A Robust Adaptive Extended Kalman Filter Based on an Improved Measurement Noise Covariance Matrix for the Monitoring and Isolation of Abnormal Disturbances in GNSS/INS Vehicle Navigation
Round 1
Reviewer 1 Report
This paper illustrates a method that combines adaptive and Robust KF by weighting and adjusting the R matrix to achieve a more stable and accurate navigation system. The results seem promising, but the methodology needs more clarification and explanation. Most of the methodology sections merely present the equations without explaining why they are used.
- Abstract: The abstract contains a run-on sentence (“And constructing a robust… abnormal disturbance”.). Please consider rephrasing it for clarity and readability.
- Abstract: What does "POS" stand for in the abstract?
- Abstract: The meaning of "the heading angle anti-interference ability" is unclear from the abstract alone.
- Page 4, Line 139: What does "relative optimal results" mean in this context?
- Page 4: How did the author determine the weighting parameters of 0.85 and 0.15?
- Measurements: It is unclear what the actual measurements in this study are. Looking at Eq. (4), it seems to refer to NED position measurements only.
- Eq. (5): Is Eq. (5) based on empirical data? Is it overly conservative to have Q and PDOP as multiplying factors?
- Eq. (6): Is V_bar the innovation vector in Eq. (6)? Please clarify this point.
- Eq. (8): What does "L" represent in Eq. (8)? Additionally, what is the initial A_bar in Eq. (11)? Furthermore, how is Eq. (10) used?
- Table I and II: There is inconsistency between the two tables. Table I includes 6 methods, while Table II contains 7 methods. Please reconcile the difference.
see comments above
Author Response
Dear Reviewer,
We thank you every for your very insightful and constructive comments to improve our manuscript. According to your comments, we have carefully revised the manuscript. Please see the detailed responses to each comment in the attachment. The comments are cited in black. The response to each comment is set out in blue. The revised manuscript will be submitted at the stage of revision.
Zhihui Yin
Author Response File: 
Author Response.pdf
Reviewer 2 Report
1. Explain POS abbreviation – row (26)
2. Almost no references outside china. GNSS consist of 4 global positioning systems across whole continents on Earth and mostly they were not developed in China. Thus too much chineese references.
3. Where AKF,EKF, RKF are from ? are they Chinesee too ?
4. No explanation of IMU abbreviation
5. In (58) there is incorrect definition of Global Navigation Satellite System (GNSS). It should be Systems – no “ System”.
6. If we introducing definition GNSS for the first time (49), there is no need to repeat GNSS description in (58) because you already explained GNSS definition in (49)
7. In (82) “Hide et. al. [18] established a residual probability function is used to compute weighting factors” – this sentence has no sense and it is incorrect
8. In Figure (2) there is GNNS instead GNSS
9. In (134) there is incorrect definition of Global Navigation Satellite System (GNSS). It should be Systems – no “ System”.
10. What is “b” second power symbol in Q symbol in equation 5 (row 196) ?
11. What is “r” symbol in equation 5 (row 196) ?
12. Explain parameters in equation 213 – I don’t see a description of them
13. Describe which Novatel Model 2.4 Data Sets and Processing Strategies (250). Provide sequence of the configuration of all Novatel commands etc. log com2 bestpos ontime 1 0 hold, etc. sbascontrol enable of conducted experiments.
14. There is no information which global satellite systems were enabled in Novatel GNSS receivers and what antenna model was used.
15. Authors made statement that the results of proposed method, compared with the classical EKF, AKF, and RKF algorithm, are 72.43%, 2.54%, and 47.82%. These conclusion is incorrect because conducted experiments were badly planned – they were planned in one GNSS geometry state and data was recorded only during 26 minutes and 28 seconds at that geometry statement. Authors should divide their experiments to different time of different days. It is not enough geometry data to make conclusions because even after about 30 minutes geometry will not change too much. There is a risk that geometry can has huge impact to the proposed filtering method which includes PDOP geometrical as a parameter into proposed algorithm.
16. Moreover, there is no information in the paper is it better to use only GPS data or even GPS+GLONASS mixed data in comparison to whole GNSS data. Maybe proposed algorithm works better only for GPS+INS?
Comments for author File: Comments.pdf
Author Response
Dear Reviewer,
We thank you every for your very insightful and constructive comments to improve our manuscript. According to your comments, we have carefully revised the manuscript. Please see the detailed responses to each comment in the attachment. The comments are cited in black. The response to each comment is set out in blue. The revised manuscript will be submitted at the stage of revision.
Zhihui Yin
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
Thank you for your paper corrections.
Can you kindly add information about exact model name of Novatel GNSS receiver?
Author Response
Dear Reviewer,
We thank you every for your very insightful and constructive comments to improve our manuscript. According to your comments, we have carefully revised the manuscript. Please see the detailed responses to each comment in the attachment. The comments are cited in black. The response to each comment is set out in blue. The revised manuscript will be submitted at the stage of revision.
Zhihui Yin
Author Response File: Author Response.pdf
