J, Hou Z, et al. Design and evaluation of a four-DoF upper limb exoskeleton with gravity compensation[J]. Mechanism and Machine Theory, 2024, 201: 105746

2024
[10]

Enhancing the performance of a biomimetic robotic elbow-and-forearm system through bionics-inspired optimiza- tion[J]

Yang H, Wei G, Ren L. Enhancing the performance of a biomimetic robotic elbow-and-forearm system through bionics-inspired optimiza- tion[J]. IEEE Transactions on Robotics, 2024

2024
[11]

Biomimetic approaches for human arm motion generation: literature review and future directions[J]

Trivedi U, Menychtas D, Alqasemi R, et al. Biomimetic approaches for human arm motion generation: literature review and future directions[J]. Sensors, 2023, 23(8): 3912

2023
[12]

Human-like arm motion generation: A review[J]

Gulletta G, Erlhagen W, Bicho E. Human-like arm motion generation: A review[J]. Robotics, 2020, 9(4): 102

2020
[13]

Impact of motor task conditions on end-point kinematics and trunk movements during goal-directed arm reach[J]

Kim B, Girnis J, Sweet V , et al. Impact of motor task conditions on end-point kinematics and trunk movements during goal-directed arm reach[J]. Scientific Reports, 2024, 14(1): 4520

2024
[14]

Hierarchical motion learning for goal- oriented movements with speed–accuracy tradeoff of a musculoskeletal system[J]

Zhou J, Zhong S, Wu W. Hierarchical motion learning for goal- oriented movements with speed–accuracy tradeoff of a musculoskeletal system[J]. IEEE Transactions on Cybernetics, 2021, 52(11): 11453- 11466

2021
[15]

The information capacity of the human motor system in controlling the amplitude of movement [J]

FITTS P M. The information capacity of the human motor system in controlling the amplitude of movement [J]. Experimental Psychology, 1954, 47:381-391

1954
[16]

Synergic control of movement: from single muscles to the whole body[J]

Latash M L. Synergic control of movement: from single muscles to the whole body[J]. Progress in motor control, 2024: 25-47

2024
[17]

Learning by demonstration for motion planning of upper-limb exoskeletons[J]

Lauretti C, Cordella F, Ciancio A L, et al. Learning by demonstration for motion planning of upper-limb exoskeletons[J]. Frontiers in neurorobotics, 2018, 12: 5

2018
[18]

Anthropomorphic reaching movement generating method for human-like upper limb robot[J]

He C, Xu X W, Zheng X F, et al. Anthropomorphic reaching movement generating method for human-like upper limb robot[J]. IEEE Transactions on Cybernetics, 2021, 52(12): 13225-13236

2021
[19]

Mapping Method of Human Arm Motion Based on Surface Electromyography Signals[J]

Zheng Y , Zheng G, Zhang H, et al. Mapping Method of Human Arm Motion Based on Surface Electromyography Signals[J]. Sensors, 2024, 24(9): 2827

2024
[20]

Upper extremity joint torque estimation through an electromyography-driven model[J]

Tahmid S, Font-Llagunes J M, Yang J. Upper extremity joint torque estimation through an electromyography-driven model[J]. Journal of Computing and Information Science in Engineering, 2023, 23(3): 030901

2023

Pith tools

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

[1] [1]

From biomimicry to robotic co-creation: rethinking the boundaries between nature and technology[J]

Tamborini M. From biomimicry to robotic co-creation: rethinking the boundaries between nature and technology[J]. Bioinspiration & Biomimetics, 2024, 19(2): 023001

work page 2024

[2] [2]

Bioinspiration and biomimetic art in robotic grippers[J]

Nguyen V P, Dhyan S B, Mai V , et al. Bioinspiration and biomimetic art in robotic grippers[J]. Micromachines, 2023, 14(9): 1772

work page 2023

[3] [3]

Recent advances in sources of Bio- Inspiration and materials for robotics and actuators[J]

Yang Y , Ai C, Chen W, et al. Recent advances in sources of Bio- Inspiration and materials for robotics and actuators[J]. Small Methods, 2023, 7(9): 2300338

work page 2023

[4] [4]

Design method for a bionic wrist based on tensegrity structures[J]

Sun J, Cao X, Song G. Design method for a bionic wrist based on tensegrity structures[J]. Biosurface and Biotribology, 2020, 6(2): 31-36

work page 2020

[5] [5]

The Latest Research Progress on Bionic Artificial Hands: A Systematic Review[J]

Guo K, Lu J, Wu Y , et al. The Latest Research Progress on Bionic Artificial Hands: A Systematic Review[J]. Micromachines, 2024, 15(7): 891

work page 2024

[6] [6]

Mechatronic design of a robot for upper limb rehabilitation at home[J]

Curcio E M, Carbone G. Mechatronic design of a robot for upper limb rehabilitation at home[J]. Journal of Bionic Engineering, 2021, 18(4): 857-871

work page 2021

[7] [7]

A review on design of upper limb exoskeletons[J]

Gull M A, Bai S, Bak T. A review on design of upper limb exoskeletons[J]. Robotics, 2020, 9(1): 16

work page 2020

[8] [8]

Mechanical design and friction modelling of a cable-driven upper-limb exoskeleton[J]

Deˇzman M, Asfour T, Ude A, et al. Mechanical design and friction modelling of a cable-driven upper-limb exoskeleton[J]. Mechanism and Machine Theory, 2022, 171: 104746

work page 2022