Schur Complement Optimized Iterative EKF for Visual–Inertial Odometry in Autonomous Vehicles

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

please reference properly a lot of inconsistencies. Lack of soundness in the approach and its validation. the background work need rewriting for better context

Comments on the Quality of English Language

could be improved which will improve the readability of the paper

Author Response

Please see the attached revised manuscript and point-by-point response to reviewers.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

As a reviewer of the manuscript, I have carefully examined the article and concluded that it does not currently meet the required standards for publication in this journal. However, if the authors address the comments and suggestions provided thoroughly and make the necessary revisions, reconsideration in the second round is possible.

While the manuscript cites VINS-Mono, ORB-SLAM3, and SchurVINS, there is no comparative evaluation against these baselines. Include quantitative comparisons against at least one optimization-based and one hybrid approach.

Add statistical significance testing for performance metrics.

The platform specs (i7-13700H, RTX3050) are given, but no runtime analysis is presented—recommendation to add Average runtime per frame, Memory usage, and CPU vs.GPU load breakdown.

Suggest the author for different experiments like Real-time benchmarking, Dynamic Scene Robustness, and Multi-View feature fusion.

Minor Comments:

• Clearly define all acronyms at first mention (IEKF, VIO, EKF, etc.).
• Reference [24] (SchurVINS) is very relevant and must be more explicitly contrasted with the proposed method.

Comments on the Quality of English Language

Nothing

Author Response

Please see the attached revised manuscript and point-by-point response to reviewers.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The structure of the paper is good, smoothly leading the reader to the final effect. Current topic filling a gap in the presented field. Visible contribution of the authors to the conducted research. This paper proposes a VIO method based on Schur complement and Iterated Extended Kalman Filtering (IEKF). The algorithm first enhances ORB (Oriented FAST and Rotated BRIEF) features using Multi Layer Perceptron (MLP) and Transformer architectures to improve feature robustness. It then integrates visual information and Inertial Measurement Unit (IMU) data through IEKF, constructing a complete residual model. Experimental results show that the algorithm achieves superior state estimation performance on several typical visual-inertial datasets, demonstrating excellent accuracy and robustness. Good work.

Below are a few minor comments:

• I suggest that the authors consider whether they are presenting an algorithm or a scheme in Figure 2. The algorithms contain yes, no conditions.
• First there should be a reference to the figure in the text, and then the figure – this applies to Fig. 3, Fig. 4, Fig. 9 and Fig. 10
• No references to formula numbers in the text.
• In my opinion, the conclusions should be expanded, which will enrich the work. The conclusions should include information about the results obtained, the significance of the work and emphasize the scientific contribution and differences in relation to others.
• The literature review was precisely performed, while the literature items should include doi

I believe that the changes introduced will enrich the article and make it more focused and transparent for potential readers.

 

Author Response

Please see the attached revised manuscript and point-by-point response to reviewers.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

Review of “Schur Complement Optimized Iterative EKF for Visual-Inertial 2Odometry in Autonomous Vehicles”

This is an interesting contribution on an important and topical application. The paper is generally clear and well written. There are however two important points concerning its estimation aspects that need to be improved before this contribution can be accepted for publication:

 

1. The authors make a big deal about using the “Schur complement” in eliminating parameters from their normal equations (6). However, standard practice in eliminating parameters from normal equations is to use reduced normal equation Cholesky-decomposition. This should be acknowledged by the authors.
2. The authors work with an iterated extended Kalman filter, but say nothing at all about its properties and conditions of convergence. This is of course not acceptable as such conditions are important to have and understand. The authors should therefore include such convergence information about the iterated algorithm, see e.g. [Teunissen (1991): On the local convergence of the iterated extended Kalman filter. Proc IAG 1991, 117-184].

Author Response

Please see the attached revised manuscript and point-by-point response to reviewers.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

better version

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the opportunity to review “Schur Complement Optimized Iterative EKF for Visual-Inertial Odometry in Autonomous Vehicles.” As a manuscript reviewer, I have carefully examined the article and concluded that it meets the required standards for publication in this journal, with only minor updates required.

Strengths

1. Innovative Integration: Combining deep learning (MLP + Transformer) with IEKF and Schur complement is a novel and powerful idea for improving both feature robustness and filter efficiency.
2. Strong Experimental Validation: Extensive tests across synthetic, public, and real-world datasets make the results credible and practical.
3. Clear Derivations: The mathematical models for IMU propagation, visual observation, and Schur complement are clearly described with supporting equations.
4. Real-Time Feasibility: The method shows competitive runtime performance, bridging the gap between high accuracy and computational efficiency.

Weaknesses & Areas for Improvement :The manuscript occasionally suffers from awkward phrasing and inconsistent grammar (e.g., "visual residual vision", "has been proposed which", etc.).Consider professional proofreading for language fluency and clarity.

Minor Comments

• Figure 2 (system diagram) would benefit from clearer labels and flow annotations.
• Figures 5–8 need better resolution or vector quality for publication.
• Consider integrating error bars in performance comparisons for statistical robustness.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

As the revision is clear, the paper can now be accepted for publication

Author Response

Please see the attachment.

Author Response File: Author Response.pdf