Tight Integration Kinematic PPP-AR Using GPS/Galileo/QZSS Overlapping Frequency Signals and Its Performance in High-Shade Environments

Round 1

Reviewer 1 Report

This research investigates the PPP-AR with tight integration among different GNSS systems. The tight integration is widely discussed in RTK, however, it is still not applied in PPP-AR. Therefore, I believe this research is scientific sound with novelty. However, after reading the while manuscript, there are many doubts from me on the method and design of the experiment.

Major comments:

1 This research discusses the tightly integrated PPP-AR. I think the main contribution of the method is when the satellite number is few in a highly-shade environment. However, the authors demonstrated it with a simulated static data with high quality data. I think the environment and the data is too ideal and can’t exposes the disadvantage of this method. I strongly recommend the author to use a real-kinematic data, which is more persuasive.

2 I agree that the ISRCB and ZTD are stable over a short period. However, in the experiment design part, they are fixed for a long time (over 5 hours). I think it is not suitable since they have already changed a lot after so long time.

3 The authors say that the integer ambiguities from the last epoch are used as constrains in the current epoch (the so-called fix and hold mode), especially in a high-shaded environment. According to my experience on this kind of data, it is a HIGH risk to use a fix and hold strategy, since it is very possible to get a wrong fix ambiguity in such environment. Again, if the authors don’t agree with that, they should try with real high-shaded data.

4 Although tightly integrated method improves the success fixing rate, which is based on the positioning mode, it doesn’t improve the correct fix rate, and it is also important to present the positioning accuracy. More positioning results and discussions should be presented.

Minor comments:

1 L91-94. (1) Why the estimation methods of DISB are completely different in PPP-AR? (2) I think relative positioning utilizes double differenced uncombined observation. (3) “estimating the DISB parameters in observation equations is not efficient and appropriate as relative positioning”, what is the meaning of this?

2 L124-125. Why the difference in receiver clock between GPS and other systems can be estimated as a constant over 30min? I believe it is better to estimated as a random walk parameter.

3 L238. When the observed satellite number is 3g, why the traditional AR approach is impossible?

4 Title of section 2.3 should starts with a new line.

5 Eq.(17), what is the meaning of dWL?

6 L319. If we have 9 satellites (3 for each system), why can’t we update the ISRCB and ZTD parameter?

7 Figure 1. It is not shown where the WL DISB and NL DISB is used.

8 The section number is not used after datasets.

9 L351. Why the threshold for AR is set as 1.8?

10 L407. I agree that using the DISBs estimated at the last epoch is reasonable. But is it still reasonable after a long time with the same DISBs?

11 L410. The authors explain that the poor stability of NL DISBs is due to the satellite-related or environment-related unmodeled errors. Why? I think the orbit, clock, and UPD has a high precision, and the environment-related error has not large impact since we are using an open sky data, the NL DISB should be quite stable (see: https://doi.org/10.1016/j.asr.2018.12.008)

12 L426. Again, why the ZTD is kept fixed. It is enough to estimate ZTD with 3 GPS, 3 Galileo, and 3 QZSS satellites. Also, it is not suitable to fix ZTD after 5 hours.

13 L452. “When the number of ob-451 served satellites is further reduced, as shown in Schemes 2–9”. In Scheme 2, the satellite number is not reduced.

14 The expression of the English should be improved.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors present the tight integration kinematic PPP-AR method using the overlapping frequency signals. The differential intersystem bias (DISB) is used and the related experiments are conducted. The topic seems interesting, but some points need to be addressed.

Major comments:

The DISB concept is used and as far as I am concerned it’s similar to the traditional ISB method. What worries me is that the discussion of the difference and improvement of the DISB compared with the ISB should be given. Also, the corresponding experiments also needs to be conducted if necessary.

Minor comments:

1.     Abstract: two “system” appears and needs to be revised.

2.     Introduction: ‘dual-frequency’ is commonly used rather than the “double-frequency”.

3.     Datasets and evaluation indicator: “UTC” can be used.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Dear Authors,

I red your manuscript with interest, as intersystem ambiguity resolution is not a frequent topic up to now. My rating of the paper is quite good but there is still need for an extensive editing concerning english grammar and also some explanations are still required to make the manuscript more understandable.

my comments below refer to the line numbering scheme; technical questions are flagged with a 'T'

1) l10

'.. navigation satellite systems (GNSS) broadcast..'

2) l12

'.. to realize a tight integration is..'

3) l13

'.. intersystem biases (DISB), which allow to retrieve the integer..'

4) l33

'.. were corrected and retrieved...'

5) l47'  'T*

here you refer to the 'latest generation' as GPS IIF, but the latest GPS generation is 'IIIA'

6) l56

'.. the same central frequency.'

7) l80

'.. PPP-AR methods are ...'

8) l92

'.. utilizes undifferenced ionosphere-free...'

9) l93 'T'

here you are taliking about 'double-differenecd unconbined observations'

but when differencing you already combine observations; do you mean

that you do not build linear combinations?

10) l94 is hard to interpret ->re-phrase

whats not appropriate in contrast to relative positioning ?

11) l96

'This may evolve to an option for future applications currently based on the traditional PPP-AR method.'

12) l98 and sevearl times more in the paper

please replace the word 'occluded' with 'obstructed' !!!

at least 10 times in the paper

13) l98/99

'.. experiment conducted in an scenario with an extremely low number of observations highlights tight ...'

14) l106

'.. we present the processing steps required as long as the receiver is able to gather  a good number of observations, such as ...'

15) l110

'This includes accounting for a change of the reference satellite and therefore UPD basis, imposing...'

16) l113

'.. conducted in a highly shaded observation...'

17) l120

here you refer to subscript i but in the equations (1) solely the subscript IF shows up

18) l123 'T'

here you refer to the wavelength of the 'IF linear combination; provide a guess of this wavelength in cm

19) l138

'.. satellite-dependent UPDs, respectively.'

20) l143

'.. sent to the user together with precise...'

21) l145 'T'

here you mention that the 'PPP calculation was completed'; I assume that this PPP coordinate calculation shall take place at the rover=user site ?

22) l147

'.. the same GNSS can be utilized to calculate a single..'

23) l155

'.. and the NL ones.. '

24) l156/157

'.. the fixed coordinates are stored.'

25) l166/167

'.. of receiver-related hardware biases: the code-range and phase-range hardware biases. The code-range part is introduced..'

26) l170

 'When SD ambiguities are calculated using two phase observations from the ...'

27) l172

'.. ambiguities of different systems.'

28) l174

'.. another problem are varying hardware..'

29) l184

'.. coordinate solution is calculated to save...'

30) l189 'T'T

I am not sure why you mention here 'is the before-corrected WL ambiguity'

my understanding here is that its the raw float WL ambiguity which has to be corrected afterwards ? slightly misleading

31) l191

'.. should be the fractional part of ..'

32) l196

'.. final hardware delay is the weighted...'

33) l199

'.. The WL DISB between systems...'

34) l201

'In this study GPS was chosen as reference system B. The...'

35) l220

'Moreover, after constraining the fractional parts..'

36) l226

'.. The NL DISB between system A...'

37) l228

'Similarly, GPS was chosen as the reference system B. The...'

38) l229

How do you mean the sentence 'When g systems are ...'?

Do you mean that g-1 quantities have to be stored ? -> re-phrase

39) l231/232

  this sentence is also problematic; maybe you mean:

'As long as enough observable satellites are available the above mentioned processing steps are conducted every epoch. On the other hand in shaded/Obstructed environment the number of observable satellites might diminish to no more than 3g. Taking...'

40) l237 'T'

are the six mentioned SD ambiguities because of the low number of only 9 observations? detail the assumed scenario here more clearly

41) l241

this seems to be the header line of chapter 2.3 - re-format

42) l261

'.. where x covers the estimated...'

43) l262

'.. are the before-constrained..'

44) l263/264 'T'

explain whats the exact difference between Qy,Qx and Qxy; its clear that Qx and Qy are the covariance matrices of vectors x and y , but whats the exact difference to Qxy?

45) first sentence chapter 2.3.2 , l266/267

'Generally the satellite with the highest elevation is kept as reference for UPD processing. In addition, the UPD basis of ...'

46) l274/275

'It is believed that the mean of NL UPDs should... transformed. The mean of the UPD...'

47) l311

'.. NL ambiguity vector at the current epoch calculated from (18), which has been fixed at the last epoch and Qn is the corresponding cofactor matrix; ...'

48) l325

'When the number is low as five, the hold times .. are compared.'

49) l327,328

'.. , respectively, the ISRCB between between systems A and C are updated when ... and between systems A and B when ..'

50) l331

'.. fixed state is lost and the current output is just the float coordinate solution.'

51) l337/338

'.. with 40 reference stations in Asia, as shown in Figure 2, to generate UPDs. The UPDs...'

52) l341 'T'

In l341 you mention to conduct a kinemtic PPP-AR experiment. Thats not totally right. You estimate in each epoch coordinates of a static receiver- the receiver is not really moving as a regular rover receiver. Moreover you simulate obstructed environments by just using e.g. only the 3 satellites with highest elevation in the second half of your hourly experimental data. Make this quite clear to the readers.

53) l343

'.. receiver type is a Septentrio...'

54) l350/351 'T'

here you note that the 'ratio was experimentally set to 1.8'

what do you exactly mean here?

55) l357

'.. We checked the position errors ...'

56) l360 'T'

I wonder about the designation of the thresholds;

why do label the thresholds in the vertical with H and the ones in the horizontal with 'V' ?

did you mix up here the labels?

57) l367

'.. using the methood presented in chapters 2.2.2 and 2.2.3.'

58) Figures 3 and 5 'T'

Are the shown graphics of WL and NL hardware delays calculated uninterupted or with the mentioned restarts every 30min?

59) l403

'.. of WL, reaching up to 1.6 ...'

60) l410

'.. our DISBs show a slightly poorer stability. The reason might be ...'

61) l413

'.. in the PPP-AR model.'

62) l426-444

here you describe your 9 test schemes; would be much better to present the diffences of the test schemes in a nice and clear table instead of these text lines.

63) l458

'.. angles are larger than..'

64) l469

its just a hint , maybe the word 'restrict' should substituted here by 'limit'

buts thats up to you

65) l480

'.. Scheme 2 estimates two more inter-system...'

66) l503 'T'

change legend of Table 1 to

'Correct Fixing Rate Pcf according to the given error thresholds'

67) l505

change legend of Table 2 to

'Success Fixing Rate and Correct Fix Rate after classification according to the number of available satellites'

68) 'T'

you might add a graphic with the estimated ZTDs, are these reasonable or is it by definition better to use an apriori model

69) l508

'.. and V15 Pcf are above 90% when..'

70) l512,513

'.. angle of the QZSS satellite ... elevation is relatively low..'

71) l525

'.. DISB estimation, accounting for a change of the UPD reference , imposing..'

72) l528

'.. in a highly obstructed observation...'

73) 'T'

a final comment; your investiagtion and the paper would really gain if you can add a real example , taking observations from a multisystem rover which starts a kinematic trajectory in open area and then slowly moves towards shaded areas (urban canyon); or at least simulate the urban canyon by a

reasonable model of obstructing buildings; is your approach able to proces this mixed system dual frequency data? (moreover in close to real-time?)

best regards

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

1. I still don’t agree with the explanation of not using real kinematic data. (1) The reference coordinate can be easily derived from RTK or RTK/INS solutions. (2) Simulated high-shaded environment is not the real case. We can validate the algorithm with a simulated kinematic data, but real data is still necessary since we discussing about high-shaded environment.

2. If ISRCB and ZTD is still estimated, then the authors should give the weight of the constrain of ISRCB and ZTD from the last epoch. However, from what I read from the manuscript, I think the authors give a very tight constraint on ISRCB and ZTD, which I think it not suitable.

3. I don’t think it is a good idea to constraint the ambiguity in high-shaded environment, unless the authors explain their special techniques in dealing with such data. Yes, I agree that it is hard to get an ambiguity-fixed solution in such environment. However, the results presented are all based on open sky data, which is not convincing.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

The authors have revised the paper and it can be accepted.

Author Response

Thank you again for all your advices!

Reviewer 3 Report

Dear Authors

the revision of your manuscript has gained a lot compared to the previous version, both in technical, but most of all in terms of grammar. You have

introduced all my hints but also a huge number of further grammar immprovements. So, at least to my view, the paper is ready now. I have listed below a very short list of grammar hints which can be introduced in another 5 minutes. Line numbers below refer to the scheme of the revised version

1) l82

'.. using UDP production'; this term is not quite clear.

Maybe you want to mention that you 'generate' or 'introduce' UPD which are estimated beforehand or in parallel ?

2) Equations (1) 'T'

In my previous review I asked you about the wavelength of the IF combination. You replied with the related formulas. I am quite aware of the

formulas and wavelengths. I wanted just to point out that in principle

the wavelength of the IF combination is just a few mm. Of course if you introduce the WL ambiguity, the IF wavelength can be converted to a narrow-lane wavelength. So thats the background of my question, as it is not definitely explained in your manuscript

3) l303

 '.. at the current epoch. The constraint is...'

4) l337

'The user receiver was a Septentrio PolaRx5 connected to a TRM59800.00 antenna.'

5) l388

'.. our DISBs showed a slightly poor stability.'

best regards

Author Response

Please see the attachment.

Author Response File: Author Response.docx