Indoor Positioning Algorithm Based on Maximum Correntropy Unscented Information Filter

Round 1

Reviewer 1 Report

This paper starts from an assumption that the noise of signal strength does not follows Gaussian distributions. And proposed a novel method, called MCUIF, based on non-Gaussian distribution. This is a very good starting point of the paper. However this paper fails to show that it would outperform the previous ones, particularly those based on Gaussian distribution and how it would outperform them.

This paper is based on non-Gaussian distribution and tries to show it by observations such as figure 1 and 2. No actual distribution in real world exactly follows Gaussian distribution and the point is how much it differs from Gaussian distribution. The authors try to show the non-Gaussian distribution of signal strength. However, the experiments shown by figure 1 and 2 are limited and insufficient to convince this assumption. 

We may need more discussion about the results shown in table 1 and figure 3, which compares the accuracy with other methods. First of all, the experiment looks immature. The experiment was conducted at a site, which seems a simple place (no floorplan was given in the paper). In order to show the strength of the proposed method, we may need more extensive (in a sense that experiments with more divers indoor places would be better) and more intensive (in a sense that more detail discussion such as how the APs are located, and how the accuracies depend on the location of APs and indoor structures) experiments.
Minor comments;
-    Labels of y-axis such as 0.5, 1.5, etc in figure 2 are meaningless.
-    What is the difference between table 1 and figure 3?
-    Is it meaningful to separate the RMSE to x-axis and y-axis?
-    The mathematical derivations in section 3.2 may be simplified and the detail would be given in annex.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The study presents a clear and profound description of an analytical concept of an RSS-based indoor positioning method MCUIF.
The authors also run an experiment showing slightly better accuracy than concurrent methods (LS, EKF, UIF), at least in table 1. But the situation seems less clear when looking at figs 6-8. 

In my opinion, the experiment needs a more robust fundament. But I can be wrong, of course. Therefore I recommend performing a test of statistical significance of the experiment, e.g., by testing a null hypothesis: "MCUIF has the same accuracy as the LS (and then EKF, and UIF) method." The test will prove that the experiment results are statistically significant even with the current measurements or need more. 

 

Multiple referencing [x, y -z] - you better refer to a source only when you use it in your text.

 
Graphs have to have a complete description. Also, It is hard to follow when the colors are not defined and not corresponding ta all the figures (where appropriate).  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

This paper presents Indoor Positioning Algorithm Based on Maximum Correntropy Unscented Information Filter. The work looks interesting.

However, some revisions are needed for manuscript to be accepted as a Journal paper. Comments from reviewer are as follows.   1. Authors have to improve the literature review referring more works in the literature. This time authors use nodes to estimate the positionaning. However, some work has been done to achieve the positioning, following natural features in the environment. For example; A Study on Hovering Control of Small Aerial Robot by Sensing Existing Floor Features, Improving Landmark Detection Accuracy for Self-localization Through Baseboard Recognition. Probably, these works would be easier to make applications. How their errors are also a little bit higher. On the other hand, some works have also been done, physically setting marks for positioning.  For example;Development of an Automated Camera-Based Drone Landing System. Authors have to discuss these kinds of works in the introduction.   2.  The experiments scene in Fig. 1 is not clear. Do you install tags in there? If so, pls indicate in the figure.   3. The theory parts are well written. But, in some equations, symbols are not described. Pls, recheck all the equation.   4. This time, authors just evaluate the proposal with simulations. I recommend authors to discuss the challenges to apply proposed method for wheel robotics or other.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

There is no reaction to my remark:

In my opinion, the experiment needs a more robust fundament. But I can be wrong, of course. Therefore I recommend performing a test
of statistical significance of the experiment, e.g., by testing a null hypothesis: "MCUIF has the same accuracy as the LS (and then EKF, and UIF)
method." The test will prove that the experiment results are statistically significant even with the current measurements or need more. 

Therefore I stay with the same evaluation - major revision. Please note that if you do not react to that remark (you can react by incorporating the remark or by explaining, why my remark is not applicable), I will recommend rejecting the manuscript.

Author Response

Dear Reviewer:

Thank you for your careful work and suggestions. We did a test of statistical significance of the experiment using the T test method, see section 4.1 for details.

Author Response File: Author Response.pdf

Reviewer 3 Report

This revised manuscript is in an acceptable level.

Author Response

Dear Reviewer:

Thank you for your careful work and suggestions. We found some grammatical and vocabulary errors, we have corrected.