Next Article in Journal
Research on Outdoor Navigation of Intelligent Wheelchair Based on a Novel Layered Cost Map
Next Article in Special Issue
The Technical Development of a Prototype Lower-Limb Therapy Device for Bed-Resting Users
Previous Article in Journal
Dynamic Envelope Optimization of Articulated Vehicles Based on Multi-Axle Steering Control Strategies
Previous Article in Special Issue
Tug-of-War-Style High-Force Fluidic Actuation for Small Diameter Steerable Instruments
 
 
Article
Peer-Review Record

Machine Learning Models for Assistance from Soft Robotic Elbow Exoskeleton to Reduce Musculoskeletal Disorders

Actuators 2025, 14(2), 44; https://doi.org/10.3390/act14020044
by Sanjana Suresh, Inderjeet Singh and Muthu B. J. Wijesundara *
Reviewer 1: Anonymous
Reviewer 2: Anonymous
Reviewer 3:
Actuators 2025, 14(2), 44; https://doi.org/10.3390/act14020044
Submission received: 11 December 2024 / Revised: 7 January 2025 / Accepted: 17 January 2025 / Published: 22 January 2025

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper presents a machine learning approach to estimate pressure input for a soft robotic elbow exoskeleton, aiming to reduce musculoskeletal disorders in occupational workers. It establishes a relationship between pressure, weight, and bending angle using experimental data, evaluates various ML models for accuracy and efficiency, and proposes a novel KNN-Linear hybrid model that outperforms others in predictive accuracy and computational efficiency. This paper is well-structured and well-written. Before publication, there are some questions to be solved.

1. The author mentioned “People with physical disabilities such as limb weakness, constantly face challenges that can negatively impact their daily activities of living; therefore, the introduction of exoskeleton devices has been vastly researched for the consideration of making use of the assistive technology.”, more state-of-the-art can be cited: DOI: 10.34133/cbsystems.0141; DOI: 10.34133/cbsystems.0122; DOI: 10.34133/cbsystems.0115.

2.What were the specific criteria used to select the optimal polynomial degree in the feature transformation process?

3.Could you elaborate on the choice of 6 as the optimal polynomial degree and how it was determined?

4.How did you ensure that the KNN-Linear hybrid model was not overfitting, especially since it performed significantly better than other models?

5.What were the main challenges encountered during the data pre-processing stage, and how were they addressed? Can you provide more details on the experimental setup used for data collection, particularly regarding the range of motion and load conditions?

6.Can you discuss any potential limitations or assumptions made in your study that could affect the applicability of the proposed method in real-world scenarios?

Author Response

Please see the attached.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors of the manuscript submitted for review aim to reduce the load on the elbow joint using a pneumatically controlled soft exoskeleton. A model has been developed that estimates the pressure required to unload a certain percentage of the load and achieve the target bending angle. The authors solve the problem using different machine learning models, which are applied and compared to identify the most suitable model for predicting the required pressure under certain conditions. The machine learning of the models was done using a special experimental setup. The article is written in a good scientific style and is well structured. Therefore the article could be accepted after minor revision.

I have the following comments regarding the proposed manuscript:

1. The abbreviation IMU is not explained as to what it means.

2. Please also explain the meaning of the abbreviation EMG.

3. It would be better if other abbreviations such as KNN, SVM, and MYO were also clarified.

4. As is known, Inertial Measurement Unit  (IMU) sensors most often consist of a three-axis accelerometer, a three-axis gyroscope and a three-axis compass (magnetometer). Could you explain how you measure the elbow angle? If you use the accelerometer, you will have an error from the summation of the gravitational acceleration and the local acceleration. If you use the gyroscope, how do you eliminate the errors, the largest of which is the one from integration. The compass gives a direct angle, but even with it there are errors that must be taken into account. Please clarify this topic.

5. In my opinion, the conclusions would be more convincing if they presented some quantitative results?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

This paper explores the use of machine learning models to enhance the functionality of a soft robotic elbow exoskeleton, aiming to reduce musculoskeletal disorders among occupational workers. The exoskeleton is pneumatically actuated and designed to assist in elbow flexion and extension by predicting the required air pressure based on specific inputs like bending angle and the weight being supported.

1. The authors claimed that the device is for elbow exoskeleton, however, the maximum weight being tested is 1500kg, which is significantly lower that the required weight, as human arm is much heavier. 

2. The size of the exoskeleton is not specified in Figure 1. It is hard to justify how this can be used for occupational tasks.

3. There is no pressure sensor indicated in Figure 1.

4. There are some formatting errors in Figure 6 and 7, the y-axis label is inside the axis line.

5. Page 10, the "unseen data" should be outside the range of 70 -150kPa, as the validation data is in the range of data being collected in Figure 2.

 

Comments on the Quality of English Language

There are typos in the paper such as "Mult-Layer Perception"

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

This paper is ready for publication.

Back to TopTop