Motion-Constrained GNSS/INS Integrated Navigation Method Based on BP Neural Network
, Cheng Yang 4, Dun Liu 5 and Haiping Zhang 6Round 1
Reviewer 1 Report
Dear author,
In your article „Motion Constrained GNSS/INS Integrated Navigation Method Based on BP Neural Network”, you show how a BP neural network can support the navigation problem of a combination of INS and GNSS, while outages of GNSS occur. As the neural network can fail, constraints have to be applied.
I have some major issues:
1. References are not found
2. No abbreviations in the abstract
3. English language: Please go through the whole text and check your phrasing.
4. Figure 1 is very general, this does not help much, if you do not apply it to your problem. So how many hidden variables are necessary, what are your x and y, please already use it in the figure?, with what do you compare DPGNSS? how does the error back propagation work? Did you use the CNN? How does CNN work, what is different to BP?
5. How often are constraints used? And which? It seems that the elevation constrain is not very effective as you use the same for ships and cars.
6. You can also test the sensitivity of your algorithm on the constraints on real data.
7. Conclusions: how general was the experiment with the ship? Calm sea or rough? Do you need different constraints for different environment? How general are the values you get for different situations in cars and ships? What does the values you get tell me?
Minor:
Throughout the text: “..when GNSS outages” is not correct. Use while, during, when outages occur,….
Page1: ..within short time” better for short integration times
Page 2: et al. not for a listing
…how to keep no good English
… which effectively improves…
… which pointed out… better setting the velocity to zero, when no jumps….
…23 used a multi-layer…
Page 3: “wild value” use a better word
… constraint in processing moving vehicles…
…h is also a system.. why also?
Page 4: P is the covariance …
… network that is based on an error back….
… output of a neural network….
… when GNSS signals are available…
Figure 2: coordinate system missing
E is the expectation value
Use a different letter for height h than for hidden variable h
… vehicle state of motion…
Page 6: Not …firstly but first
Page 7: MC –BP full name in headline
… the flowchart of the algorithm proposed in this paper…
Page 8: et al. again
1. Parag. of chapter 3: many plurals should be used
… firstly again
What is a pseudo-random error in this case?
No Sentence: The MEMS inertial…..
… to simulate GNSS outages, only the inertial navigation…
Page 9: …the GNSS was artificially interrupted artificially…
Page 10: … which is proposed in this paper…
Table 3: why same constraint in elevation for ships and cars?
Best regards.
Author Response
We would like to thank you for your kind suggestions. Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This manuscript is very interesting. There are some points should be revised. Followings are the details:
Comments for author File: Comments.pdf
Author Response
We would like to thank you for your kind suggestions. Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Manuscript:
Motion Constrained GNSS/INS Integrated Navigation Method Based on BP Neural Network
submitted by:
Ying Xu, Kun Wang, Changhui Jiang, Zeyu Li, Cheng Yang, Dun Liu, Haiping Zhang
GNSS/INS integrated navigation system has been widely used in the Intelligent Transportation System. However, the positioning error of integrated navigation system rapidly divergent when GNSS outages.
The structure of the manuscript is considered and clear. In the introduction, the background and comprehensive review of the problem's literature were presented. The Authors present GNSS/INS Classical Loosely Coupled Integrated Procedure and Motion Constrained GNSS/INS Integrated Navigation Method Based on BP Neural Network. In Experiment and Analysis section, the Authors present real data experiments divided into two experimental scenes:
land and offshore. Land experiment was realised onboard a car described i the manuscript. But there is no information about a vessel.
I can not belive, that a vessel (Table 1) was going in 5cm/s - what does mean normal velocity? SOG (speed over ground)? 1 m/s (2kn) is very low speed and it is hard to manoeuvre them. Roll amd pitch can be measured in presented accuracy using MRU sensors used in hydrography in MBES systems.
Figure 8b presents the vessel trajectory about 2000m. Because the speed was not defined, total time of the experiment is unknown.
Specific suggestions (the mamuscript does not have numberedlines:
Page 1: BP abbreviation is not defined
Page 1: Shared should not be written in capital letter
Page 6: EKF is enough, in the manuscript is: EKF filter - it means extended Kalman filter filter
Page 8: Cpprdinates are described in following form: 36.0264°N, 120.3045°E, 0. Because the marine navigation is on the horizontal surface the last zero can be deleted
Page 8, Figure 6a: it is hard to understand, how long the trajectory was. Easier is to use rectangular coordinates or line scale
Page 8, Figure 6b: is enougg to write pitch and roll (without angles). It is unreal, that pitch and roll are in presented form
Page 8, Figure 6: Please standarize: Latitude(o), Lomgitude/(o), Pitch, roll/(o), Time/s - sometimes is "/", sometimes is "()"
Page 8 (above the figure): the eror is 0.05cm? a half of 1mm?
Page 10: Figure 9 - there should be also a vessel
Page 11: 3.2.2 R3.2.2 Results Analysis... R3.2.2. should be deleted
Page 11, Figure 10: the Y axix should be corrected: positioning error is enough, because the figure is well described below.
Please take into considetation changing the time described in seconds x105 to minutes. 1.149x105 is hard to understand (figure 10-13)
The maximum positioning errors in the north, east and up direction are 296.21m, 336.70m and 24.61m respectively. Are results good in the Authors opinion?
Author Response
We would like to thank you for your kind suggestions. Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear author,
thank you very much for the updated version. It is now much better to read. I have only minor further comments:
line 20: is rapidly divergent or rapidly diverges
line191 "is" missing
line 201 ... and the actual GNSS position increment
line 207 ... was the gradient decent method
figure 4 please indicate the difference between a and b in the figure caption
line 268 an If is missing.
Best regards.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Manuscript:
Motion Constrained GNSS/INS Integrated Navigation Method Based on BP Neural Network
submitted by:
Ying Xu, Kun Wang, Changhui Jiang, Zeyu Li, Cheng Yang, Dun Liu, Haiping Zhang
Second review
I would like to thanks the Authors for taking into consideration my critical suggestions.
Figure 9 is complited by a photo of a boat. Please note, this is not a ship (military). I respected, measurements was realisen onboard a vessel, but it is a boat.
Table 2: Positioning acuracy (GNSS) is different in horizontal and vertical directions. If you use only horizontal , like during hydrographic surveys using vessl, it should be dfined: horizontal positioning accuracy (or vertical, or both of them)
Table 3: if normal velocity means vertical one, maybe it should be corrected
It is enough to use pitch and roll, with no angles. Please read your response to reviewer
Elevation is described in degrees. Like latitude and longitude, azimuth and elevation are angles. In RPH (roll-pitch-heave) sensors, roll and pitch are angles and heave is linear aling Z-axis
Please check, if lateral velocity in offshore constraint velocity should be " / "
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
