Research on the Posture Control Method of Hexapod Robot for Rugged Terrain
Round 1
Reviewer 1 Report
Title: Research on the posture control method of hexapod robot for rugged terrain
- Presentation of the paper
This paper presents an algorithm for a hexapod robot to walk on stair and sand. An adjustment on pose and gravity center position is presented. A simulation is presented with some practical results inside a lab environment. The results are very well detailed, clear and seems accurate but still needs some improvements.
The contribution of this research works is not well described in the introduction. As hexapod walk and walking pattern is widely published with many stabilisation algorithms, it is difficult to make the difference with previous works whenever the authors suggest the walking stability is improved in this research works.
- General comments (review)
Presentation of Fig. 1 should be improved. It's difficult to understand the arrows in the plane XOZ. Should you add one leg in the figure for example R2 in fig 3?
The control process diagram including compensation should be presented in a separate figure. The algorithm should be also presented, such as presented in section 2.2.
Please, should you enumerate all your algorithms in the paper? You may use LaTex example with the package Algorithm.
Some text editing should be done in sections 2.2 and other sections for the equations inside the text.
In section 3, line 165, what is he stability margin value (phase and gain margin) ? as this is a control theory formulation, you should add this in your results, in particular with high perturbation, such as pushing hexapod.
In section 3, level and height of the torso is not presented in any result.
In Fig. 3 you have a translating Spring-Damping elements for each angle in rotation. Why this is in translation? Do you want to control damping and spring values in rotation as suggested in (8) ? How the values are changed? Current cars are doing such a thing with his suspension varying in function of the ground type. I do not see any equation related to the value adaptation.
In eq. (8) and (12), translation acceleration and angular velocity variation are pretty difficult to use in a feedback loop for a compensation. It's much more destabilising the motion of the robot. How did you implement this? I think somethings is missing after eq. (12).
In Fig. 8, how control feedback is done? are you using a PID, H, state space, or other command ?
In eq. (13), how H is measured? do you have an external MoCap ? Should you add results for H ?
Fig 11, 12 and 13 show an improvement on stability, but VSDM should be compared to another method since the controller should also be improved using recent methods. I would suggest to use the same stabilizing algo used in a drone in a feedback control loop.
Quantification of stability should be presented in order to understand how much VSDM is efficient.
Should you add a statistic analysis such as ANOVA with a p-value for Fig. 11, 12, 13 ? You should compare another method by comparing result with an ANOVA.
Fig. 11, 12 and 13 are well presented, but Fig 8 seems to be done by another person and is not consistent with other fig.
Are the difference between blue and red lines significant? what should be a threshold value to be significant?
Grammar:
- line 124: somethings is missing: 'be converted The trajectory'
- line 185: somethings is missing: component in [ , , ] x y z 185 F = F F F For straight
Author Response
Please see the attachment
Author Response File: 
Author Response.docx
Reviewer 2 Report
The paper is focused on the posture control of an hexapod robot walking on rugged or heavy rugged terrain, describing different approaches according to the control goal: trajectory planning, force control, virtual suspension, and geometric methods (master polygon, support polygon). The developed methods rely on the geometric parameters of the terrain, identifying a set of gait patterns for the legs that depend on the size of the obstacles. The paper also considers the balance of forces in the body, proposing a suspension model assimilated to a mass-spring-damper system. Some of the methods are validated in simulation and real experiments.
GENERAL COMMENTS
The main contribution of the paper is the description of different control methods for an hexapod robot walking on a rugged terrain, which provides the reader a broad understanding of the problems of this kind of robots. However, the paper fails in the presentation of the content in three aspects:
1) Modelling: authors should start by defining properly the kinematic model of the robot with the reference frames, joint angles, link lengths... Using a 3D model (rendered view) of the robot within an illustrative scenario becomes really useful for this purpose as it provides a clean view for the readers. Otherwise, it is very hard to understand Figure 1 and subsequents.
2) Notation: related to previous point, there is a huge number of geometric parameters described by words instead of being represented graphically, becoming tedious to track their meaning. Also the subscripts are tiny. Use math symbols in word. Define and represent the coordinate systems according to their physical meaning. For example, X0Y0Z0 are drawn randomly in Figure 2.
3) Coherence: the different methods are presented in raw, without any link between them. It would be convenient to expalin at the begining the general control scheme, identifying the methods developed in the different sections.
English requires significant improvements to facilitate the understanding of the text.
PARTICULAR COMMENTS
1. Consider adding a sequence of images of the hexapod robot walking along with the diagram in Figure 1 to facilitate the understanding of the trajectory generation.
2. The format of the text must be improved (line 149 and others). Again, the subscripts are tiny.
3. Eq(6), I guess A^-1 is pseudoinverse, correct? Indicate clearly.
4. It is not clear to me how Eq. (8) becomes Eq. (9). Scalar and matrix notation are identical. In Eq. (11), how can K/M be divided if they are vectors?
5. The geometric parameters in Section 4 should be graphically represented.
6. In the simulation and experimental results, author should describe first the robot/model used: dimensions of body and links, joint limits and speed, picture of the robot...
7. Authors should provide graphical results showing the XYZ trajectory of the robot, or at least the evolution of the joints position during the walking.
8. The data plots should be represented in time (seconds) not in samples.
9. X0-Y0 axes in Figure 3 are in incorrect order.
Author Response
Please see the attachment
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors have added significant content to the paper improving the contribution and my understanding. I have however some remarks to discuss with the authors in order to be published.
1-For the fig. 1, you will need to add your source for the insect presented and permission fi it's not an open picture. Please add adequate citation and add insect name and reason if this one. Cockroach is largely cited as a stable insect and could be used to explain hexapod walking motion.
2- line 191, shoud be: Figure 6. Two diagrams...
3- Fig. 16 should be compared with another method. For example, when it's written without VSDM controller, you should compare with another common compensation already publised. As stated, when you compare between with and without, it's not really an improvment from current state of the art, only a improvment compared to without. This reduced a lot the contribution of the paper. You should then modulate your improvment, discuss your results in the conclusion and state the limit of your system.
When I did my controller, I compute a trajectory for the torso and with the Jacobian, it's possible to find the leg trajectory. Then, we use a search algorithm until each leg reach the ground using a FSR at the tip of the leg, as any industrial robot. If the leg do not reach the ground (after a limit defined by the workspace of the leg), the stability is then reduced and then try to find a torso position in order to improve stability before the next walk motion. An IMU, like a drone adjust the torso when it's possible.
As you state there is no algorithm, the reviewer assum there is no solution when a leg reach its workspace limit. This is a limit of your controller and should be added in the discussion/conclusion of the current paper.
Right now, there is no proof of:
- improve the stability of the robot’s walking (line 18)
-improve the stability of the movment (line 209)
- improvement of walking performance (line 169)
-improve the stability of the motion (line 506)
If there is no improvment, remove this or make a proof.
4- In Fig. 14, I think you should correct: Adjustment strategy on based on master polygon
5- In Fig. 20, 22 compute the difference with/without VSDM and state how much this difference is significant. There is enough difference ? For me, 2-3 deg. is not significant to be different. State what is your threshold. The reviewer agree with you, the trajectory is not the same, but an ANOVA with MatLAB (anova function) could be used easilly. Otherwise, it' difficult to state the motion of the torso is smoother: ok, but the trajectories are differents and the leg reach a pebble in one trajectory and not the other. Is it the difference ? Discuss the limit of your method.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
Authors have taken into account my suggestions, improving significantly the clarity in the presentation.
Author Response
Thank you for your time and efforts on this paper. We are open to revise any further suggestion from you.
