Design of a Wrist Rehabilitation System with a Novel Mixed Structural Optimization Applying Improved Harmony Search

Round 1

Reviewer 1 Report

The paper presents the mechanical design of a wrist rehabilitation system. The design was optimized using a methodology called Improved Harmony Search.

In the introduction section, authors describe the impact of the stroke in the daily life of patients. Moreover, they present a brief state of the art related to control architecture applied to rehabilitation robots for post stroke rehabilitation.

The core of the paper is the mechanical design of the two-link mechanism which is a simple problem since robotics point of view. It is not clear what is the novelty of the design considering the final objective: stroke rehabilitation.

Authors based their works in a well know procedure proposed by reference [39] related to mechanical issues. However, they don’t take into account the regulatory issues related with medical robots in order to validate the selected materials. The reviewer considers this as a minor issue.

Authors focus their work on stroke rehabilitation; however they do not consider clinical criteria into mechanical design such as the patient’s forces or the potential spasticity of the patients.

Author Response

The paper presents the mechanical design of a wrist rehabilitation system. The design was optimized using a methodology called Improved Harmony Search. In the introduction section, authors describe the impact of the stroke in the daily life of patients. Moreover, they present a brief state of the art related to control architecture applied to rehabilitation robots for post stroke rehabilitation. The core of the paper is the mechanical design of the two-link mechanism which is a simple problem since robotics point of view.

1. It is not clear what is the novelty of the design considering the final objective: stroke rehabilitation.

The redaction in Abstract (lines 5-10) and in Section 1 (lines 61-62) was changed to clarify the point.

2. Authors based their works in a well know procedure proposed by reference [39] related to mechanical issues. However, they don’t take into account the regulatory issues related with medical robots in order to validate the selected materials. The reviewer considers this as a minor issue.

Paragraph in lines 102-108 was modified and reference 43 was included to explain appropriately the selection criteria of the manufacturing materials considered for this rehabilitation system.

3. Authors focus their work on stroke rehabilitation; however, they do not consider clinical criteria into mechanical design such as the patient’s forces or the potential spasticity of the patients.

The first paragraph in Subsection 2.1 (lines 69-73) was modified and references 41 and 42 were included to mention some clinical criteria consider for the mechanical design, specifically anthropomorphic data of the target population.

Reviewer 2 Report

This article describes a method for the development of a wrist rehabilitation system, with an approach based on an optimization problem solved by the Improved Harmony Search algorithm.

The description of the proposed method is fully detailed, after correctly reviewing previously published works.

Developing a wrist rehabilitation system can be considered relevant from different perspectives and it should not only be described in a journal like Applied Sciences, but also materialized and tested. Although I consider this paper interesting, from an engineering point of view, the implementation of the best results to be physically compared would be required. If experimental results were developed, they should be correctly shown and, if not, the reasons should appear in the paper.

Author Response

This article describes a method for the development of a wrist rehabilitation system, with an approach based on an optimization problem solved by the Improved Harmony Search algorithm. The description of the proposed method is fully detailed, after correctly reviewing previously published works.

1. Developing a wrist rehabilitation system can be considered relevant from different perspectives and it should not only be described in a journal like Applied Sciences, but also materialized and tested. Although I consider this paper interesting, from an engineering point of view, the implementation of the best results to be physically compared would be required. If experimental results were developed, they should be correctly shown and, if not, the reasons should appear in the paper.

As mentioned in Section 6 (lines 194-196), a future work in this project is the construction of the complete wrist rehabilitation system. However, results have been validated by the Finite Element Method, as explained in a new paragraph in Section 5 (lines 167-173) that also includes the new Fig. 4 to show the simulations that validate the quality of the obtained results.

Reviewer 3 Report

In this paper, an improved harmony search optimization method is proposed for the design of a wrist rehabilitation system. Generally, the paper is clearly written and the idea is easy to follow. But it still suffers from 1) insufficient representation, and 2) limited experiment. Authors are suggested to address the following concerns:

1. The novelty of this paper is not very clear. Although the authors stated the novelty in the Abstract, but I don’t think it is a really new idea or remarkably new finding. All optimization algorithms that can deal with constraints can make such statement. The real novelty or contribution of this paper should be further highlighted.
2. The motivation of adopting HS as the main optimization algorithm is unclear. In the literature, there are many state-of-the-art meta-heuristics/evolutionary algorithms that can handle constraints. Related work should be further reviewed.
3. The constraints handling method used in this paper are also not very clear. The author called it the Deb rule. But how to calculate the CVS? And why use a penalized value of 1000, rather than a larger value, e.g., 10^10?
4. In Page 7, the reason of the setting for tuning parameters is unclear. The authors should discuss in more detail how the algorithm parameters are computed/selected and their dependence on the trackled problem. The super-parameter settings for all tested algorithms need more discussions as regards how they affect the results and how sensitive are the results to these settings. The following papers suggest how to tune super-parameters: J. J. Wang and T. Kumbasar, “Parameter optimization of interval Type-2 fuzzy neural networks based on PSO and BBBC methods,” IEEE/CAA J. Autom. Sinica, vol. 6, no. 1, pp. 247-257, Jan. 2019; and "Dendritic neuron model with effective learning algorithms for classification, approximation and prediction," IEEE Transactions on Neural Networks and Learning Systems, 30(2), pp. 601 - 614, Feb. 2019. The methods regarding the parameter sensitive analysis in the above two references should be discussed and, if possible, used for finding a reasonable combination setting of the super-parameters.
5. There is no performance comparison. The proposed method should be compared with some traditional methods to show its effectiveness. Without performance comparison, readers are hard to know how excellent the proposed method is.
6. Minor revision: In Eq. (14), the ranges of [2810, 7870, 8000] seems to be wrong.

Author Response

In this paper, an improved harmony search optimization method is proposed for the design of a wrist rehabilitation system. Generally, the paper is clearly written and the idea is easy to follow. But it still suffers from 1) insufficient representation, and 2) limited experiment. Authors are suggested to address the following concerns:

1. The novelty of this paper is not very clear. Although the authors stated the novelty in the Abstract, but I don’t think it is a really new idea or remarkably new finding. All optimization algorithms that can deal with constraints can make such statement. The real novelty or contribution of this paper should be further highlighted.

The redaction in Abstract (lines 5-10) and in Section 1 (lines 61-62) was changed to clarify this point.

2. The motivation of adopting HS as the main optimization algorithm is unclear. In the literature, there are many state-of-the-art meta-heuristics/evolutionary algorithms that can handle constraints. Related work should be further reviewed.

HS was selected because the manuscript is intended for publication in a special number of Applied Sciences dedicated to developments with such algorithm. However, an additional metaheuristic (DE in its version rand/1/bin) was applied to solve the proposed problem, in order to contrast the performance of ImHS with another solution method. The results of this comparison were included in Section 5, in the paragraph lines 160-166 and in the new Table 5.

3. The constraints handling method used in this paper are also not very clear. The author called it the Deb rule. But how to calculate the CVS? And why use a penalized value of 1000, rather than a larger value, e.g., 10^10?

Text in lines 137-146 was modified and reference 47 was included to explain the selection of the constraint handling method and its use. The paragraph also explains the use and calculus of CVS. In respect to the penalty value of 1,000, it was selected considering the ranges of the variables expressions (16), where it is applied. Those expressions correspond to constraints related with the weight of the links, that are in the order of 10^-1.

4. In Page 7, the reason of the setting for tuning parameters is unclear. The authors should discuss in more detail how the algorithm parameters are computed/selected and their dependence on the trackled problem. The super-parameter settings for all tested algorithms need more discussions as regards how they affect the results and how sensitive are the results to these settings. The following papers suggest how to tune super-parameters: J. J. Wang and T. Kumbasar, “Parameter optimization of interval Type-2 fuzzy neural networks based on PSO and BBBC methods,” IEEE/CAA J. Autom. Sinica, vol. 6, no. 1, pp. 247-257, Jan. 2019; and "Dendritic neuron model with effective learning algorithms for classification, approximation and prediction," IEEE Transactions on Neural Networks and Learning Systems, 30(2), pp. 601 - 614, Feb. 2019. The methods regarding the parameter sensitive analysis in the above two references should be discussed and, if possible, used for finding a reasonable combination setting of the super-parameters.

In Subsection 4.2 (lines 149-152), it is explained that for this problem ImHS was tuned by trial and error, because the simplicity of that process. However, it is also mentioned that exist diverse strategies for a better tuning, and references 48 and 49 are included to support the point. Additionally, in Section 6 (lines 197-198) the implementation of a strategy for dynamic parameter tuning to improve the performance of ImHS is considered as a future work.

5. There is no performance comparison. The proposed method should be compared with some traditional methods to show its effectiveness. Without performance comparison, readers are hard to know how excellent the proposed method is.

As mentioned in point 2, an additional metaheuristic (DE in its version rand/1/bin) was applied to solve the proposed problem, in order to contrast the performance of ImHS with another solution method. The results of this comparison were included in Section 5, in the paragraph lines 160-166 and in the new Table 5.

6. Minor revision: In Eq. (14), the ranges of [2810, 7870, 8000] seems to be wrong.

The corresponding correction was carried out, to express that they are no ranges but discrete values.

Round 2

Reviewer 2 Report

Although I consider that the implementation of the best results to be physically compared would make this paper even more interesting, it is understandable that the construction of the complete wrist rehabilitation system will require a considerable amount of time.

Including the new Fig. 4 to show the simulations that validate the quality of the obtained results is a very positive change.

Author Response

Thank you very much for your revision. All your comments were properly addressed.

Reviewer 3 Report

Authors have addressed all the issues according to my previous comments. The related work has been enriched and the indistinct description as well as deficient analysis has been further refined. More discussions have also been added. This paper has been revised thoroughly to reach the standard for publication. Consequently, I advise to accept this paper.

Author Response

Thank you very much for your revision. All your comments were properly addressed.