Impact Analysis of Time Synchronization Error in Airborne Target Tracking Using a Heterogeneous Sensor Network
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsREVIEW OF
Impact analysis of time synchronization error in airborne target tracking using a heterogeneous sensor network
BY
Seokwon Lee, Zongjian Yuan, Ivan Petrunin, and Hyo-Sang Shin
The article presents the results of a practical study of a very interesting model of observing a moving object. The model specificity lies in the fact that it describes the receipt of measurements from many sensors under the assumption of a random error in time reference. Using the example of a simple moving of an uncontrolled aerial drone and active and passive radar sensors, the task of target tracking is solved. An extended Kalman filter is used together with complemented by well-known observation fusion techniques. Simulations were carried out on the simple model example, the results were analyzed, and recommendations were formulated.
The article is beautifully structured, well-designed, written neatly, in clear and concise language. The article needs to be published.
The reviewer's small suggestions for improvement or correction of errors.
1. There is no comma before the surname of the third author.
2. The authors tried to decipher all the abbreviations. Exceptions: in the keywords UAS (transcript below), there is no GNSS and GPS. That's it, that's it. It would be nice to explain the difference between GNSS and GPS. Just for the sake of accuracy.
3. Introduction. The proposal "Previous investigations have revealed that synchronization accuracy depends on various factors, such as the quality of the clock source, timestamp resolution, network topology, and environmental conditions (e.g. temperature) [21,22]." It would be nice to mention other reasons and features of the formation of the time of receiving data from sensors. Of course, the authors know that in addition to radar sensors, acoustic sensors are widely used, also for tracking unmanaged drones, but underwater, not aerial. The same tasks, the same tools, the same algorithms, but a different environment – aquatic. This gives factors such as temperature, salinity, and water pressure. Here are the details: eBook ISBN: 9780080982915.
4. Also in the introduction, one could mention other areas where there are, but rarely taken into account and poorly studied, random measurement arrival times. In addition to the mentioned underwater drones, these are also chemical processes, for example https://doi.org/10.1016/j.isatra.2018.11.004
5. Line 134. See the symbol /Delta bar.
6. In the same place, it is necessary to explain the measurements in miles and microseconds, including explaining where the standard deviation 62.293 comes from.
7. Line 203. The reference [34] is not too correct. [34] is an excellent monograph, but it is more correct to refer to the original source, and this is a work
Bernstein I., Friedland B. Estimation of the State of a Nonlinear Process in the Presence of Nongaussian Noise and Disturbances // J. of the Franklin Institute, 1966. V. 281. No. 6. P. 455-480.
8. Line 214, “covariance matrix". This is not true, it is only a heuristic estimation of covariance, the true value is unknown due to the nonlinearity of the model. Similarly, in line 218 “covariance posteriori”, only a heuristic approximation.
9. Using the extended Kalman filter is not very good to leave without comments. This is the most primitive non-linear filtering algorithm and strong arguments are needed to use it. They are here. First, it should be noted that for the model proposed by the authors (more precisely, for the more general case when the random time of receipt of measurements depends on the state of the object, for example, on the distance to it), an optimal Bayesian filter has been obtained https://doi.org/10.25728/arcRAS.2023.50.90.001 Its structure is such that it is impossible to talk about practical implementation, therefore other filters are needed.
10. It would also be very appropriate to mention other filters that work in models with random observation times. First of all, this is the mentioned work https://doi.org/10.1016/j.isatra.2018.11.004 , where the well-known unscented Kalman filter is used. Secondly, the work https://doi.org/10.3390/drones7070468, where the conditionally optimal Pugachev filter is used (Pugachev, V.S. Estimation of Variables and Parameters in Discrete Time non-Linear Systems; Autom. Remote Control 2018, 40, 4, 512-521), in which the random measurement delay time depends on the distance to the object and changes during tracking. By the way, in the tasks considered by the authors, the use of a conditionally optimal filter can be extremely useful, it is quite simple to implement it. The results may be much better, if only because this filter has guaranteed properties, and, unlike the extended Kalman filter, is guaranteed to be stable.
11. Formula (36). The first equality is missing the bold x. The last equality is missing the superscript asterisk of the x.
12. Table 2, line “Track maintenance/deletion [3, 5]/[3, 3] [3, 5]/[3, 3]”.
13. Line 356. Fig.??
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis paper explores the influence of time synchronization on sensor fusion and target tracking, specifically focusing on a target-tracking system built upon a track-to-track fusion architecture. It delves into the mathematical modelling of time synchronization errors and discusses how these errors affect various components of the target tracking system, including local association, filtering, and track fusion. The paper highlights that increasing time synchronization errors leads to performance degradation in association and filtering. Through simulation studies, the paper validates the impact of time synchronization errors in sensor networks, providing valuable insights for improving target tracking system designs. However, there are several lacunae existing in the present form of the manuscript. Most of the hypotheses and values are considered and placed in the manuscript without properly mentioning their source and description. It needs a thorough major revision to improve the quality from its present form. A few of the suggestions that may improve the quality of the article are as follows:
─ The concept proposed in this mathematical modelling and the approximations considered in the preliminaries need more elaborate descriptions. For example:
└ Authors have used the statistical data to model clock errors by assuming the mean and maximum values (422μs, and ≤ 1ms) without proper justification or appropriate citation.
└ Authors have used the offset bound as δ = 1ms, and the time error distribution as Δi ∼ N(mθ, σθ2) where mθ and σθ are used as 422μs and 62.293μs according to the statistical model. However, there is no proper justification or appropriate citation.
─ Even more relevant references are required to establish the validation of those assumptions.
─ The authors have not cited any recent references in this article. Following are some suggested references that should be added to the manuscript:
└ 10.1016/bs.adcom.2023.04.006
└ 10.1109/JIOT.2023.3282202
─ There are some inaccuracies within the tables, figures, text, and sentences that are unclear. A few examples are as follows:
└ Table 1 has some inaccuracies such as the sensors and parameters mentioned. Also, it shows different frequencies considered for Radar and RF sensors.
└ Line #356 has a typographical mistake in the sentence, “statistical model (Fig. ??).”
└ From Figure 4 to Figure 8: some of the legends are missing in their sub-figures.
└ The caption of figures and tables needs throughout revision for
Comments on the Quality of English LanguageExtensive editing of the English language is required to eliminate the typographical and grammatical errors present in the manuscript.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Comments and Suggestions for AuthorsMy concerns have been resolved.
Comments on the Quality of English LanguageMinor editing of the English language required to thoroughly cross-check the grammatical and typographic errors.
