Next Article in Journal
Enhancing Reinforced Concrete Beams: Investigating Steel Dust as a Cement Substitute
Next Article in Special Issue
Vehicle Driving Safety of Underground Interchanges Using a Driving Simulator and Data Mining Analysis
Previous Article in Journal
Practical Aspects of Correlation Analysis of Compressive Strength from Destructive and Non-Destructive Methods in Different Directions
 
 
Article
Peer-Review Record

Comparative Analysis of Machine-Learning Models for Recognizing Lane-Change Intention Using Vehicle Trajectory Data

Infrastructures 2023, 8(11), 156; https://doi.org/10.3390/infrastructures8110156
by Renteng Yuan 1, Shengxuan Ding 2 and Chenzhu Wang 2,*
Reviewer 2: Anonymous
Reviewer 3: Anonymous
Infrastructures 2023, 8(11), 156; https://doi.org/10.3390/infrastructures8110156
Submission received: 2 September 2023 / Revised: 11 October 2023 / Accepted: 14 October 2023 / Published: 25 October 2023
(This article belongs to the Special Issue Recent Progress in Transportation Infrastructures)

Round 1

Reviewer 1 Report

The manuscript describes research conducted to recognize lane change maneuvers on multilane highways from drone images using machine learning techniques.

Beyond the evident layout problems when equations are inserted, the authors should have included line numbering in the document. Without it, it is very difficult for reviewers and authors carry out an adequate review process.

There are some aspects that must be improved before the manuscript is adequate for publishing, if applicable.

The abstract does not convey adequately what the study consists of. It should rather follow a typical objective – materials and methods – results – conclusions structure.

Section 2.3. The description of the variables used is vague. A detailed description should be included. This aspect is paramount.

The characteristics of the dataset used should be described, in addition to adding the appropriate references. For example, the relative positions of the vehicles may vary during the frame lapse analyzed. In which moment are the positions designated?

Does image recognition identify the vehicle's direction indicators?

Do the variables include the transverse position of the vehicle in the lane?

Section 3 includes a lot of theory but how the method was implemented is missing.

The metrics are apparently good enough. Have cases in which more than one of the vehicles involved change lane analyzed? Have data been removed to eliminate this difficulty?

Neither throughout the article nor in the conclusions themselves is it clear what is contribution of the authors and what is an analysis of previous developments. This applied to the three methods analyzed: SVM, EM and LSTM.

The wording of the article is of a low standard, which makes it difficult for readers to understand, the methodology and the analysis of results. Sections and paragraphs are not well connected, somewhat repetitive with redundant content. This issue needs not only careful review by authors but also English proofread performed by native speakers.

Author Response

Specific response is in the attached word.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper focuses on Lane changing processes and conducts a comparative analysis of three widely recognized machine learning techniques, specifically Support Vector Machines (SVM), Ensemble Methods, and Long Short-Term Memory (LSTM) networks. Overall, the manuscript is well organized with novel research topic, clear logic, and solid experiments. However, several problems remain to be clarified in detail:

1. The contributions in this paper are not clear. Please clarify the difference between this research and previous research.

2. In Section “Introduction”, the authors should add a chapter structure of this paper.

3. In 2.2 Indicator Calculation”, some key indicators should be given.

4. In 2.2 Indicator Calculation”, the principle of the nonlinear low-pass filter should be introduced.

5. In “2.3 Input indicator”, it is necessary to add additional tables to list the input indicators.

 

 

Please proofread the grammatical and typing mistakes of all paragraphs thoroughly in the manuscript.

Author Response

Specific response is in the attached word.

Author Response File: Author Response.docx

Reviewer 3 Report

This paper shows the comparison between four methods which are modelling lane change behaviour.

This comparison focuses on accuracy, training time, as well as the so-called confusion matrix.

The results show quite some differences between the four methods. The method that shows the best results, will be used in further research activities.

The paper gives a good description of the way these methods have been compared. The conclusion of the paper is very clear.

Author Response

Specific response is in the attached word.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The manuscript has been improved in the latest version. However, I consider that the responses to comments #4, #5 and #6 must be reflected in the manuscript, explaining the limitations of the study in the text explicitly.

Author Response

The response to the reviewer has been attached in the word.

Author Response File: Author Response.docx

Back to TopTop