Mechatronic Design and Development of a Lower-Limb Exoskeleton System Based on Knee Joint Biomechanical Principles Using Electro-Pneumatic Actuation with an Embedded EMG Controller for Experimental Validation in Elderly Gait Rehabilitation Support

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents ExoKnee, a low-cost 3D-printed pneumatic knee exoskeleton controlled by EMG signals, designed to support gait rehabilitation for elderly stroke survivors, especially in low- and middle-income countries. The study is well grounded through biomechanical design, FEA, dynamic simulation in MATLAB, and expert validation using the Content Validity Coefficient. The paper is clearly written and addresses an important clinical and engineering challenge. Nevertheless, a few minor revisions are suggested to clarify certain technical parameters and strengthen the overall robustness of the manuscript before final publication.

1. Boundary Conditions and Loading in FEA (Section 2.1.3)

As noted by Expert 2 (Appendix), the 25 N knee joint force in FEA is relatively conservative for human interaction. Clarify simulation boundary conditions (support locations, fixation constraints) and briefly discuss structural reinforcements or alternative materials (beyond PLA+) for heavier patients to ensure safety.

2. Anthropometric Adaptability (Section 2.1.2 & Discussion)

Using ISO 7250-1:2017, the authors fixed the structural offset angle at 3.71° and linkage lengths at 173 mm/180 mm. Briefly explain how the design can be modularly scaled/adjusted to fit anatomical variations in elderly adults and prevent human–robot misalignment.

3. EMG Calibration in Clinical Settings (Section 2.1.4)

The fixed 500 ADC EMG activation threshold (based on an author’s MVC) is suitable for proof of concept, but stroke survivors often have muscle weakness/spasticity. Expand Future Work/Discussion to explain implementation of a normalized calibration protocol (e.g., adaptive thresholding via individual %MVC) in future trials.

Author Response

Comment 1: As noted by Expert 2 (Appendix), the 25 N knee joint force in FEA is relatively conservative for human interaction. Clarify simulation boundary conditions (support locations, fixation constraints) and briefly discuss structural reinforcements or alternative materials (beyond PLA+) for heavier patients to ensure safety.
Response 1: We thank the reviewer for this observation. The boundary conditions have been clarified in Section 2.1.3: the rotational knee joint was fixed at its upper mounting interface, simulating attachment to the quadriceps support, while the 25 N load was applied at the lower articular surface. For the calf support, fixed constraints were applied at the strap attachment points, with the 50 N pneumatic force applied at the central actuator interface. Additionally, a discussion of structural reinforcement options has been added, including increased infill density beyond 78%, internal ribbing, and alternative materials such as PETG or carbon fiber reinforced filaments for heavier patients. These adaptations are identified as future design directions pending biomechanical load characterization from clinical trials.

Comment 2: Using ISO 7250-1:2017, the authors fixed the structural offset angle at 3.71° and linkage lengths at 173 mm/180 mm. Briefly explain how the design can be modularly scaled/adjusted to fit anatomical variations in elderly adults and prevent human–robot misalignment.
Response 2: We agree with this comment. A clarification has been added in Section 2.1.2 explaining that the configurable linkage design allows for mechanical adjustment of the anchoring lengths r_thigh and r_leg to accommodate different user dimensions, and that the structural offset angle α can be adjusted during assembly to align with individual knee geometry. A formal sizing protocol covering the full range of elderly adult dimensions defined in ISO 7250-1:2017, including alignment tolerance analysis and interface pressure evaluation, is planned as part of future clinical development.

Comment 3: The fixed 500 ADC EMG activation threshold (based on an author's MVC) is suitable for proof of concept, but stroke survivors often have muscle weakness/spasticity. Expand Future Work/Discussion to explain implementation of a normalized calibration protocol (e.g., adaptive thresholding via individual %MVC) in future trials.
Response 3: We agree with this comment. The Discussion and Conclusions sections have been expanded to address this limitation explicitly. It is acknowledged that EMG signal characteristics differ substantially in elderly or post-stroke users due to muscle weakness, spasticity, co-contraction, and increased skin impedance. The single-subject calibration performed is therefore not generalizable to clinical populations and serves exclusively as a proof-of-concept demonstration. Future work will implement normalized calibration protocols based on individual MVC percentage (%MVC) and adaptive thresholding strategies, as stated in the revised Conclusions section.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript presents the design and preliminary development of a 1-DoF electro-pneumatic knee exoskeleton (“ExoKnee”) intended for elderly gait assistance, combining a 3D-printed mechanical structure, EMG-triggered actuation, finite element analysis, MATLAB/Simulink-based dynamic simulations, and an expert-based content validity assessment. The topic is relevant and potentially suitable for the journal, especially because affordable rehabilitation devices for low-resource settings are important. However, in its current form, the manuscript has substantial methodological and interpretive weaknesses. The main problem is the gap between the claims made in the title, abstract, and discussion, and the evidence actually provided in the paper. At present, the study demonstrates a proof-of-concept prototype with simulation, static structural checks, and expert opinion, but not convincing biomechanical or clinical validation of gait assistance in elderly users.

 

Major comments

1. The central claims are stronger than the evidence provided.
   The manuscript repeatedly frames ExoKnee as a system for “elderly gait assistance” and suggests meaningful rehabilitative functionality, yet the actual validation consists mainly of simulation, static FEA, and expert judgment. The only direct human-related element reported in the methods is an MVC calibration performed on one of the authors for EMG threshold setting. No gait trials, no quantitative human-subject biomechanical data, and no functional outcome measures are presented. Therefore, the manuscript should be rewritten much more cautiously as a proof-of-concept prototype study, unless actual human testing data are added.
2. The manuscript is inconsistent regarding human-subject testing.
   The introduction states that the prototype was tested “with a human subject,” yet the methods/results do not provide a proper human experimental protocol, participant demographics, inclusion criteria, task description, outcome variables, or statistical treatment. The conclusion mentions ethical approval and future clinical trials, which further suggests that no real clinical or functional human validation has yet been performed. This inconsistency must be resolved. If the only human involvement was EMG threshold calibration on one author, that should be stated transparently and should not be presented as device validation.
3. There is a mismatch between the claimed modeling framework and the reported methods.
   The manuscript states that the system integrates OpenSim modeling and MATLAB/Simulink validation, but the methods section only describes MATLAB R2024a / Simscape Multibody. I could not identify a real OpenSim workflow, model, or result in the presented methods/results. Please clarify whether OpenSim was truly used. If not, remove this claim from the introduction and contribution statements.
4. The dynamic model and equations require substantial correction and clarification.
   Equation (4) appears to duplicate Equation (3) instead of explicitly defining the gravitational term. This is a major technical issue because the mathematical model is one of the central contributions. In addition, several physical parameters in Table 2 are insufficiently justified. For example, the reported moving-segment mass of 0.18 kg appears inconsistent with the statement that the system represents the lower segment of the user’s leg plus exoskeleton structure; likewise, the equivalent inertia and viscous friction coefficient are not adequately explained. The authors must clearly derive all model terms and explain parameter identification.
5. The control concept is not internally coherent.
   The hardware implementation is described as a threshold-based ON/OFF EMG trigger driving a relay and solenoid valve, whereas the simulation section presents a closed-loop trajectory-tracking architecture with continuous reference-angle following under constant, stepwise, and sinusoidal inputs. These are not equivalent control paradigms. The manuscript currently mixes a binary trigger prototype with a continuous reference-tracking simulation without clearly explaining how the simulated controller maps to the real hardware. The authors should either:
   (a) explicitly separate “prototype hardware logic” from “future continuous-control simulation,” or
   (b) provide a unified control framework and experimental verification.
6. The quantitative validation is too weak for the conclusions drawn.
   The results section uses qualitative expressions such as “stable,” “smooth,” “acceptable settling time,” and “bounded tracking error,” but no actual tracking-error metrics are reported. There is no RMSE, maximum absolute error, rise time, overshoot, delay, pressure response, or experimental comparison between the model and the prototype. Similarly, the therapy routine reports a maximum torque of only 0.6 N·m, but this value is not contextualized against expected knee assistance requirements. Without stronger quantitative analysis, the claim of robust gait-assistive behavior is not yet supported.
7. The FEA is not sufficient to support claims of structural adequacy in real rehabilitation use.
   The structural analysis is limited to two components and two static load cases (25 N and 50 N). These loads appear low relative to realistic assisted movement, gait, sit-to-stand, misalignment effects, accidental loading, or repeated-use fatigue. The reported stresses and displacements are extremely small, which is unsurprising under such mild loading conditions. For a wearable rehabilitation device, the authors should justify the loading scenarios biomechanically and preferably include more realistic cases: dynamic loading, worst-case joint torque transmission, strap/interface loads, safety under unexpected events, and at least a discussion of fatigue and repeated cycles.
8. The expert-validation methodology is overinterpreted.
   The CVC analysis evaluates expert agreement about the questionnaire/instrument and the perceived suitability of the concept, not the real-world performance or therapeutic efficacy of the device. Yet the discussion treats the mean CVCtc value of 0.8747 almost as evidence that the device already demonstrates good functionality and therapeutic potential. This is too strong. In addition, several items in Table 3 are below 0.80, but these weaker scores are not critically discussed. The authors should reposition this section as a preliminary expert appraisal, not as device validation.
9. The discussion includes literature comparisons that are not evidence-based for ExoKnee itself.
   The manuscript compares ExoKnee with systems reporting improvements in gait speed, sit-to-stand time, metabolic cost, torque density, EMG unloading, and machine-learning accuracy, and then suggests that ExoKnee has a “competitive functional assistance profile.” However, none of these performance metrics were measured for ExoKnee in the present study. As written, the discussion overstates the study’s contribution. The literature comparison should be reframed as contextual, and the authors must clearly state that ExoKnee has not yet been evaluated using those functional outcomes.

Minor comments

1. There are multiple language and style issues throughout the manuscript that require thorough English editing. Examples include “Bio-desing Methodology,” “Expert Validation Assesment,” and “Exonee device.”
2. Figure readability should be improved. Several figures appear crowded and the axes/text in the simulation plots are too small for comfortable reading in journal format.
3. The title uses “knee-joint biomimicry,” but the paper does not yet convincingly demonstrate biomimetic knee kinematics beyond a simple single-axis flexion-extension architecture. This term should be justified better or softened.
4. The anthropometric adaptation is asserted repeatedly, but the manuscript does not provide a sufficiently detailed dimensional design framework, user size range, alignment protocol, or adjustment mechanism.
5. The actuator specification should be checked carefully for internal consistency, including reported stroke and nomenclature.
6. The manuscript should explicitly report total device mass, distribution of mass, and whether the pneumatic supply is portable or requires an external compressor, as this strongly affects practical usability for elderly users.
7. Safety is discussed mostly from a structural viewpoint. The manuscript should also address fail-safe behavior, unintended activation, valve failure, excessive extension, and emergency stop strategy.
8. Appendix material is extensive, but some of it would be more useful if summarized in the main text with clearer interpretation, especially for the expert questionnaire and programming logic.

Final assessment

The manuscript addresses an important problem and presents an interesting low-cost prototype with interdisciplinary design effort. However, the present version remains too preliminary for publication in its current form because the paper overclaims clinical and functional relevance relative to the evidence actually shown. A substantially revised manuscript would need to:
(1) reposition the contribution as proof-of-concept, or else add real human-subject functional data;
(2) correct and strengthen the mathematical/control sections;
(3) provide proper quantitative validation metrics;
(4) justify the FEA and biomechanical loading more rigorously; and
(5) tone down discussion and conclusions so that they match the available results.

Comments on the Quality of English Language

Minor comments

1. There are multiple language and style issues throughout the manuscript that require thorough English editing. Examples include “Bio-desing Methodology,” “Expert Validation Assesment,” and “Exonee device.”
2. Figure readability should be improved. Several figures appear crowded and the axes/text in the simulation plots are too small for comfortable reading in journal format.
3. The title uses “knee-joint biomimicry,” but the paper does not yet convincingly demonstrate biomimetic knee kinematics beyond a simple single-axis flexion-extension architecture. This term should be justified better or softened.

Author Response

Comment 1: The paper frames the work as a rehabilitation/gait-assistance study, while the evidence provided is primarily mechanical design, simulation, a prototype description, one-subject threshold calibration, and expert opinion. That gap between framing and evidence is the central academic weakness of the manuscript. The paper would be much stronger if it were positioned explicitly as an engineering proof-of-concept rather than as a validated rehabilitation technology.
Response 1: We fully agree with this observation, which we consider the most important revision in this manuscript. The paper has been comprehensively repositioned as an engineering proof-of-concept study throughout all sections. Specifically, the title, abstract, introduction, methods, discussion, and conclusions have been revised to consistently frame ExoKnee as a proof-of-concept device rather than a validated rehabilitation technology. Language implying clinical efficacy has been replaced with cautious, evidence-appropriate statements, and all comparisons with clinical literature are now explicitly framed as contextual rather than performance claims.

Comment 2: The knee is simplified to a single-axis hinge, despite the discussion later acknowledging the importance of instantaneous center of rotation tracking in human knees. This simplification may be acceptable for a first prototype, but it must be discussed as a major biomechanical limitation, especially for elderly or post-stroke users who are particularly sensitive to joint misalignment and interface discomfort.
Response 2: We agree. A dedicated paragraph has been added at the end of Section 2.1.1 explicitly acknowledging the single-axis hinge simplification as a major biomechanical limitation. The paragraph discusses how the human knee exhibits a moving instantaneous center of rotation that a fixed-axis mechanism cannot fully replicate, and identifies polycentric or cam-based mechanisms as important directions for future development.

Comment 3: Anthropometric adaptation is asserted more than demonstrated. The manuscript repeatedly claims anthropometric and ergonomic adaptation, but does not provide a sizing protocol, adjustability ranges, alignment tolerance analysis, or pressure/interface analysis at the attachment points. Without those data, 'anthropometrically adapted' remains a design intention, not a validated result.
Response 3: We agree with this comment. The term "anthropometrically adapted" has been replaced throughout the manuscript with "anthropometrically informed," which more accurately reflects the current state of the design. A clarification has been added in Section 2.1.2 explaining the modular adjustability of the anchoring lengths and offset angle, and explicitly acknowledging that formal sizing protocols, alignment tolerance analysis, and interface pressure evaluation are planned as future work.

Comment 4: A specification inconsistency appears around the actuator stroke. The text and figure caption describe the pneumatic actuator as having 100 mm stroke, but the actuator model designation shown is 'DSNU-S-12-50-…', which suggests a possible mismatch that must be clarified because it directly affects kinematics, achievable ROM, and simulation validity.
Response 4: We thank the reviewer for identifying this inconsistency. Upon verification against the FESTO product datasheet, the correct model designation for the 100 mm stroke actuator is DSNU-S-12-100-P-A-MQ. The model reference has been corrected throughout the manuscript including Section 2.1.1, Figure 3 caption, and Table 2. A clarifying note has been added explaining that the numeric value in the model designation refers to the stroke length in millimeters, confirming that the 100 mm stroke value used in the kinematic model and simulation is correct.

Comment 5: A wearable exoskeleton assisting human movement is subject to dynamic loading, inertial effects, off-axis moments, misalignment loads, strap-induced compression, cyclic loads, and accidental impacts. The chosen static loads of 25 N and 50 N appear too low to establish safety for a real human-contact rehabilitation device.
Response 5: We acknowledge this limitation. The FEA section has been revised to clarify that the selected load cases represent conservative static conditions appropriate for a proof-of-concept structural evaluation. A dedicated paragraph has been added acknowledging that dynamic loading scenarios, off-axis moments, cyclic fatigue analysis, and biomechanically representative loads derived from gait data are necessary for comprehensive structural validation, and these are identified as priorities for future work in the Conclusions section.

Comment 6: Additively manufactured polymer parts are highly sensitive to layer orientation, interlayer adhesion, humidity, creep, and cyclic degradation. Treating the material as if it were homogeneous isotropic stock material likely overestimates safety margins.
Response 6: We agree. A paragraph has been added in Section 2.1.3 explicitly acknowledging that the FEA assumed homogeneous isotropic material properties for PLA+, which represents a known simplification. The anisotropic behavior of fused deposition modeled parts due to layer orientation, interlayer adhesion variability, humidity sensitivity, and cyclic degradation is now discussed, and it is noted that the reported safety factors may overestimate actual structural margins. Future analyses incorporating orthotropic material models calibrated from tensile testing of printed specimens are identified as necessary next steps.

Comment 7: The EMG threshold was set after an MVC trial on one of the authors, and the threshold was fixed at 500 ADC units. This is not a valid basis for claims about elderly or post-stroke rehabilitation, where muscle amplitude, fatigue, spasticity, cocontraction, skin impedance, and signal quality differ substantially.
Response 7: We fully agree. The manuscript now explicitly acknowledges in Section 2.1.4 that the single-subject calibration performed on one of the authors is not generalizable to elderly or post-stroke populations, where muscle weakness, spasticity, co-contraction, and signal variability differ substantially. This limitation is transparently stated, and the calibration is framed exclusively as a proof-of-concept demonstration of the control logic. Future work will implement population-specific calibration protocols based on individual MVC percentage and adaptive thresholding strategies.

Comment 8: Several important cross-disciplinary literatures are missing, like AS&SPO: A Digital Signal Phase Adjustment Method for High-precision Spool Displacement Measurement in Electro-hydraulic Control Valves.
Response 8: We thank the reviewer for this suggestion. Following the journal's revision guidelines (point IV), the authors conducted a critical analysis of the suggested reference. As ExoKnee employs a pneumatic actuation system controlled by EMG threshold detection, the cited work pertains to electro-hydraulic precision valve control, which represents a fundamentally different actuation domain. Its inclusion would not directly enhance the technical content or contextualization of this manuscript. Therefore, the authors have opted not to include this reference. Should the reviewer provide further justification for its relevance to the pneumatic-EMG domain, the authors remain open to reconsideration.

Comment 9: The control is effectively binary ON/OFF, not assist-as-needed control. The device is activated when the filtered signal crosses threshold, then the relay may remain active for up to 5 seconds. That is a very coarse control architecture for gait assistance and may be unsuitable for phase-specific support, smooth torque delivery, or safe interruption.
Response 9: We agree with this observation. A paragraph has been added in Section 2.1.4 explicitly acknowledging that the binary ON/OFF control architecture does not support phase-specific torque modulation, smooth force delivery proportional to muscular effort, or safe mid-cycle interruption. These characteristics are recognized as important requirements for naturalistic gait assistance. The current implementation is now clearly framed as a proof-of-concept demonstration of EMG-triggered pneumatic activation only, and future control architectures incorporating proportional EMG-driven pressure modulation and gait phase detection are identified as necessary next steps.

Comment 10: The manuscript describes a 'closed-loop control architecture' with desired joint angle tracking, but the actual implemented hardware described in the prototype lacks a true closed sensory loop on knee angle. What is shown experimentally is threshold-triggered valve actuation, whereas the angle-tracking controller appears to belong to simulation. The paper needs a much clearer separation between simulated controller and implemented controller.
Response 10: We fully agree. A dedicated paragraph has been added at the beginning of the simulation description in Section 2.1.4 explicitly distinguishing between two control levels. At the hardware level, the fabricated prototype operates under a binary ON/OFF threshold-based strategy. At the simulation level, a closed-loop trajectory-tracking architecture was developed in MATLAB/Simulink to characterize the system's dynamic response under idealized conditions. The text now clearly states that the simulation was not designed to replicate the hardware controller, but rather to evaluate dynamic behavior as a foundation for future closed-loop implementation.

Comment 11: The reported peak knee torque of 0.6 N·m is extremely low for a device claiming functional gait or transitional assistance. If this value is correct, the assistive effect is likely negligible for real human use; if incorrect, the model/reporting needs correction. Either way, this must be resolved.
Response 11: We thank the reviewer for identifying this issue. The torque value has been corrected in the revised manuscript. The original simulation used an initial approximation of the moving segment mass (0.18 kg) that did not account for the anthropometrically representative lower leg mass. The simulation has been revised incorporating a segment mass of 3.85 kg derived from Plagenhoef et al. [42] and ISO 7250-1:2017 anthropometric data. The corrected simulation yields a joint torque ranging between approximately −3.5 N·m and +1.2 N·m throughout the therapy cycle. The peak assistive torque of 1.2 N·m is acknowledged to be constrained by the 12 mm bore diameter of the selected actuator, and this limitation is discussed explicitly in Section 3.1 along with recommendations for larger bore actuators in future iterations.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript describes “ExoKnee,” a single-DOF, 3D-printed knee exoskeleton actuated pneumatically and triggered by quadriceps EMG, with design support from CAD/FEA, analytical modeling, and Simulink/Simscape simulations. Its main claimed contribution is an accessible, locally manufacturable rehabilitation device for elderly users in low- and middle-income settings, complemented by an expert-questionnaire-based validation reporting a mean CVC of 0.8747.

Here are my comments:

The paper frames the work as a rehabilitation/gait-assistance study, while the evidence provided is primarily mechanical design, simulation, a prototype description, one-subject threshold calibration, and expert opinion. That gap between framing and evidence is the central academic weakness of the manuscript. The paper would be much stronger if it were positioned explicitly as an engineering proof-of-concept rather than as a validated rehabilitation technology.

The knee is simplified to a single-axis hinge, despite the discussion later acknowledging the importance of instantaneous center of rotation tracking in human knees. This simplification may be acceptable for a first prototype, but it must be discussed as a major biomechanical limitation, especially for elderly or post-stroke users who are particularly sensitive to joint misalignment and interface discomfort.

Anthropometric adaptation is asserted more than demonstrated. The manuscript repeatedly claims anthropometric and ergonomic adaptation, but does not provide a sizing protocol, adjustability ranges, alignment tolerance analysis, or pressure/interface analysis at the attachment points. Without those data, “anthropometrically adapted” remains a design intention, not a validated result.

A specification inconsistency appears around the actuator stroke. The text and figure caption describe the pneumatic actuator as having 100 mm stroke, but the actuator model designation shown is “DSNU-S-12-50-…”, which suggests a possible mismatch that must be clarified because it directly affects kinematics, achievable ROM, and simulation validity.

Load cases are likely non-conservative. A wearable exoskeleton assisting human movement is subject to dynamic loading, inertial effects, off-axis moments, misalignment loads, strap-induced compression, cyclic loads, and accidental impacts. The chosen static loads of 25 N and 50 N appear too low to establish safety for a real human-contact rehabilitation device.

Additively manufactured polymer parts are highly sensitive to layer orientation, interlayer adhesion, humidity, creep, and cyclic degradation. Treating the material as if it were homogeneous isotropic stock material likely overestimates safety margins.

Single-subject calibration on one author is inadequate. The EMG threshold was set after an MVC trial on one of the authors, and the threshold was fixed at 500 ADC units. This is not a valid basis for claims about elderly or post-stroke rehabilitation, where muscle amplitude, fatigue, spasticity, cocontraction, skin impedance, and signal quality differ substantially.

Several important cross-disciplinary literatures are missing, like AS&SPO: A Digital Signal Phase Adjustment Method for High-precision Spool Displacement Measurement in Electro-hydraulic Control Valves.

The control is effectively binary ON/OFF, not assist-as-needed control. The device is activated when the filtered signal crosses threshold, then the relay may remain active for up to 5 seconds. That is a very coarse control architecture for gait assistance and may be unsuitable for phase-specific support, smooth torque delivery, or safe interruption.

The manuscript describes a “closed-loop control architecture” with desired joint angle tracking, but the actual implemented hardware described in the prototype lacks a true closed sensory loop on knee angle. What is shown experimentally is threshold-triggered valve actuation, whereas the angle-tracking controller appears to belong to simulation. The paper needs a much clearer separation between simulated controller and implemented controller.

The reported peak knee torque of 0.6 N·m is extremely low for a device claiming functional gait or transitional assistance. If this value is correct, the assistive effect is likely negligible for real human use; if incorrect, the model/reporting needs correction. Either way, this must be resolved.

Author Response

Comment 1: The central claims are stronger than the evidence provided. The manuscript repeatedly frames ExoKnee as a system for 'elderly gait assistance' and suggests meaningful rehabilitative functionality, yet the actual validation consists mainly of simulation, static FEA, and expert judgment. No gait trials, no quantitative human-subject biomechanical data, and no functional outcome measures are presented. The manuscript should be rewritten much more cautiously as a proof-of-concept prototype study, unless actual human testing data are added.
Response 1: We fully agree with this central observation. The manuscript has been comprehensively revised to consistently position ExoKnee as a proof-of-concept prototype study throughout all sections, including the title, abstract, introduction, methods, discussion, and conclusions. Language implying clinical efficacy or validated rehabilitation functionality has been replaced with cautious, evidence-appropriate statements. The abstract now explicitly states "proof-of-concept device" and all claims regarding gait assistance are framed as design intentions pending clinical validation rather than demonstrated outcomes.

Comment 2: The introduction states that the prototype was tested 'with a human subject,' yet the methods/results do not provide a proper human experimental protocol, participant demographics, inclusion criteria, task description, outcome variables, or statistical treatment. The conclusion mentions ethical approval and future clinical trials, which further suggests that no real clinical or functional human validation has yet been performed. This inconsistency must be resolved.
Response 2: We agree and apologize for this inconsistency. The phrase "tested with a human subject" has been removed from the introduction. The revised text now transparently states that the only human involvement in this study was a single-subject MVC calibration performed on one of the authors for the purpose of setting the EMG activation threshold, and that this does not constitute clinical or functional validation. The conclusion has been revised accordingly to clarify that clinical trials with elderly patients and post-stroke survivors are planned as future work following ethical approval.

Comment 3: The manuscript states that the system integrates OpenSim modeling and MATLAB/Simulink validation, but the methods section only describes MATLAB R2024a / Simscape Multibody. I could not identify a real OpenSim workflow, model, or result in the presented methods/results. Please clarify whether OpenSim was truly used. If not, remove this claim from the introduction and contribution statements.
Response 3: We thank the reviewer for identifying this inaccuracy. OpenSim was not used in the present study. The reference to OpenSim modeling has been removed from the introduction and all contribution statements. The revised manuscript accurately states that dynamic modeling and simulation were conducted exclusively using MATLAB R2024a with the Simscape Multibody toolbox. We apologize for this oversight, which originated from a previous version of the manuscript developed for a different submission.

Comment 4: Equation (4) appears to duplicate Equation (3) instead of explicitly defining the gravitational term. In addition, several physical parameters in Table 2 are insufficiently justified. The reported moving-segment mass of 0.18 kg appears inconsistent with the statement that the system represents the lower segment of the user's leg plus exoskeleton structure; likewise, the equivalent inertia and viscous friction coefficient are not adequately explained.
Response 4: We thank the reviewer for identifying these issues. Equation (4) has been corrected to explicitly define the gravitational torque term as τ_grav(θ_k) = M_sys · g · d_cm · sin(θ_k), which was inadvertently duplicated from Equation (3) in the previous version. Regarding Table 2, all parameters have been justified in the revised manuscript: r_thigh and r_leg were derived from ISO 7250-1:2017; α was measured experimentally from the fabricated prototype; M_sys has been corrected to 3.85 kg based on Plagenhoef et al. [42] anthropometric data representing the combined lower leg and exoskeleton mass; d_cm was estimated from CAD-based mass distribution analysis; I_eq was estimated using a first-order rigid segment model; B was treated as a variable parameter consistent with FESTO actuator design philosophy; and D_p, S, and P_nom were obtained directly from the FESTO datasheet.

Comment 5: The hardware implementation is described as a threshold-based ON/OFF EMG trigger driving a relay and solenoid valve, whereas the simulation section presents a closed-loop trajectory-tracking architecture with continuous reference-angle following. These are not equivalent control paradigms. The manuscript currently mixes a binary trigger prototype with a continuous reference-tracking simulation without clearly explaining how the simulated controller maps to the real hardware.
Response 5: We fully agree. A dedicated paragraph has been added at the beginning of the simulation description in Section 2.1.4 explicitly distinguishing between two control levels present in this work. At the hardware level, the fabricated prototype operates under a binary ON/OFF threshold-based strategy where EMG signal detection triggers relay activation. At the simulation level, a closed-loop trajectory-tracking architecture was developed in MATLAB/Simulink to characterize the system's dynamic response under idealized continuous reference inputs. The revised text clearly states that these are not equivalent paradigms and that the simulation serves as a foundation for future closed-loop control implementation rather than a representation of the current hardware logic.

Comment 6: The results section uses qualitative expressions such as 'stable,' 'smooth,' 'acceptable settling time,' and 'bounded tracking error,' but no actual tracking-error metrics are reported. There is no RMSE, maximum absolute error, rise time, overshoot, delay, pressure response, or experimental comparison between the model and the prototype.
Response 6: We agree with this comment. The qualitative descriptions have been replaced with quantitative metrics in the revised manuscript. For Case (i), the system maintained a constant angular position of 0° with rod displacement oscillations below 10⁻⁶ mm. For Case (ii), the system tracked each angular setpoint transition within the prescribed 2.5 s intervals with no observable overshoot, and rod displacement decreased from 100 mm to 0 mm across successive steps. For Case (iii), the system followed a sinusoidal trajectory with angular displacement ranging between approximately −65° and −15° at a period of 0.6 s. Additionally, a preliminary comparison between simulated and physical prototype behavior has been added: the simulation predicted a ROM of 90° while the fabricated prototype achieved 66° (from 175° to 109°), representing a deviation of 26.7%, with actuation time of approximately 3 to 5 seconds consistent with the simulation intervals.

Comment 7: The structural analysis is limited to two components and two static load cases (25 N and 50 N). These loads appear low relative to realistic assisted movement, gait, sit-to-stand, misalignment effects, accidental loading, or repeated-use fatigue.
Response 7: We acknowledge this limitation. The FEA section has been revised to clarify the boundary conditions and justify the selected load cases as conservative static conditions appropriate for the proof-of-concept scope of this work. A paragraph has been added acknowledging that dynamic loading scenarios, off-axis moments, cyclic fatigue analysis, and biomechanically representative loads derived from gait data are necessary for comprehensive structural validation. These are explicitly identified as priorities for future work in the revised Conclusions section.

Comment 8: The CVC analysis evaluates expert agreement about the questionnaire/instrument and the perceived suitability of the concept, not the real-world performance or therapeutic efficacy of the device. Yet the discussion treats the mean CVCtc value of 0.8747 almost as evidence that the device already demonstrates good functionality and therapeutic potential. Several items in Table 3 are below 0.80, but these weaker scores are not critically discussed.
Response 8: We agree with this important methodological observation. The Discussion section has been revised to reposition the CVC analysis as a preliminary expert appraisal of the conceptual design's coherence, clarity, and relevance, explicitly clarifying that it does not constitute empirical measurement of device performance or therapeutic efficacy. Additionally, the three items below the 0.80 threshold (item 2: CVCtc = 0.7330; item 3: CVCtc = 0.7163; item 8: CVCtc = 0.7497) are now critically discussed, and their lower scores are contextualized as consistent with the acknowledged limitations regarding FEA validation, anthropometric sizing protocols, and clinical gait trials.

Comment 9: The manuscript compares ExoKnee with systems reporting improvements in gait speed, sit-to-stand time, metabolic cost, torque density, EMG unloading, and machine-learning accuracy, and then suggests that ExoKnee has a 'competitive functional assistance profile.' However, none of these performance metrics were measured for ExoKnee in the present study.
Response 9: We agree. The Discussion section has been revised to reframe all literature comparisons as contextual rather than performance claims. A paragraph has been added explicitly stating that the functional metrics reported in cited studies have not been evaluated for ExoKnee in the present work, and that these comparisons are presented solely to situate the design within the broader landscape of assistive knee devices. The phrase "competitive functional assistance profile" has been removed and replaced with language appropriate to a proof-of-concept study pending clinical validation.

Comment 10: There are multiple language and style issues throughout the manuscript that require thorough English editing. Examples include 'Bio-desing Methodology,' 'Expert Validation Assesment,' and 'Exonee device.' Figure readability should be improved. Several figures appear crowded and the axes/text in the simulation plots are too small for comfortable reading in journal format.
Response 10: We thank the reviewer for identifying these issues. All typographical errors have been corrected throughout the manuscript, including "Bio-design Methodology," "Expert Validation Assessment," and "ExoKnee device." Figure 7 has been redesigned with a professional layout using a 2×3 tiledlayout format, larger font sizes (11pt), clearly labeled axes with units, and exported at 600 DPI resolution for journal-quality readability. Figure 10 has also been updated to reflect the corrected simulation results.

Comment 11: The title uses 'knee-joint biomimicry,' but the paper does not yet convincingly demonstrate biomimetic knee kinematics beyond a simple single-axis flexion-extension architecture. The anthropometric adaptation is asserted repeatedly, but the manuscript does not provide a sufficiently detailed dimensional design framework, user size range, alignment protocol, or adjustment mechanism.
Response 11: We agree with both observations. The term "biomimicry" in the title has been replaced with "biomechanical principles," which more accurately reflects the current design stage. Regarding anthropometric adaptation, the term "anthropometrically adapted" has been replaced throughout with "anthropometrically informed," and a clarification has been added in Section 2.1.2 explaining the modular adjustability of the anchoring lengths and offset angle. A formal sizing protocol, alignment tolerance analysis, and interface pressure evaluation are acknowledged as necessary steps for clinical deployment and are planned as future work.

Comment 12: The manuscript should explicitly report total device mass, distribution of mass, and whether the pneumatic supply is portable or requires an external compressor, as this strongly affects practical usability for elderly users. Safety is discussed mostly from a structural viewpoint. The manuscript should also address fail-safe behavior, unintended activation, valve failure, excessive extension, and emergency stop strategy.
Response 12: We agree. The revised manuscript now explicitly reports that the 3D-printed structural components have a total mass of approximately 0.18 kg, with the FESTO pneumatic actuator adding approximately 0.1 kg, resulting in a lightweight wearable assembly. It is also transparently acknowledged that the current configuration requires an external compressed air supply at 6 bar, which limits immediate portability, and that miniaturized pneumatic supply solutions will be evaluated in future iterations. Regarding safety, the revised text now addresses fail-safe valve configurations, emergency stop mechanisms, software-based extension limits, and unintended activation prevention as important design requirements for future clinical versions of the device.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors improved the paper.

Reviewer 3 Report

Comments and Suggestions for Authors

ok