An In-Vehicle Smartphone RTK/DR Positioning Method Combined with OSM Road Network

Round 1

Reviewer 1 Report

The technique is interesting, but the update switch (§2.4) lacks theoretical justification. The KF filter covariance matrix will not be affected by undetected bias, which will then ignore some errors that may contaminate the filter, as seen on Fig 4.b for example. The proposed switch seems to work, but the authors should insist that it is based on empirical testing, rather than on strong theoretical foundations.

Apart from that, the rest of the paper is interesting.

Here are some suggestions to improve the wording.

In the legend of the figures and text, change "base" to "reference"
In all the text, change "plane" to "horizontal"
In all the text, change "this paper counts" to "this paper analyzes"

abstract: "OSM" -> OpenStreetMap
p2: "According to research by Kong, Q" -> replace by reference [21]
p2: "method of the heading angle" -> method for the heading angle
p3: "altitude angle" -> elevation angle
p2: "To solve this problem, We" -> remove capital 'W' from "We"
P4 : "[B L H] represents for latitude, precision and" -> reprensents latitude, longitude and
p5 : "because of the acceleration noise" -> because of the accelerometer noise
p8: Eq(16) first outcome should be "straight line" instead of "curve" ?
p8: "or a length of more than 15 m": shouldn't you filter out short segments associated with potential turns? -> "length of less than 15 m"??
p9: "campus of Southeast University": add the name of the city
p10: in the hardware description, briefly mention the reference trajectory equipment (red square in Fig 6)
p11: "This paper also counts" -> This paper also analyzes?
p11: Fig 8 legend is wrong, should mention "after correction"
p12: "This paper also counts" -> This paper also analyzes?
p12: Table 1 could be discarded, as analysis of drifting errors by RMS is not very interesting. The overall result is already visible in Fig 7-9, and could be enough. If discarded, mention to Table 1 in the section "4. Discussion" should also be discarded and replaced by mentions to Fig 7-9.
p16: " the position There is a less than 1% drop ": sentence to be re-written

Author Response

Thank you for your suggestions. All your suggestions are very important.

The description of the measurement update switch has been revised in Section 2.4. In the fourth paragraph of Section 2.4 of the paper, it is indicated that the measurement update is a rough prediction of the horizontal position accuracy and serves as an empirical indicator.

Additionally, I've made the following modifications as you suggested:

1. In the legend of the figures and text, change "base" to "reference"
2. In all the text, change “plane” to “horizontal”
3. In all the text, change "this paper counts" to "this paper analyzes"
4. abstract: "OSM" -> “OpenStreetMap”
5. “According to research by Kong, Q” -> “According to the reference [21]”
6. "method of the heading angle" -> “method for computing the heading angle”
7.  "altitude angle" -> “elevation mask”
8.  "To solve this problem, We" ->  “In order to solve this problem, it is possible to”
9.  "[B L H] represents for latitude, precision and" -> “represents the latitude, precision and geodetic height”
10.  "because the acceleration noise" -> “because the accelerometer noise”
11. Eq(16) first outcome is also a curve. Eq(16) is only used to judge the curve
12.  "or a length of more than 15 m" -> “or a length of lower than 15 m”
13.  "campus of Southeast University" -> “campus of Southeast University in Nanjing, China”
14. In the first paragraph of Section 3.1, we added a description of the geodetic receiver -> “a geodesic GNSS receiver with a positioning accuracy of 3 to 5 cm”
15. "This paper also counts" -> “This work also analyzes”  In the second paragraph of section 3.2
16. Title of Figure 8. “before” -> “after”
17. Table 1 shows the heading and position accuracies of dead reckoning after heading correction. We want to show that this accuracy meets the conditions for combined positioning with RTK.
18. “the position There is a less than 1% drop” -> “there is a decrease of the forward accuracy about 1%”

Thank you again for your suggestions. They all have important guiding significance for my future scientific research work.

Reviewer 2 Report

1, RTK/DR loose combination of research value is open to question, when entering the tunnel, the satellite navigation signal failure, the DR navigation results are not reliable. The method proposed in the manuscript lacks innovation.

2, Lack of comparison with the most advanced methods. In addition, the experiment only uses the same brand of cell phone for comparison, which is not convincing.

In summary, the manuscript does not meet the publication requirements and is recommended to be rejected.

 

Author Response

Thank you for your suggestions. All your suggestions are very important.

1. I agree with this point: Smartphone RTK positioning in tunnels will not work, and INS or DR are also unreliable. In the past two years, three international competitions on smartphone positioning have been held, the primary purpose of which is to solve the problem of smartphone positioning in urban environments. The GNSS/INS/other fusion positioning scheme based on filtering or factor graph is still a hot spot in these competitions. This work aims to improve smartphone positioning accuracy in urban areas. We pay more attention to the practical value of the algorithm so that the designed positioning algorithm can be directly embedded in the smartphone APP. When the RTK positioning results jump, DR and navigation maps are the preferred ways to assist positioning. Other than that, I can not think of a better way.
2. The most advanced smartphone positioning method has yet to be determined. Among the more credible papers currently published in China, Professor Zhang Xiaohong published a paper stating that the dynamic positioning accuracy of smartphones can reach 0.5 meters in an open environment. The dynamic positioning accuracy of our smartphone positioning app is about 0.7 meters in an open environment. Since the test environment is different, more information is needed to determine which algorithm is better. Of course, I would like to compare it with the best smartphone positioning method in the world. This will motivate me for future research.
3. The smartphone RTK positioning method in this work cannot fix the carrier phase ambiguity, so the test results using multiple brands of smartphones are relatively close. In future research, I will borrow more smartphones to verify the algorithm.

Thank you again for your suggestions. They all have important guiding significance for my future scientific research work.

Reviewer 3 Report

Please, apply the changes highlighted in yellow colour in the attached file.

Comments for author File: Comments.pdf

Author Response

Thank you for your suggestions. All your suggestions are very important.

Following your suggestion, I have substantially revised the paper. Your suggestion has dramatically improved the English expression and terminology. They all have important guiding significance for my future scientific research work.

Reviewer 4 Report

Dear authors,

 

In this draft, experiments were conducted under three different conditions using two types of smartphones and their performance was compared in the actual driving process using the vehicle. The first is RTK-based GNSS positioning, the second is loosely combined with DR using an inertial sensor and the RTK-based GNSS positioning results, and the third is to estimate an accurate bearing angle by combining with OSM map information in a straight-line driving situation.

 

I agree to a certain extent with the contributions to be claimed in this draft, but I think it needs to be supplemented in various ways for publication.

First of all, we need to drastically improve English expressions and terminology. The use of variables that are not commonly used should be avoided, and even if commonly used variables are used, their definitions should be clearly described for variables that first appeared in this paper.

For example, we don't usually use a, b, and c in the H matrix. Also, to avoid confusion when a is already being used in a different sense.

The above problems can be seen in equations (3), (9), (11), (13), (17), and (18).

Also, the title of the picture should be re-corrected so that it can be a more appropriate title. In particular, in the case of Figure 8, "after correction" seems more appropriate than "before correction". Also, the pre/real shown in the legend in Figure 4 seems to require explanation in the text. According to the text, the green line is the plane variance and the red line is the RTK positioning error. It is necessary to review whether this is an appropriate expression and whether the y-axis is expressed as "plane error".

And Tables 3, 4, and 5 should be reorganized so that redundant information can be minimized, so that readers can clearly see only the necessary information.

 

In the technical part, the distance to the reference station that transmits the correction information of the GNSS RTK method used in the smartphone, the number of used constellations and satellites, etc. must be additionally specified.

 

Also, the ratio of time when measurement update is stopped due to RTK jumps is very high, and an analysis of whether or not this affects the overall performance should be added together. For this, it is necessary to analyze how much the position can be reliably calculated using DR even if no measurement update occurs during that time through the time between jumps of the RTK that occurs on average.

Finally, it should be mentioned whether the separation distance between the smartphone and the antenna for reference positioning is compensated.

 

Author Response

Thank you for your suggestions. All your suggestions are very important.

1. In the latest manuscript, I have drastically improved the English expression and terminology. For some variables that appear in the article, we replaced some confusing variables in equations (3), (4), (5), (9) and (13).
2. We modified the title of Figure 8. “before correction” -> “after correction”
3. We have added descriptions of Figure 4 for a better understanding. The green line (pre) represents horizontal variance which treated as a rough prediction of horizontal position error. the red line (real) represents the actual RTK horizontal position error of the smartphone. After the description is added, the title of the Y-axis may be more appropriate.
4. The original table 3 and table 4 are merged, in order to reduce data redundancy. Table 3 is sorted by test number, and Table 4 is sorted by smartphone model.
5. In the first paragraph of Section 2.1, we added the distance information from the reference station to the smartphone. In the last paragraph of Section 2.1, we added the constellation and satellite information used for RTK positioning.
6. For the measurement update switch, we have added the statistics of horizontal position accuracy, as shown in Table 5. We also added the analysis of Table 5. “the horizontal position accuracy of RTK during the closed measurement update period is lower than the overall accuracy, which shows that the measurement update switch can effectively identify the period with significant RTK error. In addition, the positioning accuracy of DR in each group of test is higher than that of RTK, which shows that DR can maintain reliable positioning when RTK positioning results are not good.”
7. In the first paragraph of Section 3.1, we added the description: “Moreover, the horizontal distance between the smartphone and the geodetic GNSS receiver is measured, which is used to calculate the precise coordinates of the smartphone.”

Thank you again for your suggestions. They all have important guiding significance for my future scientific research work.

Round 2

Reviewer 2 Report

The article still doesn't explain how to measure the antenna phase center of the mobile phone, so it can't be used as an exact reference. It is suggested to make it clear.

Author Response

Thank you for your comments and suggestions. They are very helpful to me.

A smartphone's antenna phase center is generally above its top midpoint. In static positioning, if the smartphone is placed horizontally, it is possible to use the position of the RTK fixed solution as the antenna phase center. During dynamic positioning, since the smartphone's pitch angle and roll angle is continually changing, the position of the antenna phase center is also continually changing. Measuring the continually changing antenna phase center is very difficult. Moreover, the smartphone used by the driver is placed close to vertically, so the influence of the antenna phase center on the horizontal positioning accuracy may be negligible. In future research, we will try to find a way to correct the antenna phase center. We added a description to the third paragraph of Section 3.1.

Thank you for your reminder, which dramatically improves the rigor of our research.

Reviewer 4 Report

Dear authors,

Thanks for revising the points I mentioned.

I think the manuscript has been improved significantly, but there are some important issues about the result in the manuscript.

1.  in Figure 4, it should be addressed that the prediction of positioning error is smaller than the actual error sometimes, especially in P40.

2. in Table 3 and 5, it should be addressed that accuracy performance between DR and RTK/DR is large in case of P40.

 

In addition, English expressions and formatting need to be refined even more. For example, "pure RTK" in abstract should be changed to "the RTK", unify the format using capital or not in Keywords, initialize or initialise

 

Best Regards,

Author Response

Thank you for your comments and suggestions. They are very helpful to me.

Precisely because the predicted positioning error is smaller than the actual error at some moments, we must add additional information to evaluate the positioning quality. Finally, we used RTK predictions of heading angle variation and positioning error to assess the positioning quality. We only need to judge the moment when the RTK positioning error is large and stop the measurement update at this moment. We have added descriptions above Figure 4.

We have added descriptions above Table 5. Since the RTK positioning accuracy of the P40 fluctuates more violently than that of the P30, the measurement update switch is triggered more times. Ultimately, the positioning accuracy difference between RTK/DR and RTK of P40 is more significant than that of P30. The reasons for the large difference in positioning accuracy between RTK/DR and DR are the same as above.

In abstract: “pure RTK” -> “RTK”

In keywords: “vehicle navigation” -> “Vehicle Navigation”

“initialize” and “initialise” seem to be equivalent. We use “initialise” uniformly.

Thank you again for your comments and suggestions. They all have important guiding significance for my future scientific research work.