Utilizing Reinforcement Learning to Drive Redundant Constrained Cable-Driven Robots with Unknown Parameters

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper presented a reinforcement learning method for driving redundant constrained cable-driven robots with unknown parameters. Simulations and experiments were also performed to evaluate the performance of the proposed method. Overall, the paper was well written and has solid contribution. However, several minor issues should still be addressed to further improve the quality of this manuscript. Below are some comments for the author to consider:

1. In Equation (7), the subscript t should be explained. Does t represent the t-th step? If yes, the authors should provide a short explanation in the text.

2. The derivation process from Equation (14) to (15) should be explained. For example, what does E_\miu in Equation (14) mean? Is it an operator? If yes, the authors should also explain it in the text.

3. In Table 2, the value for the parameter T_s(s) is missing. Please address this issue.

4. The literature study part could be further extended. Currently, there are also many studies integrating redundant continuum structures into cable-driven robots to achieve motion control (see the references below). From this perspective, the authors are also recommended to mention those work in the Introduction section (e.g. in the first paragraph) as related background. Below are some studies on cable-driven continuum robots:

"Optimization of Stress Distribution in Tendon-Driven Continuum Robots Using Fish-Tail-Inspired Method". https://doi.org/10.1109/LRA.2022.3147456

"Development of a Multi-Cable-Driven Continuum Robot Controlled by Parallel Platforms". https://doi.org/10.1115/1.4048967

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Dear authors

Thank you for the research presented in the article. They are very useful for further analysis in the field of machine learning. 

However, it would be good if corrections/additions were made in order to enhance the content:

1. a general control algrotym is missing I mean some flowchart.

2. in addition to the parameters of the computer, please state the environment in which the tests were performed and which library you used.

3. please standardise the symbols in the equations up to 7, because from 8 and onwards there is no ‘bold’. please standardise the form of the equations.

4) Please underline in your conclusions whether you are considering other methods than the DDPG for further research?

5) Please enrich the literature 20 items is far too few, there are already several hundred articles on RL.

Best regards

Author Response

Please see the attachment.

Author Response File: Author Response.pdf