Comments to author (Associate Editor)
=====================================

Please follow the reviewers' suggestions in correcting and
improving your contribution.


Reviewer 6 of ITSC 2023 submission 773

Comments to the author
======================

This paper study an important problem in ITS, and the
approach is valid. The results are convincing.

Reviewer 7 of ITSC 2023 submission 773

Comments to the author
======================

Accurate prediction of vehicle trajectories is vital for
advanced driver assistance systems and autonomous vehicles.
This paper propose an approach for interaction-aware
personalized vehicle trajectory prediction that
incorporates temporal graph neural networks., in which it
utilizes Graph Convolution Networks (GCN) and Long
Short-Term Memory (LSTM) to model the spatio-temporal
interactions between target vehicles and their surrounding
traffic. Human-in-the-loop simulation is employed to
collect personalized naturalistic driving trajectories and
corresponding surrounding vehicle trajectories. I think the
topic is interesting.

(1) The comparisons of trajectory prediction error (RMSE)
between the proposed methods and other vehicle trajectories
models, such as physics-based motion models and
maneuver-based motion models is concluded in TABLE I.
However, I think the computation time is also a key
parameter to be considered. It may be cost more times
compared with other vehicle trajectories models, as
calculating the relationship between multiple cars requires
a considerable amount of computation. This paper should
show more computation time cusumpation result to show the
innovation clearly.




Reviewer 11 of ITSC 2023 submission 773

Comments to the author
======================

This paper is well-presented and technically sound.
However, it suffers from the following issues:
1, The study fails to introduce the impact of learning
personalized driving trajectory, which lays the foundation
for this study. It would be helpful for readers to
understand the impactfulness of the study

2, The objective of the study is a bit ill-formed, if not
misleading. My first question is how to measure the
characteristics of a driving trajectory. Without
establishing a comprehensive set of metrics that evaluate
the quality of personalization, it is hard to understand
the quality of the learned strategy. The only
personalization metric used in the paper is the MSE of
driving trajectories, which is insufficient to evaluate the
characteristics learned by the model in my opinion. In
addition, a driver possesses different driving styles,
which might be taken into consideration if the objective is
to learn a personalized trajectory.

3, The benchmark should be more inclusive. It will be
interesting to see how the proposed framework perform
against IRL.