Using a Two-Stage Method to Reject False Loop Closures and Improve the Accuracy of Collaborative SLAM Systems

Round 1

Reviewer 1 Report

• l. 50: The description of figure 1 subplots is incorrect. Subplot b) and c) are switched.
• Figure 1. It would be useful for the reader if you give more details about the optimization with g2o (number of vertices, edges, needed iterations, optimization algorithm, etc.)
• l 64. The authors affirm that "In a collaborative SLAM system, generally, the pairwise internal consistency information between the loop measurement value and the local odometer measurement could be used to reject false inter-robot loop closure". But this affirmation is not only true for collaborative SLAM, but also for single robot SLAM.
• Related works.
  • As the work related to graph SLAM system, references such as "A tutorial on graph-based SLAM. IEEE Intelligent Transportation Systems Magazine, 2(4), 31-43." where they describe what a graphSLAM works, or "Graph SLAM Built over Point Clouds Matching for Robot Localization in Tunnels. Sensors 2021, 21, 5340." where they describe a process to carry out a graphSLAM through scan matching. These references would help the reader.
  • It would be also useful for the reader if a short summary of loop closure techniques is included.
• l. 109. It is not clear what  Sünderhauf proposes, and what the switchable constraints means. The whole sentence should be explained.
• l. 147. The authors affirm that there lack credible speedometer measurements, but it is not justified in the given reference [20]. This affirmation is debatable. A proper justification should be included.
• l. 175. It is confusing for the reader that the robot are called a, b, c..., and they are called alpha and beta in future points. Please, keep a consistence between the robot names.
• l. 176. i,d=2,3 can me confusing for the reader, please separate both variables.
• l. 178. A dot is forgotten before X.
• Figure 2. It is not clear how the relative position between robots is calculated (green lines). It is also unspecified how the loop closures are done in a single robot (I guess it may be any technique, but it should be specified).
• l. 187. It is written that 'Z' is represented in Figure 2, but it is not.
• l. 201. It is not clear what anchor points means.
• l. 215. The text says "As shown in Figure 1, the transformation between...", but Figure 1 does not show that, neither another figure.
• l. 221. Errors is not correctly written.
• l. 254. Do you mean dividing subsets, or dividing in subsets?
• l. 256. It is not completely clear what set is being generated. Do you  mean a set of loop closure of each vertex?
• l. 313. "correct s". Is it correct?
• Figure 4. It is not clear what lines and arrow mean, and what color means. Please, explain deeper what they are.
• l. 331. Difficult to read the sentence starting in "from the starting..."
• l. 333. It is not clear where the false loop is represented in Figure 4.
• l. 341. Unspecified second symbol.
• l. 353. How is it possible that computing time get smaller and smaller with increase of the vertex and edge density. This fact should be justified.
• Figure 5 and 7. In caption you refer to yellow line, but there is no yellow lines.
• Table 3. Do you consider to include the computation time of each algorithm. It is an important part of any algorithm that should be taken into account.
• Conclusions. It would be useful if you specify if the algorithm is used online or offline.

 

Author Response

Dear Editor and dear Reviewers:

Thank you very much for your valuable comments and suggestions. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our study. We have studied comments carefully and have made correction which we hope meet with approval. All modifications in the paper are made in revision mode. Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript titled "Using A Two-Stage Method to Reject False Loop closures and Improve the Accuracy of Collaborative SLAM systems" is well structured article. The formulation of two stage method is clearly presented. The results validation is also clear. Overall the manuscript presented a novel contributions to the dealt field. However, the manuscript needs thorough editing as per the comments mentioned below:

1. There are lot of short forms, but it seems the author(s) didn't add the full forms (e.g., SLAM,  CSAIL, PCM). So it would be better to add them.
2. There were lot of we, our etc., please avoid them and rewrite the sentences.
3. This should be "RANSAC (Random Sample Consensus)" the other way round. Please check it.
4. In lines 112 and 113, "However, the computation complexity is thereby increased", but reasons are not mentioned clearly.
5. In line 127, it should " ℓ₀-norm", please check and correct it.
6. Eq. 1 representation is confusing, please correct it, in the form maximization optimization function.
7. In line 209, what are those "other factors"?
8. Use the symbol for "chi-square test"
9. Line 234, looks rudimentary "Figure 3 shows a flow chart of this algorithm", try merging with other paragraphs to have correct sequence.
10. In Figure 3, why "Generating pose graph for single robot", and  "Eliminating loop closures of single robot" are mentioned twice. Please revise it.
11. In section 5, we do not see any valid "Prospect", it should be revised carefully. 

Author Response

Dear Editor and dear Reviewers:

Thank you very much for your valuable comments and suggestions. Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our study. We have studied comments carefully and have made correction which we hope meet with approval. All modifications in the paper are made in revision mode. Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The responses were satisfactory and the manuscript can be published.