Research of Flexible Assembly of Miniature Circuit Breakers Based on Robot Trajectory Optimization

Round 1

Reviewer 1 Report

The paper is well written and enjoyable to read. State of the art is pretty good. Some figures could be approved regarding resolution and self-explanation. For example, the axes in Figure 8 are too small. The statement in Figure 9 is not clear.

Line 167: Please give a complete list of states which are considered in trajectory optimization.

Line 170: please argue for the " relatively high efficiency" of the approach. Are there references? Could you give a reason for this statement? Why fifth degree?

Figure 10 is very good.

Figure 12: Could you please the oscillation in your graphs? How is this affecting your results? How did you prove optimality with these oscillations?

The statement in Figure 17 is not clear.

Please at least give some ideas for validation.

Please include a critical discussion of your assumptions and your results and add some next steps.

Author Response

Thank you for your comments. The revised manuscript in attached.  The responses to your comments are as follows:

The statement in Figure 9 is not clear.

After a careful consideration, we decide to delete Figure 9 since it only shows how the model looks like in the simulation software, while Figure 8 has illustrated the issue of the axes.

Line 167: Please give a complete list of states which are considered in trajectory optimization.

The states considered in the optimization are the positions, velocities, accelerations and jerks of the joints, as explained in line 176.

Line 170: please argue for the " relatively high efficiency" of the approach. Are there references? Could you give a reason for this statement? Why fifth degree?

The quintic (fifth-degree) polynomial is a comment approach in trajectory planning. The reason why it is chosen is now described in line 179-183.

Figure 12: Could you please the oscillation in your graphs? How is this affecting your results? How did you prove optimality with these oscillations?

The oscillations in the graphs are now explained by line 308-313. The key idea is to keep the continuity and smoothness of the curves, and meanwhile to make the peaks of the oscillations as close to the constraints as possible, so that the performance can be improved.

The statement in Figure 17 is not clear.

Figure 16 (previously Figure 17) shows the process of how randomly-postured parts are picked, adjusted and positioned, which is now described in line 381 to 385.

Please at least give some ideas for validation.

In the revision we add Table 7 which shows the practical optimization results of the processes of different MCB parts on our experimental assembly platform. It is believed that the practical operation is a good way to validate the simulation. It can be observed that significant improvement is achieved by our method. Descriptions are in line 390-395.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper presents an interesting improved version of a PSO-based approach for robot kinematics scheduling under multiple constraints and fitness criteria.

The overall approach is sound and valid; however there are a few drawbacks that must be addressed:

1) The flowchart in Figure 10 is formally incorrect and unclear in practice; the "swicth function" node seems inappropriate, as it is not a decision (see next steps) and not directly related to what happens at the second half of the (different) flow paths.

2) The authors describe that the fitness function includes strain-minimization factors; however, according to Eq.13-15 only time factors are implemented. This means that no other target is taken into account.

3) Figures 12 and 14 are the most informative in terms of optimization merits from the proposed approach. However, there are two evident problems: (a) the alternative algorithms should be presented too, in order to check what is different in practice; (b) regarding strain-minimization, as a fitness factor described earlier (see section 1), the optimized scheduling actually multiplies max acceleration and max jerk by a factor of more than x4. Either limiting these is outside the scope of the proposed approach or the experimental evidence are in contrast to what the authors describe as the general context of the optimization task at hand.

4) English language must be carefully reviewed and edited, as some terms and words are incorrect (e.g. "In practical,", "there is a contradictory", etc).

Author Response

Thank you for your kind comments. The revised manuscript in attached. The responses to your comments are as follows:

Please include a critical discussion of your assumptions and your results and add some next steps.

Now a paragraph (line 405-416) has been appended to the conclusion to discuss the weaknesses of presented work as well as our further work of improvement.

1) The flowchart in Figure 10 is formally incorrect and unclear in practice; the "switch function" node seems inappropriate, as it is not a decision (see next steps) and not directly related to what happens at the second half of the (different) flow paths.

We apologize that we did not explain the flowchart clearly. In Figure 9 (previously Figure 10), the switch function is exactly a decision step that determines whether a particle should be directly forwarded to the optimization iterations or be updated to meet the constraints. The related explanation is in line 277-281.

2) The authors describe that the fitness function includes strain-minimization factors; however, according to Eq.13-15 only time factors are implemented. This means that no other target is taken into account.

Thank you for the kind comment. However, in the article we only investigated the time-optimal optimization, while the optimization with strain-minimization was not described. The fitness function was applied to improve the PSO for time factors, which were constrained by the position, velocity, acceleration, and jerk. The fitness function was not related to the strain which was unable to be measured in our experimental system. It is a very good point that we can add strain factors in the optimization in our future work.

3) Figures 12 and 14 are the most informative in terms of optimization merits from the proposed approach. However, there are two evident problems: (a) the alternative algorithms should be presented too, in order to check what is different in practice; (b) regarding strain-minimization, as a fitness factor described earlier (see section 1), the optimized scheduling actually multiplies max acceleration and max jerk by a factor of more than x4. Either limiting these is outside the scope of the proposed approach or the experimental evidence are in contrast to what the authors describe as the general context of the optimization task at hand.

(a) In the simulation we also applied the standard generic algorithm for comparison, as shown in Table 6. We add Table 7 which shows the practical optimization results for different MCB parts on our experimental assembly platform.

(b) In this article we do not discuss strain minimization, but only focus on time optimization. Therefore, to shorten the operating time, the acceleration and jerk are unavoidably increased. Our point is, to keep the continuity and smoothness of the curves, and meanwhile to make the peaks of the oscillations as close to the constraints as possible, so that the performance can be improved. Our results show that although the acceleration and jerk increases, they are still within the constraint range, which supports our purpose at the beginning. We are sorry for the unclarity, and now it is explained in line 308-312.

4) English language must be carefully reviewed and edited, as some terms and words are incorrect (e.g. "In practical,", "there is a contradictory", etc).

Many thanks for your kind advice. Errors are now corrected.

Author Response File: Author Response.pdf

Reviewer 3 Report

The article presents an approach based on the optimization of the robot trajectory to solve the problem of high rigidity in the current MCB production and presents the flexible automatic assembly of multiple MCB parts with one robot. In my opinion, the topic is interesting and quite innovative. A good idea for solving the proposed task is to use the PSO time-optimal trajectory method with a multi-optimization mechanism. The results presented by the authors are quite convincing and confirm the validity of the use of this method. This article is well-worded with good consistency logic. The article presents a series of tests that are intended to enable the reader to understand the method used. However, I have questions for the users:

I propose that the authors expand their conclusions more,

I suggest adding more references as there are a many articles on this topic.

Author Response

Thank you for your kind comments. The responses to your comments are as follows:

I propose that the authors expand their conclusions more.

Now a paragraph (line 396-405) has been appended to the conclusion to discuss the weaknesses of presented work as well as our further work for improvement.

I suggest adding more references as there are many articles on this topic.

References of the late research are updated and introduced (ref. 13, 19, 22, 23, 24).

Author Response File: Author Response.pdf