All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Authors have addressed all the comments from the reviewers. Hence, it is recommended to accept this paper in its current form.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
The author has prepared well the manuscript based on the comments that have been given by last reviews. The control part especially has been enhanced a lot in the current manuscript. So, the paper is accepted in its current form.
Overall OK
Overall OK
Thank you to the authors for their efforts in revising the manuscript. While the reviewers have acknowledged the improvements made in the revised version, several significant issues still require attention. Therefore, the authors should carefully revise the manuscript in accordance with the provided feedback.
No comments
No comments
No comments
The drawings and photos of simulation and actual experiment results are not beautiful and the resolution is not high. The author should update the new version of all the figures in the article. Furthermore, the subtitle of the drawings should also be standardized according to a single format and font size.
The contribution of this revised manuscript is clear and good. All comments from the reviewers were very good, and the author’s answer was very good and satisfied. Therefore, my recommendation is to accept the revised manuscript that has ID #101404 for publication.
The contribution of this revised manuscript is clear and good. All comments from the reviewers were very good, and the author’s answer was very good and satisfied. Therefore, my recommendation is to accept the revised manuscript that has ID #101404 for publication.
The contribution of this revised manuscript is clear and good. All comments from the reviewers were very good, and the author’s answer was very good and satisfied. Therefore, my recommendation is to accept the revised manuscript that has ID #101404 for publication.
Author has enhanced the current version but still there are many things to enhance:
- There is no correlations between the camera transformation with path planning and control. There should be a relevancy between all these components: Camera>>transformation>>mapping>>path planning>>control. In fact, each component have been designed and discussed alone. The feedback of controller is the heading angle which is measured by accelerometer&magnetometer and compared with the heading angle coming from path planning that is determined from camera transformation. All are related to each other, but we couldn't see that relevancy in manuscript.
Author should reconstruct the system with showing this relevancy.
Author should also correlate the equations between theses sub-systems.
- Author used the same symbols for equations with different definitions : such as x and y in Equations 1-4 (image coordinate system) are same as in Equations 9-20 (real coordinate system). In fact, there must be correlation between all these equations with utilizing different symbols.
- All equation symbols must be defined before or after the equations, for example K in equation 2. Please check all equations.
- The feedback of the PID is not clear, is it measuring the rotational angles of the both wheels by accelerometer and magnetometer, then how they can be converted into heading angle to be compared with the theoretical heading input.
Also how to derive the heading angle from path planning is missing.
Figure 13 still doesn’t clarify the input and feedback of the PID controller.
Also how to convert the right and left rotational angles into PWM, i.e. how much current that will be supplied to motor.
More elaboration on the input and feedback are needed.
- The accelerometer and magnetometer sensors are very important component in this system, but there are no signal processing or figures showing their results of these two sensors in the whole manuscript.
- The results of the controller are still missing, as we don't see any figures showing the controller performance, i.e. the tracking errors of the controller is not presented.
- There should be a comparison between heading angle coming from path planning (theoretical value) and the heading angles measured by magnetometer sensor (actual value).
After reviewing process, this paper requires a "Major Revision". Authors should carefully revise the paper based on all the comments suggested by the reviewers. In the "Abstract", please provide the quantitative performance achievements of the proposed works. In the "Conclusion", it should reflect what author has achieved and the conclusion should be clearly stated.
1. The language and presentation style are not suitable for the structure of an academic research paper. The author needs to improve or use the support of Academic English Center to improve the quality of the article.
2. The Introduction section is still too sketchy and does not highlight the existing problem towards the solution goal of the proposed method. Therefore, the novelty and contribution have not been clearly clarified. The author should also provide a Related works section for more comparison and clarification of the contribution of the author's method. Finally, in the conclusion, it is necessary to write and clarify the results achieved, limitations and future developments.
3. The image quality of many figures is too small, for example figure 13. The author should be consistent in image representation, size and quality to ensure the magazine's requirements.
1. In Figure 1, the author provides synchronization in the time step of the AMR control process using a PID controller and feedback of the "robot heading angle" using a magnetometer Sensor with the "calculation of the AMR rotation angle" ” based on “Blod analysis and removable area extraction”. With modern control methods, noise reduction and data transmission and processing capabilities are required to avoid delays between the input and output of the AMR controller. This is one of the main obstacles that affects stability and sustainability in robot control.
2. Results of segmented images from the Deeplabv3+proposed network model, so how does the number of image frames relate to the sampling time step when setting up the PID controller? Because the article has important results related to embedded robot control systems, the author should provide indicators of the training model in terms of quality and processing speed to ensure real-time control process.
3. The scenario when controlling the robot in an environment is too simple, in Fig. 15. The author should provide additional scenarios with static obstacles obstructing the robot's path, from which the robot needs to develop a navigation strategy to ensure it avoids obstacles and meets the requirements such as short paths. best, smoothed path, avoiding static obstacles. Furthermore, does the author's proposed method help the robot avoid moving obstacles? If so, it needs a detailed explanation.
4. From image space to real space, there needs to be a conversion and adjustment process. The author should add the formula and calculation process in the revision. Furthermore, when path planning is obtained, the robot moving along the trajectory also has deviations from the obtained path planning. The author should also learn and supplement this issue.
5. In the semantic segmentation model structure, the author proposes Deeplabv3+Network, so it is necessary to clarify the structure and advantages of the proposed network here? Why choose that structure?
1. The conclusion and abstract are still too sketchy. After adding and rewriting the Introduction and Related works sections, the author needs to complete the Abstract and Conclusion sections to clearly show the newness, contributions, limitations and applications. application of the proposed method in practice, as well as future developments
In general, the contribution of this manuscript is good. Therefore, my recommendation is to accept (Minor Revision) this manuscript that has ID #101404 after several adjustments must be made before publication.
My specific comments are:
• In the abstract, the result of this work must be described briefly with data in order to show the effectiveness of the proposed work.
• The author did not describe the drawbacks of each conventional technique in the introduction paragraph.
• Add the problem definition of this work in the introduction paragraph because it is not clear.
• Please include the references for all equations.
• In conclusion, please add an enhancement percentage (%) that demonstrates the proposed algorithm efficiency for your method when used with another method.
• In conclusion, please add an enhancement percentage (%) that demonstrates the proposed algorithm efficiency for your method when used with another method, as well as a percentage (%) that demonstrates the difference in results between simulation and experimental works
• The author did not describe the drawbacks of each conventional technique in the introduction paragraph.
• Add the problem definition of this work in the introduction paragraph because it is not
The article presents a method for finding the drivable path using deep learning. I have the following concerns regarding this paper:
1. The work presented in this paper doesn't show any contribution comparing with the previous works, as there are many papers published on how to use deep learning based image processing for finding the robot path.
2. The mathematical modelling of the robot is not well presented as there are only 3 equations (6-8) to describe the motion of the robot. Equation 7 is wrong as the heading angle will affect to velocity in x and y. The center of gravity and its affect to the kinematics equation is missing. The authors doesnt include non-holonomic condition in the equations. Author must revise the kinematic modelling of the robot.
3. The author mentioned that he used PID controller to control the robot, but he didn't show how to develop the controller. how he used the kinematics or dynamics models in the controller. How the controller compensates the errors. no figures to show the tracking errors.
4. This work presents a very simple movement of the robot based on deep learning and image processing, which finds an arbitrary path based on the heading angle measurement in the environment. So, there is no path planning in this work. This work should be entitled as SLAM localization of robot using deep learning.
5. From previous comments, I can conclude that the work lacks of many robotics fundamentals and doesn't show any novelty in this field. Thus It is not acceptable for publication in its current version. Perhaps the author can work on the comments and submit in future.
The comments about the experiments are as follows:
1. The author has explained clearly the part of deep learning but he failed to present the robot modelling, control and path determination.
2. A Very simple explements has been performed in corridor with clear borders and Garbage box . It should be performed with some others challenging environments.
3. The ground truth is not clear how the author determines it.
4. As the author has developed new database, he should mention how he has labelled the data.
5. How to convert the signals of the control system (which is missing) into PWM. The authors shows in Figure 3 a general diagram for both control system and PWM. No figures to show nether the control system or PWM signal.
The validation is not prepared well as follows:
1. The authors has presented the results of segmentation for the proposed method deeplabv3+ and ResNet50 with many deep learning methods to find the drivable path, however there is no results or comparison of the control system which is very important to know the accuracy of robot movement.
2. There is no continuous path of the robot from the beginning until the end of the movement. So, we cant see how the path looks like.
3. No results to show how the proposed algorithm will deal with obstacles.
The contribution and robotics fundamentals are missing in this work, thus, I would recommend to reject the paper in its current version. I encourage the author to work hardly on the paper and submit in the future.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.