VS-SLAM: Robust SLAM Based on LiDAR Loop Closure Detection with Virtual Descriptors and Selective Memory Storage in Challenging Environments

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper discusses a LIDAR loop closure detection system that is claimed to be more robust by proposing a Virtual Descriptor and Selective Memory Storage techniques in a framework called VS-SLAM. This approach is designed to address two main problems that often hamper the performance of the loop closure detection, particularly in geometrically similar environments.

The first problem is the high sensitivity of descriptors to translational changes. Existing methods struggle to maintain the stability of the representation when position shifts occur, thus reducing the detection accuracy. To overcome this, VS-SLAM introduces a new method that improves the translational invariance of the descriptor so that the system becomes more accurate in detecting loop closures.

The second problem is the large memory consumption due to the calculation complexity of the previous method. As the mapped area increases, the memory requirement increases significantly, thereby burdening system resources. To address this challenge, VS-SLAM adopts Selective Memory Storage Technology, which enables more efficient memory management without compromising the detection accuracy.

Although various methods have been developed previously, most are still unable to solve the fundamental problem of the descriptor sensitivity to translational changes while addressing the issue of large memory resource consumptions. Therefore, it is claimed that the VS-SLAM framework presents a more adaptive and efficient solution for handling both challenges.

Based on the test results conducted on ten sequences with five test iterations for each sequence, VS-SLAM is claimed to improve accuracy by 66.41% in structurally similar environments and reduce memory consumption by 93.45%. However, we have made the following comments regarding the current version of the manuscript:

1. In line 141, a brief explanation of the purpose of the citation [25-27] should be provided so that readers can better understand the context.
2. In lines 144, 255, 370, and 381, the use of the phrase “thanks to”, which is less formal, should be replaced by a more scholarly phrase that fits the academic style. In addition, please check other words or phrases in the paper that can be informal or less scientific and adjust them to academic writing standards.
3. In line 147, the use of the word “first” for the first step in the proposed framework is appropriate. However, to be more consistent, the subsequent steps should also be stated sequentially, if possible. This will help the readers follow the flow better.
4. It is better if each equation is mentioned in the paragraph that explains it, so that the connection between the theory and the formula is clearer.
5. Based on the explanation in lines 362-364, Figure 5 needs to be given a more detailed explanation of the meaning of the figure, as well as a sign or visual marker that shows the problem discussed in that line. This will help readers, especially those who are not familiar with this topic, better understand the meaning of the figure.
6. The subtitles in Figure 5 are extremely large. Ensure that the prescribed format is followed.
7. Figure 6 also requires a more detailed explanation of its meaning and additional visual cues or markers related to the discussion in lines 368-370. This will help readers to understand the relevance of the image to the content of the text.
8. The subtitles in Figure 6 are extremely large. Make sure to follow a predetermined format.
9. In Table 2, the test results only compare the VS-SLAM with and without the selective storage memory. It would be better if the comparison also includes the previous method to demonstrate the robustness of the proposed approach.
10. Consider including an example for the accuracy visualisation to support the data presented in Table 2.
11. In lines 404-405, please review the statement “the localization error of SC-LVI-SAM 404 is 0.11 m” and the data presented in Table 2, as there is a discrepancy between the statement and the data in the table.
12. In lines 418-419, please review the correctness of the statement and its relationship with the data presented in Table 2, because there is a discrepancy between the statement in the text and the data in the table.
13. Ensure that the format and writing of the footer in each table are in accordance with established standards.
14. It would be better if comparison images were provided to show the results of determining the key descriptor in the point cloud.
15. To improve visual appeal, please consider providing a graph showing the memory usage during the trial, so that the increase in the memory usage can be analysed.
16. The discussion on accuracy and error is still unclear. The values of 93.45% and 66.41% mentioned in the Abstract are not clearly explained in the Discussion section. There will need to be an additional explanation of how these values were obtained, so that readers can better understand the calculations.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

English is quite good, but needs improvement in some parts, as mentioned in the comments section.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Author proposed VS-SLAM, a novel robust SLAM based on LiDAR loop closure detection with virtual descriptors and selective memory storage,this is very meaningful for the development of SLAM. Unfortunately, many key contents are missing from the vague expression:

1. In Chapter 3, the VS-SLAM algorithm investigated by the author is primarily based on LiDAR loop closure detection, while also incorporating visual inertial odometry as a prior. Please provide a detailed description of the relative contributions of LiDAR and vision-based perception within the VS-SLAM algorithm in terms of sensor utilization.
2. Algorithm Dimension: As for Figure 1, the system is developed based on LVI-SAM.Please explain the similarities and differences and improvements of the proposed VS-SLAM algorithm and LVI-SAM algorithm in detail, thereby highlighting the innovation of virtual descriptors and selective memory storage.
3. Experiment Dimension:From the title of the paper, VS-SLAM algorithm should be applied in challenging scenarios. But the influence of different lighting conditions, scene complexity, motion trajectory or posture is not considered in experiment. In order to highlight the effectiveness of the algorithm, the authors need to add experimental results in multiple “challenging environments” to the experimental section.
4. The textual content of the article is accurate and logically structured. However, the tables and graphs are disorganized and not properly aligned with the text. For example, referred to Figure 3 in Chapter 5, but Figure 3 is actually located in Chapter 4.Please author review the entire article thoroughly to prevent these issues.

Comments on the Quality of English Language

 The English could be improved to more clearly express the research.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author has revised most of the comments that have been given in the previous review. However, the revisions that have been made still require some improvement, as in the following notes:

1. In figure 9, the red bar is “SC-LVI-SAM or VS-SLAM-w/o-st”. This is confusing.
2. Double check your statement on lines 448–451. Your explanation does not match the graph in Figure 9.
3. On lines 368-374, consider improving the last sentence to match the paragraph’s main idea. In general, the explanation of the results and discussion section are quite good but still need improvement in the flow of the narrative so that readers can understand the discussion more easily.
4. The layouts of the figures and tables should be improved according to the template guidelines. Explanations and images are too far away, which can make them inconvenient for readers.
5. Recheck the suitability of the image title format/template, especially for images with multiple panels (a and b).
6. Recheck the suitability of the table format based on the official journal template.
7. Future research directions may also be highlighted in the conclusion.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

The English quality is quite good, but the flow of explanations could be improved for better coherence and readability. The English could be improved to more clearly express the research

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

Accept in present form.

Author Response

We sincerely appreciate the reviewer's valuable feedback during the initial review and their support in recommending our manuscript for acceptance.