Vehicle Trajectory Prediction Based on Adaptive Edge Generation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is an interesting paper that introduces a dynamic vehicle trajectory prediction model based on adaptive edge generation. Some aspects that must be improved are listed below.

1. The main contribution of the paper is the adaptive encoder generation. Figure 2. shows it, but, is too general, more details are required. A new figure should be added including more details about it. e.g. inputs, outputs, operations, ...

2. More detail about the trajectory prediction horizons. How many steps were predicted? This aspect is important because it is part of the manuscript contribution.

3. Some information about the dataset used for the experiments was not provided. What are the features or characteristics of this dataset, are there missing values ​​(percentage, treatment, ...), outliers, etc?

4. The equations (16-18) for ADE, FDE and MR should be described in detail. N, t, Ptx, Ttx, Pty, ... what are they? Also, RMSE or MAPE could be included.

5. The information provided about the baseline models is poor. A sub section for baseline models should be added where [38],[23],[39] can be described appropriately.

6. Just two decoders were compared. MLP and GRU. Why LSTM, BiLSTM, BiGRU... were discarded?

7. Also, a discussion section is required:

 - A comparison and discussion about the proposal and baseline models.

 - A comparison and discussion between the proposal and related works.

 - Some recommendations to improve the proposal and future work.

Author Response

Thank you for your letter and for the reviewer's comments concerning our manuscript entitled " Vehicle Trajectory Prediction Based on Adaptive Edge Generation" (Manuscript ID: electronics-3184400). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. We have studied comments carefully and have made correction which we hope meet with approval. 

Please see the respond in attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript describes a vehicle trajectory prediction using an adaptive edge detection framework. The following minor comments and questions should be addressed:

1. Page 8, line 318. Seems like p_0 should be the initial coordinate and p_1 should be the final coordinate?

2. Equation (14). Why did the authors choose an equal weight between the two terms? The authors should discuss the effects of different scaling of the two terms in the loss.

3. Section 4.2. The authors should explain how they got to this set of hyperparameters.

4. Section 4.3. The results in the ablation study seems to be coming from a single model training. It might be best to repeat the model training for a few times to show the mean and standard deviation of each metric.

5. Table 4. Memory usage is also an important measure, especially for mobile applications like on cars. This should also be reported.

 

Author Response

Thank you for your letter and for the reviewer's comments concerning our manuscript entitled " Vehicle Trajectory Prediction Based on Adaptive Edge Generation" (Manuscript ID: electronics-3184400). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. We have studied comments carefully and have made correction which we hope meet with approval. 

Please see the respond in attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

In this paper, the authors proposed a method to predict the trajectory of a vehicle based on an adaptive edge generator. In order to alleviate the high computational demands and complexity of a fully connected graph, we proposed a method to reduce the computational burden for prediction by implementing various connection methods according to various properties of dynamic and static nodes. The content of the paper is well structured, and the dataset and experimental method for performance evaluation seem to be appropriate.
- C1. However, since the history is not properly displayed in the actual trajectory visualization, it is difficult to determine whether the proposed method is accurate, so please add a line.
- The English is well written for easy reading, and only one period below needs to be corrected. recognition by computational systems.[17]

Author Response

Thank you for your letter and for the reviewer's comments concerning our manuscript entitled " Vehicle Trajectory Prediction Based on Adaptive Edge Generation" (Manuscript ID: electronics-3184400). Those comments are all valuable and very helpful for revising and improving our paper, as well as the important guiding significance to our research. We have studied comments carefully and have made correction which we hope meet with approval. 

Please see the respond in attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I agree with the revisions.