Testing Scenario Identification for Automated Vehicles Based on Deep Unsupervised Learning

Round 1

Reviewer 1 Report

The article is well-written and provides a detailed description of the proposed methodology and experimental results. 

The following points need to be addressed 

In line 150, you mentioned that ξ is Euler's constant with a specific value. While the value you provided is an approximation of Euler's constant, it's worth noting that Euler's constant is an irrational number and cannot be represented exactly. If precision is crucial, you may want to mention that the value provided is an approximation. Provide the comment on this?

In Table 4, it would be beneficial to include a brief explanation of the evaluation metrics (SC, CH, DB) to help readers understand their significance in measuring clustering performance.

In Section 4.3, where the analysis of scene identification results is presented, consider briefly explaining the importance or relevance of the analyzed parameters and their impact on scenario classification. This can help readers understand the significance of the analysis.

The article requires minor English editing during proof-reading

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors have to incorporate the following suggestions for improving the readability of the paper. 

1. A flow chart representing the proposed methodology is to be included.

2. The samples of the driving data parameters for different scenarios have to be given in a Table.

3. What are the input and outputs considered for the 1D-RACE neural network?  And some samples of the training data have to be given.

4. What are the features considered for clustering using K-mean algorithm?,   and samples of the features have to be given.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The authors proposed an automatic driving test scenario identification method based on deep unsupervised learning by using a combination of multiple algorithms.

The paper presents all the steps the authors followed and should be easily reproduced by other researchers if needed.

I recommend the publication of this paper since it presents a specific method that should help in the near future the development of an automatic driving test scenario library which is of high interest these days.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report

1. What is the main question addressed by the research?

Naturalistic driving data (NDD) is valuable for testing autonomous driving systems under various driving conditions. Automatically identifying scenes from high-dimensional and unlabeled NDD remains a challenging task. In this paper, the authors present a novel approach for automatically identifying test scenarios for autonomous driving through deep unsupervised learning.

2. Do you consider the topic original or relevant in the field? Does it address a specific gap in the field?

Yes. The problem solved in this paper is relevant to this journal.

3. What does it add to the subject area compared with other published material?

The main contributions of this paper include:

(1) Utilizing the IF to achieve the segmentation of typical and extreme driving scenarios, resulting in separate datasets for extracting typical and extreme scenes.

(2) Designing a novel neural network, the 1D-RCAE, which can learn and extract features from data without the need for labels, in contrast to traditional machine learning methods.

(3) The residual learning mechanism is introduced to optimize the training process and enhances the feature extraction capability of the network.

(4) The application of IE optimizes the K-means algorithm, enhancing the accuracy and robustness of the clustering process.

4. What specific improvements should the authors consider regarding the methodology? What further controls should be considered?

- What is the motivation of using Isolation Forest (IF) algorithm to design the method?

- How to combine the authors’ work with these works on vehicle safety, such as u-safety urban safety analysis in a smart city, when urban safety index inference meets location-based data, vehicle safety improvement through deep learning and mobile sensing. More discussion should be added in the paper.

- More discussion on potential research directions and future works could be added in this paper.

5. Are the conclusions consistent with the evidence and arguments presented and do they address the main question posed?

Yes.

6. Are the references appropriate?

More technical papers about autonomous driving technologies could be investigated and analyzed. For example:

- A novel urban emergency path planning method based on vector grid map

- Real-time cache-aided route planning based on mobile edge computing

- A multi-objective hyper-heuristic algorithm based on adaptive epsilon-greedy selection

- An efficient learning-based approach to multi-objective route planning in a smart city

7. Please include any additional comments on the tables and figures.

- Tables and figures are easy to understand.

- A table of abbreviations could be added.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The authors addressed all the comments raised. 

While Proof read, Kindly check line number 79. The reference is missing in the PDF version.

Proof reading is required

Reviewer 2 Report

The revised version of the paper is acceptable for publication 

Reviewer 4 Report

All problems have been solved.