An Improved Initialization Method for Monocular Visual-Inertial SLAM

Round 1

Reviewer 1 Report

1 it is necessary to expand the description of the literature, including the number and younger, highlighting the main advantages and disadvantages

2 conclusion should be expanded

3 It is necessary to formalize the field of application - information is provided on any robots (including UAVs), and an example is given on ground robots 

4 should compare monocular and binocular technology

Author Response

Thank you for your kind suggestions. Our response is in the attachment. Please check it.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper is very difficult to read:

• The usage that appeared in this paper is strange. For example, “Since the measurements of IMU and camera output with different rates”, “Because the gravity direction possesses a great effect on the acceleration estimation”, “Since the measurements of inertial and visual odometry are the relative movement.”…
• The symbols and denotations in the equations are not normal, and the corresponding descriptions of them are not sufficient. So, I cannot check whether the mathematic equations are correct.

So, I think that the authors should make a major revision to the paper and resubmit it.

Author Response

Thank you for your kind suggestions. Our response is in the attachment. Please check it.

Author Response File: Author Response.pdf

Reviewer 3 Report

Please find below a list of detailed comments that should be addressed to improve the manuscript:

1) The background on SLAM approaches provided in the Introduction is unsatisfactory and deserves to be extended. In particular, before delving into the specific visual-inertial technologies that are at the basis of the present work, authors should provide a more broad overview of the different technologies that can be used as well to perform simultaneous localization and mapping. This topic is indeed attracting a lot of interest in the emerging contexts of 5G communications, since directional antenna arrays and higher bandwidths can be fruitfully exploited to achieve high accuracy. In this respect, some interesting and quite recent papers that can be added are:

- H. Wymeersch et al, "Downlink Single-Snapshot Localization and Mapping With a Single-Antenna Receiver," in IEEE Transactions on Wireless Communications, 2021;

- Y. Ge et al, "5G SLAM Using the Clustering and Assignment Approach with Diffuse Multipath" Sensors, 2020

but authors are encouraged to add also other references they may be aware of.

2) Almost on the same line of my previous comment, the Introduction misses a clear explanation of why the INS sensors are generally used for positioning and tracking purposes. I would suggest to extend also this part, providing a pointer to the existing literature that shows the benefits of fusing inertial measurements to achieve highly accurate localization (and possibly mapping). Indeed, such information are crucial when operating in harsh propagation environments (e.g., rich of multipath) where the typical GNSS information are highly inaccurate or completely unavailable. From a search on Google scholar, some interesting and recent references that can be added are:

- G. Ricci et al, "A Pseudo Maximum likelihood approach to position estimation in dynamic multipath environments", Signal Processing, 2021

- C. Wang et al, "A High-Accuracy Indoor Localization System and Applications Based on Tightly Coupled UWB/INS/Floor Map Integration," IEEE Sensors Journal, 2021.

3) A Table summarising the main advantages and drawbacks of the existing initialisation methods discussed in Sec. I could be helpful to better clarify the main novelty of the present contribution.

4) In Sec. 2.1, the details on the transformation matrices from the body frame to the camera frame need to be provided for completeness.

5) If possible, I would suggest adding an even coincise complexity analysis in Sec. IV to corroborate the performance in terms of RMSEs and highlight possible trade-off between the achieved accuracy and the computational complexity, if any.

6) In Fig. 5, the labels on the x-axes should be "time [s]" instead of "tims(s)".

Author Response

Thank you for your kind suggestions. Our response is in the attachment. Please check it.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

(1) The expression is definitely very confused, but the problems with the symbols and denotations in the equations make it more difficult to read. There are two subscripts (one is on the left and the other is on the right side) in most of the symbols. This usage is not popular, and I suggest modifying them with the commonly used ones (refer to the book, for example, OpenGL). 

(2) Many symbols miss the descriptions, so, it is difficult to know the meanings of these symbols. I suggest double-checking the equations and the symbols and adding the corresponding description to each symbol that appears in the paper.

Author Response

Thank you for your kind suggestions. Our response is in the attachment. Please check it.  The description and definition of symbols has embodied in the revised manuscript.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors correctly addressed all my comments.

Author Response

We would like to sincerely thank you for your comments and suggestions.

Round 3

Reviewer 2 Report

The paper has been revised two times, but it still needs to improve the quality in the part of preliminaries, and experiments. The comments on the paper can be found in the attachment. Besides, I suggest: (1) making a dictionary of the notations, and it will be better for readers. (2) You may use Grammarly to correct the basic problems with grammar and expressions. "https://app.grammarly.com/"

Comments for author File: Comments.pdf

Author Response

Thank you for your kind suggestions. Our response is in the attachment. Please check it. 

Author Response File: Author Response.pdf