The Rescuer’s Navigation in Metro Stations Based on Inertial Sensors and WiFi

Round 1

Reviewer 1 Report

The authors introduce a recent bio-inspired approximation algorithm to solve a navigation problem in a challenging environment: indoor and dynamic. The role of the bio-inspired algorithm, namely, dingo optimization (DOX), is in correcting the drift affecting target dead reckoning. Such a correction is governed by a particle filtering process that uses the WiFi fingerprint of the target. In such a context, the DOX algorithm is used to compensate for the weight loss that affects particle filtering in the long run. The authors provide experimental results to support the validity of their solution.

I am skeptical about the validity of the approach, so I ask the authors to provide details about the following aspects.

In the article, the authors note that the models used for the state and the observation have complementary errors: one tends to drift, while the other may mismatch the target. There is no evidence that such errors compensate for each other, and do not accumulate instead, especially in a crowded environment.

The paper introduces DOX to compensate for particle weight degradation. The association of the two techniques is computationally heavy, as the authors note in the conclusions. However, particle weight degradation affects long runs, while the scenario suggests short ones to avoid mismatches and keep computational load acceptable. The value of 60 steps in figure 5 should be justified. So it is questionable whether a particle re-population is needed at all.

The main contributions listed after line 51 are not convincing:

- The first one is of sure interest, but of marginal novelty

- Regarding the second point, it is not clear to me (see below) if the use of DOX brings a significant improvement. In addition, the point is not introductory, since the topic of particle filter is not yet introduced, and DOX adoption is consequential to particle filters.

- Finally, an experiment may justify the validity of a contribution, but it is not a contribution per se. In addition, an experiment with two items sounds very limited: a better explanation is worth it in the introduction, not among the main contributions.

In section 4.5.2 the comparison with APF (which should be included in the "Related Works" section if applicable) is poorly justified.

Summarizing I am not convinced that the suggested solution effectively contributes to solving the problem, for the reason that it introduces computational costs that are not clearly justified.

 

 

Author Response

Please read the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

This paper presents a DOA algorithm to address the particle impoverishment problem. The paper is well organized and its contribution is clearly described. 

 

I have some questions on how the experimental data was obtained.

1) Many figures have position error over time. Does that mean it is measured over a single path by traveling it once? There is not much information on experimental method. 

2) DOA is only on method for solving the particle impoverishment problem? I am wondering if you compare the DOA with other alternative method for   solving the particle impoverishment problem? 

 

Author Response

Please read the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The revision does not answer my questions. However, the added "contributions" (please fix the typo) section clarifies that the paper concentrates on theoretical aspects validated by experiments and suggests a use case that, to some extent, matches the theoretical framework. In that sense, the article appears worth publishing.

If the authors agree with my view, they may take it into consideration for improving the impact of their paper using the time left for a "minor revision".

Author Response

We have answered the reviewers' comments point-to-point, and if there are any answers that are not good enough, we hope that the reviewers will point out which questions are not answered.

Thanks for the reviewer’s valuable and constructive comments. We have revised the manuscript carefully, hoping to reach your expectation. Detailed modifications are listed below.
Comments from Reviewer #1:
The authors introduce a recent bio-inspired approximation algorithm to solve a navigation problem in a challenging environment: indoor and dynamic. The role of the bio-inspired algorithm, namely, dingo optimization (DOX), is in correcting the drift affecting target dead reckoning. Such a correction is governed by a particle filtering process that uses the WiFi fingerprint of the target. In such a context, the DOX algorithm is used to compensate for the weight loss that affects particle filtering in the long run. The authors provide experimental results to support the validity of their solution.
I am skeptical about the validity of the approach, so I ask the authors to provide details about the following aspects.
1.In the article, the authors note that the models used for the state and the observation have complementary errors: one tends to drift, while the other may mismatch the target. There is no evidence that such errors compensate for each other, and do not accumulate instead, especially in a crowded environment.
Reply: Thank you for your comments. The dead reckoning has high short-term accuracy and large long-term run error. Due to the ambiguity of fingerprints caused by the environment, there are mainly mismatches in WiFi fingerprint matching. Fusion algorithms combine the respective advantages of both. Since the error sources are different, the fusion algorithm can eliminate the error of a single navigation source. References [1] and [2] use fusion algorithms to eliminate individual navigation source errors.

2.The paper introduces DOX to compensate for particle weight degradation. The association of the two techniques is computationally heavy, as the authors note in the conclusions. However, particle weight degradation affects long runs, while the scenario suggests short ones to avoid mismatches and keep computational load acceptable. The value of 60 steps in figure 5 should be justified. So it is questionable whether a particle re-population is needed at all.
Reply: Thank you for your comments. It is difficult to solve the analytical solution of the posterior distribution of particle filter. Instead of solving the analytical solution, the approximate solution has been well developed. A large number of particles are used to approximate a posteriori distribution. To deal with the problem of a posteriori distribution in sampling, a resampling technique is proposed. After multiple sampling, the particles with high weight become less. The particle impoverishment problem is inevitable. How to solve the particle impoverishment problem is worthy of further research. When low-weight particles are generated, use the DOX operation to make the low-weight particles migrate to high-weight particles. Compared with the traditional particle filter, the introduction of DOX will increase the computational burden. But if DOX is not introduced, the particle impoverishment problem happens after many iterations, which leads to wrong position estimation. Reference [3] proposed an intelligent particle filter algorithm. 

3.The main contributions listed after line 51 are not convincing:
- The first one is of sure interest, but of marginal novelty
- Regarding the second point, it is not clear to me (see below) if the use of DOX brings a significant improvement. In addition, the point is not introductory, since the topic of particle filter is not yet introduced, and DOX adoption is consequential to particle filters.
- Finally, an experiment may justify the validity of a contribution, but it is not a contribution per se. In addition, an experiment with two items sounds very limited: a better explanation is worth it in the introduction, not among the main contributions.
Reply: Thank you for your comments. In the revised manuscript, we have made changes to our contribution to this work. Separate the contribution section into a separate section, and move the contribution section after the experimental section. 

The use of particle filters to filter out navigation position errors and improve positioning accuracy has been widely used. However, after resampling particles many times, there are fewer and fewer high-weight particles, and the particle impoverishment problem happens. To solve that problem, this work introduces DOA. The main innovation is the design of a group attack strategy, a persecution strategy, and a scavenger strategy. The particle search space is expanded, the diversity of particles is increased. The navigation performance of rescuers in metro stations is effectively improved. Another innovative point is that the solution uses inertial sensors and existing access points, which does not need to deploy additional infrastructure, and has the advantage of low cost.

Please read page 12 for detail in Section 5.

4.In section 4.5.2 the comparison with APF (which should be included in the "Related Works" section if applicable) is poorly justified.
Reply: Thank you for your comments. In the revised manuscript, we have moved APF to the related work section. 
Yu et al. [4] proposed an auxiliary particle filter (APF) to eliminate indoor pedestrian navigation position errors. APF uses floor plans to effectively reduce the cumulative error of DR. APF sets the weight of all valid particles to be equal, and sets the weight of invalid particles to zero, which is easy to cause the particle impoverishment problem. As the number of iterations increases, the number of high-weight particles decreases, which leads to larger position errors.

5.Summarizing I am not convinced that the suggested solution effectively contributes to solving the problem, for the reason that it introduces computational costs that are not clearly justified.
Reply: Thank you for your comments. Compared with traditional particle filters, the computational complexity is relatively increased. The reason for the increase in computational cost is that when low-weight particles are generated, DOA is used to allow low-weight particles to migrate to high-weight particles, which improves the posterior distribution of the particle filter, thereby improving navigation accuracy. 

[1].Xie, H., Gu, T., Tao, X., Ye, H., & Lu, J. (2015). A reliability-augmented particle filter for magnetic fingerprinting based indoor localization on smartphone. IEEE Transactions on Mobile Computing, 15(8), 1877-1892.
[2].Li, X., Wang, J., Liu, C., Zhang, L., & Li, Z. (2016). Integrated WiFi/PDR/smartphone using an adaptive system noise extended Kalman filter algorithm for indoor localization. ISPRS International Journal of Geo-Information, 5(2), 8.
[3].Yin, S., & Zhu, X. (2015). Intelligent particle filter and its application to fault detection of nonlinear system. IEEE Transactions on Industrial Electronics, 62(6), 3852-3861.
[4].Yu, C., Lan, H., Liu, Z., El-Sheimy, N., & Yu, F. (2016). Indoor map aiding/map matching smartphone navigation using auxiliary particle filter. In China Satellite Navigation Conference (CSNC) 2016 Proceedings: Volume I (pp. 321-331). Springer, Singapore.