Trends in the Control of Hexapod Robots: A Survey

Round 1

Reviewer 1 Report

This study is about overview of hexapod robot control. The authors intend to overview the trends in this field of research and determine in which stage is the design of autonomous and adaptable controllers for hexapods. The paper is well written and but there are minor errors as follows.

You can add the number of each item in each percentage for better understanding. (Fig.1 Fig.2 etc.)

The format of Table 4 is wrong.

P15 Fig.1 label is wrong.

P15 Fig.1 RRT is motion planning algorithm and LiDAR is LiDAR is not a vision sensor.

Author Response

Thank you for your kind suggestions on the manuscript. We modified the relevant part of our manuscript according to your advice. Each of your questions was answered as follows.

Point 1: You can add the number of each item in each percentage for better understanding. (Fig.1 Fig.2 etc.)

Response 1: We followed your suggestion and included the percentage in all items of Figures 1 to 7 (Pages 11-17). We also took the opportunity to revise the data and update the values throughout the sections 4 and 5 (Pages 11-15).

Point 2: The format of Table 4 is wrong.

Response 2: Following your comment, we updated the format of Table 4 (Page 11). Considering the same issue, we also revised the format of Table 3 (Pages 7-8).

Point 3: P15 Fig.1 label is wrong.

Response 3: Following your comment, we changed the label «Fig. 1», to «Fig. 4» (Page 14).

Point 4: P15 Fig.1 RRT is motion planning algorithm and LiDAR is LiDAR is not a vision sensor.

Response 4: We updated the data from Fig. 7(b) and verified the percentages, according to your comment (Page 17).

Reviewer 2 Report

The authors present a systematic review of modeling and control methods applied to hexapod (6-legged) robots, with a particular focus on control for adaptive and autonomous behaviors. Overall, the literature search is well structured and focused and seems to provide a thorough summary of the relevant work. My major comment to the authors is to encourage them to identify additional trends in the reviewed literature, in addition to the categorization, so as to meet their goals of showing where the field is heading (see detailed comments for more information).

Major comments:

My major concern is that the analysis the authors present, which shows the percentages in each of the categories they break papers into, does not sufficiently achieve their goal of showing trends in the field of hexapod control. Ideally, I would like the discussion to add something past showing the data from the tables in a new format, perhaps showing changes in the control approaches over time or discussing how some of these metrics may influence each other (i.e. are learning methods more often seen in simulation or experimentally?). The authors discuss something along these lines briefly when talking about the use of computer vision in some controllers, but I would encourage the authors to draw other relevant conclusions from the literature review, where appropriate and supported by the data. Some specific suggestions of how the discussion section and the presented results might be strengthened by some additional analysis (note I do not think it is necessary to include all of these, but a subset would likely strengthen the paper):

• Show trends over time – show whether certain approaches are becoming more popular over the course of the last 10 years
• Show cross metrics of (1) environment and controller, (2) sensor and controller, (3) simulation/experimental choice and controller type.
• Adding any metrics on the success of the implementations, if these are consistently reported.

I would like to see some short clarification of the “inherent static stability” assumption one of the first times it is mentioned. Since significant later discussion is spent on controllers that calculate SSM and DGSM, adding a brief description of why (or when) hexapods are considered inherently statically stable would be useful.  

Clarification of what is meant by uneven terrain environments, what is the range of obstacles seen and their sizes relative to the hexapod body lengths. Similarly, does outdoor environments only cover material changing or are some outdoor environments also uneven terrain?

The dynamics section suggests that these models are distinguished by their use of external force and torque measurements, allowing for a fuller sense of the environmental effects. However, references [11-14] use contact force measurements in their models. Is the proper breakdown in categories kinematic versus dynamic, or is it quasi-static versus dynamic? If it is kinematic versus dynamic, I’d encourage the authors to revise their description at the top of section 3.1.2 to better distinguish what makes the dynamic models different.

Minor comments:

I would change the heading for 3.3 from “RL” to “Reinforcement Learning”

The English grammar and spelling are very good overall. Here are a few minor typos/grammar suggestions:

• Line 33-34: “able of navigating” -> “capable of navigating” or “able to navigate”
• Line 90: “resumed” should be “presumed”?
• Line 182: “meant to humanitarian demining” -> “meant for humanitarian demining” or “meant to aid in humanitarian demining”
• Line 202-203: unsure what is meant by “a layer identification of tracking method”
• Line 245: “planning locomotion algorithm” -> “locomotion planning algorithm”
• Line 309: “in a smooth surface” -> “on a smooth surface”
• Line 330: “able climb ramps” -> “able to climb ramps”
• Line 450-451: “and how the state-of-the-art is concerned with this issue.” Awkward phrasing, I would consider rewording.
• Line 486: “considered in its early beginnings” Awkward, consider rewording.
• Line 487: “takes the advantage of using the gathered” -> “takes advantage of the gathered”
• Line 498: missing reference

Please check the style and information in the references for errors and missing items, making sure to match the suggested style. To point to a few spots, there are some missing data in the following citations:

• [8], [25], [55] List all authors
• [22] Update with ICRA 2020 citation information
• [35] Missing “no.”
• [37] Missing journal and issue information

Author Response

Thank you for your kind suggestions on the manuscript. We modified the relevant part of our manuscript according to your advice. Each of your questions was answered as follows.

Point 1: Show trends over time – show whether certain approaches are becoming more popular over the course of the last 10 years

Response 1: We followed your suggestion and included Figure 1, which portrays the not only the increasing number of publications regarding the design and control of hexapods, but also the rising interest in bio-inspired controllers and learning methods particularly in the past five years (Page 11).

Point 2: Show cross metrics of (1) environment and controller, (2) sensor and controller, (3) simulation/experimental choice and controller type.

Response 2: Following your comment, we included cross metrics of environment and controller, simulation/experimental choice and controller type, and environment and traditional controller, in Figures 3(b), 4(b), and 6, and updated the discussion presented in Pages 13-16 accordingly. Since we could not find any significant conclusion regarding the cross metric of sensors and controllers, we chose not to include it in our modifications.

Point 3: Adding any metrics on the success of the implementations, if these are consistently reported.

Response 3: Following your comment, we revised all the publications presented in our manuscript. However, the majority did not discuss any metrics on their success.

Point 4: I would like to see some short clarification of the “inherent static stability” assumption one of the first times it is mentioned. Since significant later discussion is spent on controllers that calculate SSM and DGSM, adding a brief description of why (or when) hexapods are considered inherently statically stable would be useful.  

Response 4: According to your comment, we added a brief clarification of the inherent static stability in section 1, lines 35-38. We also took the opportunity to improve the description of the SSM (lines 148-165) and include a comparison between the SSM and the DGSM and the ZMP (lines 198-224).

Point 5: Clarification of what is meant by uneven terrain environments, what is the range of obstacles seen and their sizes relative to the hexapod body lengths. Similarly, does outdoor environments only cover material changing or are some outdoor environments also uneven terrain?

Response 5: We updated the description at the top of section 3 (lines 97-105=, according to your comment.

Point 6: The dynamics section suggests that these models are distinguished by their use of external force and torque measurements, allowing for a fuller sense of the environmental effects. However, references [11-14] use contact force measurements in their models. Is the proper breakdown in categories kinematic versus dynamic, or is it quasi-static versus dynamic? If it is kinematic versus dynamic, I’d encourage the authors to revise their description at the top of section 3.1.2 to better distinguish what makes the dynamic models different.

Response 6: Following your comment, we revised and updated the description at the top of section 3.1.2 (Page 5).

Point 7: I would change the heading for 3.3 from “RL” to “Reinforcement Learning”. The English grammar and spelling are very good overall. Here are a few minor typos/grammar suggestions: Line 33-34: “able of navigating” -> “capable of navigating” or “able to navigate”; Line 90: “resumed” should be “presumed”?; Line 182: “meant to humanitarian demining” -> “meant for humanitarian demining” or “meant to aid in humanitarian demining”; Line 202-203: unsure what is meant by “a layer identification of tracking method”; Line 245: “planning locomotion algorithm” -> “locomotion planning algorithm”; Line 309: “in a smooth surface” -> “on a smooth surface”; Line 330: “able climb ramps” -> “able to climb ramps”; Line 450-451: “and how the state-of-the-art is concerned with this issue.” Awkward phrasing, I would consider rewording. Line 486: “considered in its early beginnings” Awkward, consider rewording. Line 487: “takes the advantage of using the gathered” -> “takes advantage of the gathered”; Line 498: missing reference. Please check the style and information in the references for errors and missing items, making sure to match the suggested style. To point to a few spots, there are some missing data in the following citations: [8], [25], [55] List all authors, [22] Update with ICRA 2020 citation information, [35] Missing “no.”, [37] Missing journal and issue information.

Response 7: We followed your suggestions and applied your corrections.

Round 2

Reviewer 2 Report

I want to thank the authors for their response to the review comments. In my opinion the work has been significantly improved and I think the analysis and discussion of the trends in the field show interesting angles on where the literature is pointed.

One minor grammar comment - line 521, I would suggest a different word than "novelty" if I'm interpreting this sentence correctly. Novelty carries a strong connotation of "originality" when I believe the authors were going more for "new" or "recent".