A Bionic Control Method for Human–Exoskeleton Coupling Based on CPG Model

Round 1

Reviewer 1 Report

The paper focuses on the development of a control system for a human-exo coupled system. The control method is based on a CPG constituted by two neurons that generate oscillations tuned to behave like the torque signals requested by the coupled system.

 The paper covers different fields that have been presented in the introduction although there are relevant parts of the state of the art that are still missing and need to be properly discussed and compared with the proposed approaches.

In particular, the selection of the adopted CPG is not well discussed. There are other attempts in the literature to use dynamical coupled systems to develop CPGs for both humanoid robots [1] and lowe-limb exo control [2]. Since the CPG is one of the core elements of the proposed work, further discussions and comparisons should be provided, placing the work in a larger context [3]. Another relevant aspect that has been not addressed is the concept of embodiment. It is a relevant trend in current research also for exoskeletons [4] and the possibility to include embodied motor neurons [5] as well as control strategies that can shape the dynamical system behaviour [6].

The controller chosen for the motor is a PD. This choice should be argued (why not a PID) as well as the selection of the gain values proposed in Section 3.

In Fig 3 and also in the text the term rejection is used although the proper term commonly used are: excitatory and inhibitory connections.

 An interesting part of the work is related to the parameter identification. It is useful to provide further details on the adopted method, the search range and others.

The error performance obtained should be properly quantified using key performance indicators.    

The oscillations considered miss the typical slow-fast behaviour characteristic of the human stepping diagram. It is not clear why the authors adopted a static Experimental Procedure instead of using a typical acquisition on a treadmill or during free walking.

The desired trajectory reported in Eq 18 should be discussed. Is it a typical trajectory extracted by the acquired data? What will happen if the trajectory is modified, are the reported consideration still valid? It is important to provide quantitative results besides the reported figures.

Minor comments

Page 6 Line 192: the dot is missing on er

Page 9 Line 292-293: there are strange symbols instead of commas. Verify also along the text.

Fig. 5: labels are missing on the axes.

The font size in all figures should be not less than the caption font size to improve readability.

 

References

[1] G. Liu, M. Habib, K. Watanabe, and K. Izumi, “The Design of Central Pattern Generators Based on the Matsuoka Oscillator to Generate Rhythmic Human-Like Movement for Biped Robots,” J. Adv. Comput. Intell. Intell. Inform., Vol.11 No.8, pp. 946-955, 2007.

[2] Gomes MA, Siqueira AAG. Trajectory generation of exoskeleton for lower limbs using synchronized neural oscillators [Internet]. Anais. 2013 ;[citado 2023 jul. 21 ] Available from: http://abcm.org.br/anais/cobem/2013/PDF/191.pdf

[3] Dutta S, Parihar A, Khanna A, Gomez J, Chakraborty W, Jerry M, Grisafe B, Raychowdhury A, Datta S. Programmable coupled oscillators for synchronized locomotion. Nat Commun. 2019 Jul 24;10(1):3299. doi: 10.1038/s41467-019-11198-6. PMID: 31341167; PMCID: PMC6656780.

[4] R. L. Hybart and D. P. Ferris, "Embodiment for Robotic Lower-Limb Exoskeletons: A Narrative Review," in IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 657-668, 2023, doi: 10.1109/TNSRE.2022.3229563.

[5] P. Arena, L. Patanè and A. G. Spinosa, "A New Embodied Motor-Neuron Architecture," in IEEE Transactions on Control Systems Technology, vol. 30, no. 5, pp. 2212-2219, Sept. 2022, doi: 10.1109/TCST.2021.3131119.

 

[6] Arena, P., Patané, L., Spinosa, A.G. A nullcline-based control strategy for PWL-shaped oscillators (2019) Nonlinear Dynamics, 97 (2), pp. 1011-1033. DOI: 10.1007/s11071-019-05028-z 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents a bionic control for lower limb exoskeletal systems. The paper includes a mathematical formulation of the coupled human-robot dynamic model and an application of the proposed model in trajectory tracking performed by a healthy participant.

The paper turns out to be very interesting and control strategies that can exploit the behaviors of CPG systems and adaptation on the patient can have a great impact in robot-mediated rehabilitation. 

I suggest several major revisions to be made to the paper before proceeding with its publication.

- The 'introduction' section gives a very good understanding of the problem and the potential of CPGs. At the same time, the reviewer lacks a 'Related Works' section to better frame this contribution against the existing literature. In this way, it is possible to understand what the current limitations in the literature are about CPG-based controls and how this contribution manages to overcome them.

- The reviewer would better understand where the presentation of the approach proposed in this paper begins. In particular, the explanation of the proposed approach seems to begin at line 231. The expected benefit of using the proposed approach should then be better stressed.

- The reviewer suggests adding an "Experimental Validation" section before the "Results" section to explain in detail what the experimental setup and protocol followed to acquire data with an exoskeleton was. the technical details of the acquisition carried out are also missing: which exoskeleton, what data was acquired, at what frequency, etc.

- In the results section, mention of a genetic algorithm appears for the first time. It is not clear to the reviewer what genetic algorithm was used, for what purpose, and why it only appears in the discussion of the results. If a genetic algorithm is used in the approach presented, this should be described in detail in the materials and methods section.

- In line 335, did the authors mean to say "accurately" instead of "precisely"?

- in Figure 5, the axis labels are missing. I recommend then reporting summary metrics in the text to describe the performance of the system as a root mean squared error between the human and simulated pair to give an idea of the error. It is not sufficient to report a qualitative graph.

- Figures 7 and 8 should have the same scale on the y-axes to allow a quicker and more intuitive comparison of results. 

- The "Discussion" section should be called "Conclusions" and should include the future developments of this research.

The English of the manuscript is satisfactory.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors have performed a major revision of the original submission. Besides accommodating almost all remarks of this reviewer, the quality of the manuscript is significantly improved in terms of readability and technical soundness. This reviewer suggests improving the font size of the labels in Fig 6-10 to facilitate the readers.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is significantly improved from the initial version.

I again suggest that the authors take the explanation of the experimental validation out of the Results section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf