Investigation of UHF Signal Strength Propagation at Warehouse Management Applications Based on Drones and RFID Technology Utilization

Round 1

Reviewer 1 Report

The paper proposes a method of inventory of objects in external warehouses, each object being provided with an RF identification tag (RFID). For this purpose, an unmanned aerial vehicle (UAV) travels a route that covers, according to an imposed algorithm, the entire surface of the warehouse. The UAV transports RFID-specific equipment that emits interrogation signals to which it receives identification signals from all RFIDs in the covered area so that, based on the received information, it can create a map with the positioning of objects in the warehouse. The authors' solution for this objective is to optimize the range of heights within which the UAV trajectory must be placed, based on RSSI experimental data obtained for several height levels of the areas where the inventoried objects are stored.

Currently, the topic is still interesting only to a small number of readers, but the foreseeable evolution of the techniques of storage and management of goods justifies the publication of the paper. However, the current presentation focuses mainly on the interpretation of the results obtained from RSSI measurements and less on the parameters related to the RFID chain, although they are also significant if analyzed properly.

For example, there are thirteen (more than half of the total!) bibliographic references related to antennas [11-23], the properties of which the authors claim to play an important role in the inventory application (lines 168-172). However, the relevant parameters of the RFID antennas and interrogator used in this paper are not specified, although they allow for easier interpretation of certain experimental results.

• Also, it is not clear whether the authors took into account the fact that, at long distances between the reader and RFID, several shelves and supports can be interposed between the objects to be inventoried, so the simplified model of direct visibility or line of sight RF signal propagation is no longer applicable.
• The results reported in Figure 2 could be explained more simply by means of radiation patterns of both interrogator and RFID antennas, polarization mismatch, signal fading due to interferences produced by multiple reflections (multi-path propagation, as mentioned in rows 212-215, but ignored by authors) – this additional info could answer the RSSI distributions in Figure 2. Recommended reading: “Complete link budgets for backscatter radio and RFID systems,” IEEE Antennas Propag. Mag., vol. 51, no. 2, pp. 11-25, April 2009 (authors: J. D. Griffn and G. D. Durgin).

The authors claim (rows 51-53) that “The UAV carrying the reader performs the inventory automatically according to the specified trajectory, faster, without errors, with a defined and repeatable route.” Furthermore, the lines 142-146 mention the issue of UAV displacement on a trajectory imposed by mapping and more accurate positioning of the deposit, specified in [9], and suggest a flight controller. However, this does not fully solve the problem, as there are other factors that should be considered in this paper, i.e. the wind can randomly change the UAV’s trajectory, therefore accurate feedback from an UAV tracking and positioning system should be provided for maintaining the right trajectory.

Recommended/suggested text replacements

Please revise the excerpts below, as your translation is not clear enough:

• [Lines 147 - 148] aims to measure resonant frequency of goods dielectric constants wirelessly
• [Lines 240 - 243] Based on the data hypothesis test for a difference in two population means and medians that were performed for each measured distance the significance of the difference can be determined. Based on the hypothesis tests it could be possible to find a maximal distance that allows the distinguishing between the tag under test and surrounding tags.

Author Response

The paper proposes a method of inventory of objects in external warehouses, each object being provided with an RF identification tag (RFID). For this purpose, an unmanned aerial vehicle (UAV) travels a route that covers, according to an imposed algorithm, the entire surface of the warehouse. The UAV transports RFID-specific equipment that emits interrogation signals to which it receives identification signals from all RFIDs in the covered area so that, based on the received information, it can create a map with the positioning of objects in the warehouse. The authors' solution for this objective is to optimize the range of heights within which the UAV trajectory must be placed, based on RSSI experimental data obtained for several height levels of the areas where the inventoried objects are stored.

Currently, the topic is still interesting only to a small number of readers, but the foreseeable evolution of the techniques of storage and management of goods justifies the publication of the paper. However, the current presentation focuses mainly on the interpretation of the results obtained from RSSI measurements and less on the parameters related to the RFID chain, although they are also significant if analyzed properly.

---

For example, there are thirteen (more than half of the total!) bibliographic references related to antennas [11-23], the properties of which the authors claim to play an important role in the inventory application (lines 168-172). However, the relevant parameters of the RFID antennas and interrogator used in this paper are not specified, although they allow for easier interpretation of certain experimental results.

Thank you so much for this comment. We agree. Based on your recommendation we added some description of the main purposes of those papers. We also added key parameters of used RFID antenna.

---

Also, it is not clear whether the authors took into account the fact that, at long distances between the reader and RFID, several shelves and supports can be interposed between the objects to be inventoried, so the simplified model of direct visibility or line of sight RF signal propagation is no longer applicable.

Thank you for this comment. That’s good point. We took into account remark you mentioned, but we agree that it may not be clear from the description that this is the use of drones for identification in outdoor warehouses, where we do not currently assume any barrier between the reader and the RFID tag. This has been added into text. In any case, as you rightly point out, this is a problem that must always be considered when designing RFID systems. 

---

The results reported in Figure 2 could be explained more simply by means of radiation patterns of both interrogator and RFID antennas, polarization mismatch, signal fading due to interferences produced by multiple reflections (multi-path propagation, as mentioned in rows 212-215, but ignored by authors) – this additional info could answer the RSSI distributions in Figure 2. Recommended reading: “Complete link budgets for backscatter radio and RFID systems,” IEEE Antennas Propag. Mag., vol. 51, no. 2, pp. 11-25, April 2009 (authors: J. D. Griffn and G. D. Durgin).

Thank you for this comment. We really appreciate the time you spent to provide us a useful reference. We carefully studied that literature and add some comments into that paragraph for explanation. 

“RF tags are further limited by the need, in passive systems, to power up their radio frequency integrated circuits through rectification of the incoming signal. There is a powerful analogy between the optical-polarizer experiment and the backscatter link of an RF tag system [25]. The measurements we performed are based on a real situation where a drone flies over an item in a warehouse. Because in real conditions it is not always possible to guarantee from which side the drone will fly over the RFID tag, we also chose RFID tags with circular polarization.”

---

The authors claim (rows 51-53) that “The UAV carrying the reader performs the inventory automatically according to the specified trajectory, faster, without errors, with a defined and repeatable route.” Furthermore, the lines 142-146 mention the issue of UAV displacement on a trajectory imposed by mapping and more accurate positioning of the deposit, specified in [9], and suggest a flight controller. However, this does not fully solve the problem, as there are other factors that should be considered in this paper, i.e. the wind can randomly change the UAV’s trajectory, therefore accurate feedback from an UAV tracking and positioning system should be provided for maintaining the right trajectory.

Thank you for this comment, maintaining a specific trajectory and eliminating environmental disturbances such as wind is essential to our success. Our initial measurements described in the article so far concern our first experiments with RFID on UAVs. In the project for which we are conducting this research, we are cooperating with experts from the field of Geoinformatics as well as experts from Korea dealing with camera image analysis and artificial intelligence. Our common solution will use a combination of several approaches to maintain the specific trajectory, trying to eliminate the influences you describe. On the one hand, the UAV will be equipped with accurate GPS, but the trajectory of the UAV will also be refined by the data from the image analysis to ensure tracking and positioning. So we modified text in article according to it. 

“The UAV, equipped with GPS and tracking and positioning system, carrying the reader can perform the inventory automatically according to the specified trajectory, faster, without errors, with a defined and repeatable route.”

---

Recommended/suggested text replacements

Thank you so much for these recommendations. We really appreciate it, and all your suggestions were implemented. We are thankful for text editing.

---

Please revise the excerpts below, as your translation is not clear enough:

[Lines 147 - 148] aims to measure resonant frequency of goods dielectric constants wirelessly

Thank you so much for this comment. We tried to explain little bit different and add some other comments to describe it. We hope that now it should be more understandable.

---

[Lines 240 - 243] Based on the data hypothesis test for a difference in two population means and medians that were performed for each measured distance the significance of the difference can be determined. Based on the hypothesis tests it could be possible to find a maximal distance that allows the distinguishing between the tag under test and surrounding tags.

Thank you so much for this remark. We agree that this explanation is not clear enough. We tried to re-formulate that sentence and now we hope that it should be more understandable.

“The significance of the difference between pairs of measurements (for tag AA and BB) were determined. The determination was based on the set of hypothesis tests for a difference in means as well as for medians of two populations (measurements for tags AA and BB) that were performed in various distances. Based on the hypothesis tests it could be possible to find a maximal distance that allows the distinguishing between the tag under test and surrounding tags.”

---

We would like to thank you again for your review, we sincerely appreciate for your time and considerations. Thank you for your suggestions.

Author Response File: Author Response.pdf

Reviewer 2 Report

The paper provides a UHF Signal Strength Propagation investigation for UAV RFID based identification of warehouse assets. The paper has some merit, however the methodology employed is questionable and the results are weak and, in this reviewer's opinion, not enough for a journal publication.

• In the paper motivation you argue that you want to find the "optimal height for reading UHF RFID tags using a reader with an antenna placed on an UAV" Your solution seems to be based on running 100 experiments and then coming to the conclusion that measured data sets are our of normal values. Therefore you set about finding statistically significant differences between the data pairs. This reviewer is not convinced of this approach (why not define an optimization problem and solving it, then compare it with field measurements?) and neither will the readers. Please clarify.
• The results are based mostly on histograms and box-and-whisker plots for 2-8 m reading distance for mean comparison. However, for median comparison you provide no such figures. Why?
• Table 4 spans 2 pages and represents raw data which could be put in an Annex or supplementary material. Please do that.
• "we decided that 95% confidence level is fully sufficient" - is this decision based on literature (if yes, please state relevant reference)? Or is it from the application requirements ? Please clarify.
• Please provide paper organization at the end of Sec 1 (i.e. The paper is organized as follows. Section 2 presents ...)
• It is not clear why you split Sec 2 into subsections, but provide only 1 subsection. Please either split text into more subsections or incorporate into the main section.
• "Mann-Whitney W test" - you must mean "Mann-Whitney U test" ?
• Please explain better Fig. 11. Readers might not be aware of "polyn." abbreviation. Does it mean "polynomial interpolation"? If yes, please explain how you performed it, what type of polynomial you used etc.

Author Response

Dear reviewer,

Thank you so much for your valuable comments and recommendations. We really appreciate time you spent to make a review of our paper. Based on your remarks we have tried to do our best to enrich our article as you suggested. The changes are indicated in an updated version in “track changes” mode. We really hope that these modifications and addition of some parts will contribute to accept the paper.

Comments and Suggestions for Authors

The paper provides a UHF Signal Strength Propagation investigation for UAV RFID based identification of warehouse assets. The paper has some merit, however the methodology employed is questionable and the results are weak and, in this reviewer's opinion, not enough for a journal publication.

In the paper motivation you argue that you want to find the "optimal height for reading UHF RFID tags using a reader with an antenna placed on an UAV" Your solution seems to be based on running 100 experiments and then coming to the conclusion that measured data sets are our of normal values. Therefore you set about finding statistically significant differences between the data pairs. This reviewer is not convinced of this approach (why not define an optimization problem and solving it, then compare it with field measurements?) and neither will the readers. Please clarify.

Thank you so much for your remarks. That is a really good point. In this paper we described the indoor tests (800 experiments in 8 various conditions) as a first step of the whole implementation process, next steps like measurements with the BAP tags or outdoor experiments should follow up.

We plan to solve the mentioned optimization problem and field measurements in next months and therefore we would like to include it in next paper(s).

---

The results are based mostly on histograms and box-and-whisker plots for 2-8 m reading distance for mean comparison. However, for median comparison you provide no such figures. Why?

Thank you so much for your note. The figures you mentioned (box and whiskers plots) are usable also for median comparison. Box-and-whiskers plot presents medians (arrowed vertical lines) as well as means (small crosses). Explanation added to the text.

---

Table 4 spans 2 pages and represents raw data which could be put in an Annex or supplementary material. Please do that.

Thank you for this comment, this is a very good reminder. We didn't realize that the table was that big and could be distracting as part of the text. For better clarity, the table has been moved to the end, in Annex A.

---

"we decided that 95% confidence level is fully sufficient" - is this decision based on literature (if yes, please state relevant reference)? Or is it from the application requirements ? Please clarify.

Thank you for this comment. The 95% confidence level is generally the most common in technical disciplines, e. g. 99% confidence level is used mostly in medical sciences etc. The usual levels are discussed for example here:

ILLOWSKY, Barbara; DEAN, Susan. Introductory statistics. 2018.

Illowsky, Barbara. Introductory statistics. Dean, Susan L., 1945-, Illowsky, Barbara., OpenStax College. Houston, Texas. ISBN 978-1-947172-05-0. OCLC 899241574

---

Please provide paper organization at the end of Sec 1 (i.e. The paper is organized as follows. Section 2 presents ...)

Thank you so much for your remarks. That’s true that for the readers clearer and easier for orientation. So, we added it.

---

It is not clear why you split Sec 2 into subsections, but provide only 1 subsection. Please either split text into more subsections or incorporate into the main section.

Thank you for this comment. Based on your suggestion we decided to remove subsection 2.1

---

"Mann-Whitney W test" - you must mean "Mann-Whitney U test" ?

Thank you so much for this note. Yes indeed, Mann–Whitney–Wilcoxon test mentioned in paper as “Mann-Whitney W test” may be also called Wilcoxon-Mann-Whitney or Mann–Whitney U test or otherwise based on the literature. A few other possible names added to the text.

• Mann, Henry B.; Whitney, Donald R. (1947). "On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other". Annals of Mathematical Statistics. 18 (1): 50–60. doi:10.1214/aoms/1177730491. MR 0022058. Zbl 0041.26103
• Wilcoxon, Frank (1945). "Individual comparisons by ranking methods". Biometrics Bulletin. 1 (6): 80–83. doi:10.2307/3001968. hdl:10338.dmlcz/135688. JSTOR 300196

or

• Fay, Michael P.; Proschan, Michael A. (2010). "Wilcoxon–Mann–Whitney or t-test? On assumptions for hypothesis tests and multiple interpretations of decision rules". Statistics Surveys. 4: 1–39. doi:10.1214/09-SS051. MR 2595125. PMC 2857732. PMID 20414472

---

Please explain better Fig. 11. Readers might not be aware of "polyn." abbreviation. Does it mean "polynomial interpolation"? If yes, please explain how you performed it, what type of polynomial you used etc.

Thank you for this comment, we agree that this abbreviation can be confusing without any description. “polyn.” was meant as polynomic regression. It’s parameters were found with a classic approach of least squares method and now are displayed in the graph as well as the unabbreviated table legend. We replaced the Fig. 11 where it is not used any abbreviation at all.

---

We would like to thank you again, we sincerely appreciate for your time and considerations. Thank you for your suggestions.

Author Response File: Author Response.pdf

Reviewer 3 Report

The paper is well written and it is original. The paper should be published after revising the following minor Changes.

Suggestion: A comparison table of UHF RFID Tags for UAVs could been added in the manuscript for clear understanding to the reader with help of recent literature.

Author Response

Dear reviewer,

Thank you so much for your valuable comments and recommendations. We really appreciate time you spent to make a review of our paper. Based on your remarks we have tried to do our best to enrich our article as you suggested. The changes are indicated in an updated version in “track changes” mode. We really hope that these modifications and addition of some parts will contribute to accept the paper.

Comments and Suggestions for Authors

The paper is well written and it is original. The paper should be published after revising the following minor Changes.

Suggestion: A comparison table of UHF RFID Tags for UAVs could been added in the manuscript for clear understanding to the reader with help of recent literature.

Thank you for this comment. We really appreciate the time you spent to review our paper. Thanks to your suggestion we added a table with various UHF RFID tags that can be used in UAVs.

---

We would like to thank you again, we sincerely appreciate for your time and considerations. Thank you for your suggestions.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

The authors have addressed most of my concerns, except for the critical flaw that I have identified related to the research design.

Also, the table with raw data was  moved to the annex, but it also remained in Section 3. It should be removed.

Author Response

Dear reviewer,

we really appreciate your comment and time you spent to make a 2 reviews of our paper. We are very grateful to you, thanks to your comments, we were able to enrich our article with valuable information.

We are really grateful to your opinion. We added some new part, we updated some passages as you suggested. At this updated paper, we have tried to clearly describe the solved issue in which a problem is first identified. We really hope that these modifications and addition of some parts will help to accept the paper.

As described in paper, we were forced to rely completely on experimental character data and its statistical comparison based on measured RSSI.

RSSI is an important, but often neglected, property of an RFID system. The reader analyzes the response and reports the tag’s data along with the signal’s RSSI. RSSI, or Received Signal Strength Indicator, is a measurement of the power received from the returned signal from an RFID tag when interrogated by a reader. 
When a reader reports a tag’s RSSI value, it is effectively reporting the power level of the tag’s backscattered response signal as it relates to the power level of the reader’s initial transmission signal. This power level is typically reported in decibels per milliwatt, or dBm. 

In UHF RFID applications, RSSI values are used to provide a general idea of how well a particular tag is responding in a read zone. When testing different tags, RSSI can be a good indication of how well each tag will respond in a similar environment. However, RSSI values do not provide an accurate measurement of the distance of the tag from the antenna in a passive RFID system. This is due to the many environmental factors that can affect a tag’s read range in a passive system; because of this, RSSI alone is not a reliable measurement to use when calculating tag distance.

Anyway, we would like to thank you again, we sincerely appreciate for your time and considerations. Thank you for your suggestions. We really hope that these modifications and addition of some parts will help to accept the paper.

Round 3

Reviewer 2 Report

I agree with the modifications brought by the authors.