@@ -32,7 +32,7 @@ Currently we consider only the attenuation per floor, however by including infor
Instead of providing those additional environmental informations by manual measurements, the optimization scheme could be used to approximate the respective model and material parameters.
Special data-structures for pre-computation combined with online interpolation might then be a viable choice for utmost accuracy that is still able to run on a commercial smartphone in real-time.
Finally, the \del{rapid computation}\add{approximation} scheme for the KDE opens up completely new possibilities when handling particle sets.
Finally, the \del{rapid computation}\add{approximation} scheme for the KDE is \add{capable of offering} completely new possibilities when handling particle sets.
Within this paper we used it to find the real global maxima for a state estimation and to accurately calculate the Kullback-Leibler divergence.
However, many other estimation schemes are thinkable, for example a trajectory based one, with multiple path-hypotheses, each weighted based on a-priori knowledge.
The KDE approach could also be used to develop better suited resampling techniques, by enabling to draw particles from the underlying density, instead of just reproducing known owns.
@@ -558,7 +558,7 @@ We hope to further improve such situations in future work by enabling the transi
To summarize, the KDE-based approach for estimation is able to resolve multimodalities.
It does not provide a smooth estimated path, since it depends more on an accurate sensor model than a weighted-average approach, but is suitable as a good indicator about the real performance of a sensor fusion system.
At the end, in the here shown examples we only searched for a global maxima, even though the KDE approach opens a wide range of other possibilities for finding a best estimate.
At the end, \add{we only used the KDE approach to provide a global maxima, even though it} opens a wide range of other possibilities for finding a best estimate.
\add{A detailed examination of the runtime performance of the used estimation methods in comparison to the state-of-the-art can be found in \cite{Bullmann-18}.}
@@ -31,6 +31,6 @@ The paper presents a smartphone-based localization system using a particle filte
-> At first, "within the 2500m2 building" was a bad misformulation, as the building is far bigger. The 2500m2 refers only to the area that is actually walkable by the visitors. As can be seen in figure 7, the building has a very large courtyard in its center. A second reason for the high number of beacons, are the very thick walls. To prevent to much attenuation, we tried to install at least two beacons per room and a third one in an approximate radius of 10 meter. As said before, this was done very quickly without analyzing the Wi-Fi coverage. As the beacons are very cheap (less then $10), they represent only a small part of the total cost of the system. They only require a power source in order to operate, which keeps the need for additional infrastructure small. Furthermore, we believe that a janitor is able to set up our system independently. This means that there is no need to pay an external contractor to utilize the system and only the hardware costs and, if applicable, the price of the software have to be calculated. Nevertheless, these considerations could not apply to all buildings and scenarios, which is why the property "low cost" is removed. Please also refer to line 563 to 586. The number of reference points (133) was added to the text. The term "fast to deploy" is discussed in great detail at the beginning of the experiments. Experiments providing the impact of the density of the reference points, as well as the access-points can be found in our previous paper, ""
Thank you again for your time and the good suggestions to further improve this work.
-> Thank you again for your time and the good suggestions to further improve this work.
@@ -56,3 +56,6 @@ Figure 6: You use the expression "Monte Carlo", are you referring to Condensatio
Results section: results and comments are ad-hoc for this environment, and it is not demonstrated that could be applied in a more general context.
-> We incorporated this suggestion into the conclusion section of this work. Please see line 961 to line 959. However, we have decided not to discuss the general usage of this approach in the experiments, since this work explicitly deals with a historical building.
-> Again, thank you very much for your time and the good suggestions. They further improved this work.
-> At first we would like to start with a short overview over all changes. The individual answers follow directly after this text. All additions to the text are highlighted in blue. Words or text passages suggested by the reviewers to be removed are highlighted in red. As you will see, abstract and introduction are completely revised to better highlight the novel contributions. We were able to implement many suggestions of the reviewers. The transition was highly extended, to achieve a better understanding of the method. We have also added a detailed description of how and by what means our system is installed in a building. This also leads to a better description of the experimental setup.
We added a complete new section, evaluating the activity recognition. Additionally, you will find mind smaller changes and addition throughout the paper as well as further improvements of the writing. In the following our answers are marked with "->".
The paper presents an improvement to a previous work of the authors where a transition, model, an activity recognition method, a recovery method for the particle filter, and an improved density estimation.
The novelty of the paper was collected in the reading and it should be more clearly listed. At the current status of the paper it is not clear. It should be itemized in the abstract and also in the obtained results.
-> Thank you very much for this advice. We tried to clarify this within the abstract as well as at the end of the introduction (see line xxx to xxx). We further added a discussion to the results, which addresses the contributions and their impact on the system.
-> Thank you very much for this advice. We tried to clarify this within the abstract as well as at the end of the introduction (see line 84 to 99 and line 9 to 17). We further added a discussion to the results, which addresses the contributions and their impact on the system.
The rapid computation declaration is not proven, given that the authors do not compare the non-gridded approach timings.
-> The terminology "rapid computation scheme" only refers to the state estimation process, not the underlying graph or the complete system performance. It seems this is not clearly formulated within the paper. The weighted-average estimator yields faster estimates of the position compared to the KDE approach as we have shown in our previous work .... This previous work does also provide an extensive comparison between other state-of-the-art KDE approximations.
However, if you refer to the comparison between gridded graph and navigation mesh, ...
Nevertheless, ...
-> The terminology "rapid computation scheme" only refers to the state estimation process, not the underlying mesh or the complete system performance. It seems this was not clearly formulated within the paper. Thanks for pointing that out. We tried to clarify this in different parts of the work, especially in the introduction. We also removed the terminology "rapid computation" and instead called it "approximation" scheme.
For clarification, the weighted-average estimator yields faster estimates of the position compared to the KDE approach as we have shown in our previous work "Fast Kernel Density Estimation using Gaussian
Filter Approximation". This previous work does also provide an extensive comparison between other state-of-the-art KDE approximations.
Does the system will also work in regular buildings? A final comment on the lessons learned in this case of the 13th century building should be in the conclusions, given that the title focus on this very specific context.
->
-> Thanks for this suggestion, we added an additional paragraph to the conclusion (see line xxx to xxx). We further included a general description on how the system can be utilized to a building. We hope this leads to a better understanding of the possibilities this system offers.
From the middle of the paper the quality of the English decreases with several sentences with errors, some of them were identified.
-> We have fixed several errors on our own as well as the recommendations made.
-> The spelling as well as the writing was improved throughout the complete work. Thanks for the many recommendations made.
Given that the authors declare that this is an update they should be more clear on what was already done. We also consider that the authors use auto citation excessively (7 in 32) and or compress the self-citation or increase the comparison with other works.
->
-> Based on this and your comments on the novelty, we completely revised the abstract and the introduction of this work. See again line line 84 to 99 and line 9 to 17. Thanks again for this suggestion, it really improves the work.
The paper is relevant and should be published after a clarification on the produced work and at least report two comparisons, one in the performance of the KDE and other on the initialization time.
->
-> Thanks for the kind words. We added a detailed description regarding the initialization time at the beginning of the experiments. As the computational performance of the KDE was already discussed within previous work and this work already evaluates the accuracy, we added an additional experiment to evaluate the activity recognition.
Some notes follow along the lines of the paper:
Ln 1- "of our award-winning"
Ln 1- "of our award-winning"
The authors should refrain to be excessive in the connotation of the work, at least in the abstract. Should not hide but can declare the award in a more soft way.
-> We have removed the remark and have formulated the remaining text passages more modestly.
Ln 9-…
"Continuous and smooth floor changes are enabled by using a simple activity recognition. Our rapid computation scheme of the kernel density estimation allows to find an exact estimation of the pedestrian’s current position. We further tackle advanced problems like multimodal densities and sample impoverishment (system gets stuck) by introducing different countermeasures, leading to a more robust localization." Too many adjectives without justification: simple, rapid, advanced...
"Continuous and smooth floor changes are enabled by using a simple activity recognition. Our rapid computation scheme of the kernel density estimation allows to find an exact estimation of the pedestrian’s current position. We further tackle advanced problems like multimodal densities and sample impoverishment (system gets stuck) by introducing different countermeasures, leading to a more robust localization." Too many adjectives without justification: simple, rapid, advanced...
-> You are right and we have rewritten the particular text-block.
Ln 14: Why low cost solution? Are the material less expensive compared to what alternative? If the authors do not want to defend this property we suggest it should be removed. (also in Ln 64)
-> The argument that our system is inexpensive is based on several considerations: With under 10 dollars per piece the Wi-Fi beacons are very cheap compared to conventional access points.
They only require a power source in order to operate, which keeps the need for additional infrastructure small. Furthermore, we believe that a janitor is able to set up our system independently. This means that there is no need to pay an external contractor to utilize the system and only the hardware costs and, if applicable, the price of the software have to be calculated. Nevertheless, as you suggested correctly, these considerations could not apply to all buildings and scenarios, which is why the property "low cost" is removed.
Ln 21: “optimization scheme enables a setup-time of under 120 min for the complete building. ”- Should indicate the mapped area.
-> Fixed in line 26.
Ln 38: “There is also a higher chance of detecting false or misplaced turns,”
The sentence appears incorrect
-> Fixed in line 43.
Ln 40: “presents a robust but realistic movement”
Too many adjectives not explained
-> Fixed in line 45.
Ln 52: "This leads to problems for methods..." This phrase should be reformulated since Wi-Fi fingerprinting is RSSI based and might actually benefit from the high signal attenuation between different rooms. The problem is probably coverage of the whole building, not parameter estimation.
-> This line was written with signal strength prediction models in mind, which is wrong in terms of fingerprinting, of course. Thank you for the hint! We tried the incorporate both, coverage and parameter estimation between line xxx and xxx.
-> This line was written with signal strength prediction models in mind, which is wrong in terms of fingerprinting, of course. Thank you for the hint! We tried the incorporate both, coverage and parameter estimation between line 56 and 83.
Ln 58 of simple and cheap- How simple and how cheap?
-> The price mostly depends on the reseller, ranging from 3 dollar to 10 dollar. The term "simple" is of course difficult to substantiate, which is why it was removed.
Ln 160 “are based on the nature of particle filter.” -> “are based on the nature of a particle filter.”
-> Fixed in line 180.
-> Fixed in line 203.
Ln 215: "Using variable shaped/sized elements instead of rigid grid-cells provide both higher accuracy for reaching every corner, and ..." Is accuracy the right word here?
-> We completely changed the transition part, describing how both, graph and nav-mesh are generated automatically, based on the building's floorplan.
We hope the new description points out why navigation meshes allow for a better (and more exact/accurate) representation of the building's floorplan compared to the grid-based approach.
-> We completely changed the transition part, describing how both, graph and nav-mesh are generated automatically, based on the building's floorplan. We hope the new description points out why navigation meshes allow for a better (and more exact/accurate) representation of the building's floorplan compared to the grid-based approach.
Ln 228: "If the destination is unreachable, e.g. due to the walls or other obstacles." This phrase is incorrect, the authors should reformulate it.
-> Thank you for noticing! We adjusted the sentence accordingly.
-> Thank you for noticing! We adjusted the sentence accordingly. See line 308.
Ln 237: "...the average acceleration..." This includes both linear acceleration and gravity, use "linear acceleration".
-> we rephrased the complete paragraph. It should now be clear how the current gravity readings are used to
determined the phone's current orientation, to undo the rotation, present within the gyroscope's readings.
-> We rephrased the complete paragraph (line 315 to 322). It should now be clear how the current gravity readings are used to determined the phone's current orientation, to undo the rotation, present within the gyroscope's readings.
Ln 258 - This equation needs revision. Should it be "p(s_i|p) ~ N(u_i,p , std²_wifi)" ? Also the wall-attenuation-factor-model only takes into account attenuation by floors, not walls.
-> No, the equation is correct. Its the actual >result< of the normal distribution when questioned for the received s_i, given the model prediction was u_i,p with uncertainty \sigma^2_wifi
-> The equation is correct. Its the actual >result< of the normal distribution when questioned for the received s_i, given the model prediction was u_i,p with uncertainty \sigma^2_wifi.
-> We now made clear that our model is something in between the log-distance and the wall-attenuation factor model. To reduce computation time on the smartphone, only floors/ceilings are considered
as this can be achieved without costly intersection tests. We also pointed out, that including walls would be more accurate, but is costly during runtime (intersection-tests).
Ln 271-272: The authors mention that their WiFi fingerprinting approximation process is faster than classical fingerprinting, but they fail to provide a reference for an example of the latter or significant metrics such as the average time per square meter for fingerprinting a whole building. Furthermore, the authors should also take into account that while there are approaches where reference measurements are recorded on small grids between 1 to 2m, there are also approaches able to record reference measurements using faster methods. One example is walking by the building while registering ground truth points and using dead reckoning techniques (see Guimarães, V. et al. A motion tracking solution for indoor localization using smartphones. In Proceedings of the 2016 International Conference on Indoor Positioning and Indoor Navigation (IPIN)).
-> Die Kritik ist berechtigt, weswegen diese Anmerkung entfernt wurde.
-> Your criticism is justified. Between line 171 and 178 we now introduce the paper you suggested as a faster solution for recording fingerprints and discuss this method in context of an historic building. The topic is then again addressed in line 364 - 367, regarding our optimization scheme.
Ln 275 - Equation 9 The d0 parameter of eq.8 shpuld also be presented in eq.9.
-> Fixed in line xxx. (Is usually assumed to be 1 and thus omitted)
-> Fixed in eq. 9. (Is usually assumed to be 1 and thus often omitted).
Ln 307 “Activity Recognition ” The threshold approach should have reports on the detection capability
-> TODO: Also die pfade nochmal berechnen und da die erkennungsrate ausspucken.
-> This small recommendation produced section 7.4. Thanks for this advice, we are very happy with the result.
Ln 316 - Equation 10 According to these rules a user could be standing and walking at the same time. The algorithm would be better represented by a flowchart or a decision tree.
-> You are right. Standing and walking at the same time should not supposed to be happen. We exchanged the equation with a decision tree. In our opinion the algorithm is now very easy to understand, thank you for the good advice.
-> You are right. Standing and walking at the same time should not supposed to be happen. We exchanged the equation with a decision tree. In our opinion the algorithm is now very easy to understand, thank you again.
Ln 316: t_baro = 0.042 m/s² ? Wrong unit, please confirm value and unit again.
-> Of course, m/s² is the wrong unit for pressure. Changed to hPa.
Ln 360 “It is obvious, that a computation of the probability density function of the posterior could solve the above, but finding such an analytical solution is clearly an intractable problem, which is the reason for applying a sample representation in the first place.” The authors should refrain to make non scientific comments like “obvious” or “clearly”. Do direct statements.
-> The objection is valid and we will fix the appropriate text passages.
-> The objection is valid and we will fix the appropriate text passages. See line 452 ongoing.
Ln 381 “Our rapid computation scheme” – As commented before. Be more concrete and remove non justified adjectives.
-> As before, the adjective "rapid computation" was removed in behalf of approximation scheme.
Ln 453 “By utilizing it to a 13th century historic building”-> “By utilizing the proposed technology in a 13th century historic building”
-> Fixed as recommended.
Ln 469: "8 cores..." The i7-4702HQ has 4 cores and 8 threads, not 8 cores.
-> This misstatement has been fixed.
Ln 551 - “above are more moderately attenuated” revise.
-> More understandable and correctly formulated in line xxx.
-> More understandable and correctly formulated in line 718 to 722.
Ln 524 - Figure 4 "...xz plane..." should be xy plane
-> Fixed.
-> Fixed in figure and in the text.
Ln 578 “This allows us to discuss everything in detail” – irrelevant or should be better expressed.
-> Removed.
Ln 711 - “The KDE-based estimation illustrates this behavior very accurate” – Revise
-> This sentence was very daring and incomprehensible. We tried to explain the situation in greater detail. Please refer to lines xxx to xxx.
-> This sentence was very daring and incomprehensible. We tried to explain the situation in greater detail. Please refer to lines 921 to 935.
Ln 713 – “the teal square” – It is not a square.
-> Replaced throughout the whole work by rectangle, area or border depending on the use.
Ln 720 – “In the end, it is a question of optimal harmony between transition and evaluation.” This is sentence is not providing information and lacks scientific objectivity.
-> The sentence was removed, as it was irrelevant.
Ln 726 “At the end, in the here shown examples we only searched for a global maxima” – revise english
-> The sentence was fixed in line xxx.
Ln 733 – “Compared to other state-of-the-art solutions, the setup time is only a few hours and does not require any expert knowledge or hardware.” This comparision is not done in the paper. The authors should be more specific given that this property of the purposed system is declared important, and so should be given with ore detail and (m^2 per hour) and compared with the coverage time of other systems.
-> As mentioned before, we added an extensive discussion to the beginning of the experiments. Here, the setup-time as well as the complete initialization process is clarified. However, we were not able to compare the coverage time with another system, as this would have required to utilize another solution to the museum. We are sorry.
Ln 754 – “opens up completely new possibilities when” – revise tone of writing.
-> The sentence was fixed in line xxx.
-> Again, thank your very much for this very detailed review. Your suggestions have really improved this work and we hope that you are as happy as we are about the result.
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
toni