A Study of Human-like Lane-Changing Strategies Considering Driving Style Characteristics

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 Respect to the submitted manuscript: A Study of Human-Like Lane-Changing
Strategies Considering Driving Style Characteristics .
The manuscript has merit for publication but here are some major revisions
suggested as
1. Please provide validation with existing literature.
2. The geometrical diagram of the present problem has no reference, or the
authors have drawn it.
3. What is the novel contribution of the present work based on literature?
4. What is the solution methodology? Please include.
5. Include more physical details with practical applications in the results and
discussion section.
6. Add key findings in the abstract.
7. There should be a section on future directions.
8. Read the manuscript carefully to remove grammar and spelling mistakes.
9. Include a flowchart of solution methodology with practical applications of
present work.
10. Why authors have used a humanized lane-changing strategy for this work.
Include advantages over other strategies or reasons for choosing it.
In view of all the points given above, the paper needs to be modified and then only can be recommended for publication in the journal. 

Comments for author File: Comments.pdf

Comments on the Quality of English Language

 Read the manuscript carefully to remove grammar and spelling mistakes 

Author Response

Comments 1: [ Please provide validation with existing literature]

Response 1: [ Thank you for pointing this out. We agree with this comment. Therefore, we have fully incorporated recent literature on the integration of reinforcement learning and game theory into the introduction as suggested. Building upon this foundation, the core innovation of this study lies in introducing driving style characteristics into the game theory framework. This enables the construction of a human-like lane-changing decision strategy that directly addresses the critical challenge of human-like behavior in complex interactions for intelligent driving. Modifications can be found on page 2, lines 83-100.]

 

Comments 2: [ The geometrical diagram of the present problem has no reference, or the authors have drawn it.]

Response 2: [ Thank you for pointing this out. To clarify the origin of the geometric figures in this paper, we state that all illustrations—including geometric diagrams, data curve plots, and framework flowcharts—are original works created by the authors to complement the research content. None of these figures are sourced from existing literature or external resources. The purpose of these figures is to present the core models, scenario settings, algorithmic workflows, and simulation results in the clearest and most intuitive manner possible for the readers.]

 

Comments 3: [ What is the novel contribution of the present work based on literature?]

Response 3: [ Thank you for pointing this out. The novelty of this study is evident in two key aspects. Methodologically, while existing research has extensively applied game theory, it has rarely integrated it with quantifiable driving style characteristics. This work innovatively incorporates driving style clustering analysis within a game theory framework, constructing a differentiated payoff function that allows the decision model to capture a continuous spectrum of behaviors, from conservative to aggressive. In terms of application, we generated behaviorally interpretable and predictable human-like lane-changing trajectories to address specific challenges, such as “mechanical lane re-entry.” Simulations show that this strategy significantly enhances behavioral anthropomorphism and system versatility while ensuring safety, offering a novel solution to the issue of rigid behavior in autonomous driving.]

 

Comments 4: [ What is the solution methodology? Please include.]

Response 4: [ Thank you for pointing this out. The methodology of this study aims to achieve human-like lane changes through a coherent technical framework. First, driving styles are quantified and characterized using K-means clustering and traffic psychology theory. Next, these style characteristics are integrated into a game theory framework to simulate the lane-changing behaviors of different drivers. Building on this foundation, a dynamic reference line switching mechanism is introduced to ensure trajectory safety and smoothness. Finally, the effectiveness of the entire strategy is validated through integrated simulation. Modifications can be found on page 15, lines 398-401.]

 

Comments 5: [ Include more physical details with practical applications in the results and discussion section.]

Response 5: [ Thank you for pointing this out. We agree with this comment. Therefore, we have revised the conclusion section to include its potential for practical application: the proposed human-like lane-changing strategy has significant potential for practical application in personalized ADAS, high-fidelity simulation testing, and connected cooperative driving. It offers a key solution for seamlessly integrating autonomous driving into human transportation ecosystems.]

 

Comments 6: [ Add key findings in the abstract.]

Response 6: [ Thank you for pointing this out. We agree with this comment. Therefore, we have added the key conclusion at the end of the abstract: The human-like lane-changing strategy proposed in this paper dynamically balances safety, comfort, and efficiency while adapting to different driving styles. This approach provides a viable solution to the behavioral adaptation challenge in autonomous driving decision systems, laying a solid foundation for the development of next-generation personalized intelligent driving systems. Modifications can be found on page 1, lines 29-31.].

 

Comments 7: [ There should be a section on future directions.]

Response 7: [ Thank you for pointing this out. We agree with this comment. Therefore, we have added directions for future research in the conclusion section. First, the development of open online driving style recognition technology to enable strategies that dynamically adapt to real-time changes in the driver’s state. Second, large-scale robustness testing and the development of dynamically adaptive weighting strategies to enhance the adaptability and safety of the strategy in complex traffic flow environments. Modifications can be found on page 15, lines 402-410.]

 

Comments 8: [ Read the manuscript carefully to remove grammar and spelling mistakes.]

Response 8: [ Thank you for pointing this out. We agree with this comment. Therefore, we have thoroughly reviewed and proofread the entire document multiple times based on your feedback, making targeted revisions to correct any grammatical errors, spelling mistakes, and unclear expressions.]

 

Comments 9: [ Include a flowchart of solution methodology with practical applications of present work.]

Response 9: [ Thank you for pointing this out. We agree with this comment. Therefore, we have incorporated the feedback by adding a comprehensive flowchart of the human-like lane-changing strategy at the end of Section 3.4 to visually illustrate the decision-making process. The revised content can be found on page 10, lines 293–294. Additionally, we have expanded the conclusion section to emphasize the practical application potential of this strategy. We specifically highlight its value in scenarios such as personalized advanced driver assistance systems and autonomous driving simulation testing, thereby clarifying the implementation prospects of our research findings. The revised content can be found on page 15, lines 398–401.]

 

Comments 10: [ Why authors have used a humanized lane-changing strategy for this work. Include advantages over other strategies or reasons for choosing it.]

Response 10: [ Thank you for pointing this out. This study introduces an human-like lane-changing strategy that directly addresses the core challenge of autonomous driving in the real world: how to coexist safely and naturally with other road users, much like human drivers. While traditional strategies may meet safety standards, they often fall short in terms of behavioral interpretability and user experience. By simulating the dynamic trade-offs that human drivers make between safety, comfort, and efficiency, and by accounting for diverse driving styles, this strategy offers behaviorally explainable and predictably interactive lane-change decisions as its key advantage.]

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript addresses a relevant and timely topic concerning the modeling of “humanized” lane changes for intelligent driving systems. The authors propose a well-structured approach combining driving style analysis, decision modeling using game theory, and a dynamic reference line switching mechanism.
The work is methodologically sound, and the simulation results clearly demonstrate the superiority of the proposed method over the classical dynamic programming approach.
Overall, the article is of good quality and presents an interesting scientific contribution. However, a few minor adjustments are necessary to improve clarity, consistency, and overall presentation.
- The text would benefit from being proofread to correct a few minor grammatical errors and improve flow (especially in sections 3.3 and 4.2).
- Some sentences are long and could be reworded to make them easier to read.
- Check the legibility of the axes and legends in Figures 6 to 11.
-    The table captions should be more explicit (e.g., specify in Table 4 that these are weighting coefficients according to driving style).

- The literature review is comprehensive, but it would be useful to briefly mention recent work (2023–2025) based on reinforcement learning or hybrid AI–game theory approaches, in order to situate the contribution within the current scientific landscape.
- Certain variables appear without being redefined later. The notation between sections 3.3 and 3.4 should be standardized.
- The “Limitations” section (Section 5) could be slightly expanded to mention the challenges of implementation in real-world conditions (sensor uncertainty, vehicle-human interactions, mixed traffic, etc.).

 

Author Response

Comments 1: [The text would benefit from being proofread to correct a few minor grammatical errors and improve flow (especially in sections 3.3 and 4.2)]

Response 1: [ Thank you for pointing this out. We agree with this comment. Therefore, we have carefully proofread and linguistically refined the entire document in multiple rounds, as recommended, focusing on correcting grammatical errors and restructuring or optimizing sentences that were unclear or lacked logical coherence. This thorough process has significantly enhanced the document’s fluency and professionalism.]

 

Comments 2: [ Some sentences are long and could be reworded to make them easier to read.]

Response 2: [ Thank you for pointing this out. We agree with this comment. Therefore, we have thoroughly reviewed the entire text, with particular focus on analyzing and restructuring complex or excessively long sentences. By breaking down sentence structures, simplifying subordinate clauses, and optimizing logical connectors, we have made the text clearer and more direct, significantly improving its readability and fluency. These revisions have been applied throughout the revised draft.]

 

Comments 3: [ Check the legibility of the axes and legends in Figures 6 to 11.]

Response 3: [ Thank you for pointing this out. We agree with this comment. Therefore, we have fully optimized the axes and legends in Figures 6 through 11 based on your suggestions. By adjusting fonts, spacing, and contrast, we have significantly improved their clarity and readability. All modified images have been updated accordingly.]

 

Comments 4: [ The table captions should be more explicit (e.g., specify in Table 4 that these are weighting coefficients according to driving style).]

Response 4: [ Thank you for pointing this out. We agree with this comment. Therefore, we have optimized the titles of all tables in the text based on your feedback to make them more specific and clear. For example, the title of Table 5 has been updated to explicitly indicate that the weighting coefficients are calculated based on driving style. Please refer to lines 262–263 on page 8 for the revised content.]

 

Comments 5: [ The literature review is comprehensive, but it would be useful to briefly mention recent work (2023–2025) based on reinforcement learning or hybrid AI–game theory approaches, in order to situate the contribution within the current scientific landscape.]

Response 5: [ Thank you for pointing this out. We agree with this comment. Therefore, we have incorporated your suggestions by adding relevant recent studies to the literature review. The new content highlights research that integrates cooperative games with deep learning to optimize multi-agent decision-making efficiency, as well as studies proposing frameworks that combine game theory and reinforcement learning to enhance the safety of autonomous driving decisions. These advancements provide a cutting-edge scientific context for our exploration of incorporating driving style characteristics within a game-theoretic framework to achieve human-like decision-making. The revised content can be found on page 2, lines 85–91.]

 

Comments 6: [ Certain variables appear without being redefined later. The notation between sections 3.3 and 3.4 should be standardized.]

Response 6: [ Thank you for pointing this out. We agree with this comment. Therefore, we have thoroughly reviewed and standardized all symbols throughout the manuscript in accordance with the reviewers' comments. Special attention was given to ensuring consistency between Section 3.3 and Section 3.4, and reference notes have been added for the first definition of variables that reappear in subsequent sections. All symbols have been formally defined.]

 

Comments 7: [ The “Limitations” section (Section 5) could be slightly expanded to mention the challenges of implementation in real-world conditions (sensor uncertainty, vehicle-human interactions, mixed traffic, etc.).]

Response 7: [ Thank you for pointing this out. We agree with this comment. Therefore, we have substantially expanded the conclusion section in Section 5 of the paper, explicitly acknowledging the study's limitations in two key areas: dynamic online identification and adaptive adjustment of driver styles, as well as systematic parameter sensitivity testing and confidence interval analysis. We have also clearly outlined future research directions and proposed methods for addressing these issues. The revised content can be found on page 15, lines 402–410.]

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The article addresses a current and relevant topic in the field of autonomous vehicles. Overall, the work is well structured and consistent with the literature in the field, but it has some weaknesses. In particular:

• The use of K-means for clustering driver styles is described superficially, with a lack of validity metrics (e.g., silhouette score, Davies–Bouldin index) and justifications for the choice of the number of clusters (k=3).
• The clustering parameters (speed, acceleration, headway) are standard, but no normalization or outlier handling techniques are specified.
• The formalization of the game is only qualitative. No numerical payoffs or equilibrium analysis (Nash, Stackelberg, or cooperative) are provided.
• The weighting of the objectives is assumed ad hoc (0.2/0.6/0.2, etc.) without any quantitative basis.
• The simulations do not include confidence intervals, sensitivity tests, or analyses of the robustness of the results.
• The comparison with the DP method is interesting but limited to a few indicators. There is no validation in more realistic environments or with mixed traffic.
• Regarding the results, the differences in performance (e.g., +7.06% average speed, +75% smoothness) are significant, but statistical significance is not stated.
• The interpretation of the results appears to be more descriptive than analytical. The possible influence of input parameters (initial speed, safety distance, traffic density) is not discussed.
• Tables 1-4 are clear but should include units of measurement and sources
• The bibliography is solid but too focused on Chinese works. There is no reference to recent Western studies.
• The authors acknowledge some limitations, e.g., the incomplete representation of individual variability and the simplification of traffic, but without proposing operational strategies to overcome them.
• A bibliography of 20 articles seems a bit sparse to me.

Author Response

Comments 1: [ The use of K-means for clustering driver styles is described superficially, with a lack of validity metrics (e.g., silhouette score, Davies–Bouldin index) and justifications for the choice of the number of clusters (k=3).]

Response 1: [ Thank you for pointing this out. We agree with this comment. Therefore, we have supplemented the rationale for selecting (k=3) in accordance with the reviewer’s comments. The elbow rule shows that (k=3) corresponds to a significant inflection point in the distortion curve, while contour coefficient analysis demonstrates optimal clustering quality, with the best cohesion and separation at (k=3). These findings collectively validate the choice of this number of clusters. The revised content can be found on page 4, lines 147–153.]

 

Comments 2: [ The clustering parameters (speed, acceleration, headway) are standard, but no normalization or outlier handling techniques are specified.]

Response 2: [ Thank you for pointing this out. We have incorporated the reviewers' suggestions by adding descriptions of the data preprocessing steps in the relevant sections. Outliers were filtered using the Isolation Forest algorithm, resulting in the removal of 545,047 records. Subsequently, the velocity, acceleration, and vehicle distance features used in the clustering analysis were Z-score normalized to ensure the quality and reliability of the clustering results. The modifications are detailed on page 4, lines 138–146.]

 

Comments 3: [ The formalization of the game is only qualitative. No numerical payoffs or equilibrium analysis (Nash, Stackelberg, or cooperative) are provided.]

Response 3: [ Thank you for pointing this out. We fully agree that rigorous mathematical formalization and equilibrium analysis of game models are crucial for advancing theoretical research. As a foundational exploration and proof-of-concept integrating driving styles with game theory, the core contribution of this study lies in demonstrating the effectiveness and potential of this approach in enhancing human-like decision-making. While the precise definition of benefit functions and the analysis of Nash equilibrium solutions are indeed important, they fall beyond the scope of this paper, which focuses on constructing and validating the core framework. We highly value your suggestions and consider them a key direction for future research. We plan to conduct rigorous mathematical modeling and equilibrium analysis upon this solid foundation. Please refer to lines 402–410 on page 15 for the revised content.]

 

Comments 4: [ The weighting of the objectives is assumed ad hoc (0.2/0.6/0.2, etc.) without any quantitative basis.]

Response 4: [ Thank you for pointing this out. We agree with this comment. Therefore, we have made key modifications to the weight determination method in response to your comments. A data-driven approach is now employed, where weights are calculated through a standardization and normalization process based on style features derived from cluster analysis. This approach provides a robust quantitative foundation for the previously defined weight values. The modifications are detailed on page 8, lines 241–263.]

 

Comments 5: [ The simulations do not include confidence intervals, sensitivity tests, or analyses of the robustness of the results.]

Response 5: [ Thank you for pointing this out. This study serves as a proof-of-concept for the proposed integrated framework, with its core contribution lying in demonstrating its fundamental effectiveness. For the robustness analysis of the results, we have included simulation validation in Section 4.3. However, more in-depth robustness analysis, such as large-scale sensitivity testing, is beyond the scope of the current work. We highly value your suggestions and have incorporated them into our future research plans. A systematic robustness assessment will be a top priority in the next phase of our work. The modifications are detailed on page 15, lines 402–410.]

 

Comments 6: [ The comparison with the DP method is interesting but limited to a few indicators. There is no validation in more realistic environments or with mixed traffic.]

Response 6: [ Thank you for pointing this out. We fully acknowledge the importance of validating algorithms in mixed traffic flows. As the current phase of this research focuses on constructing the core algorithmic framework and demonstrating its principles, large-scale mixed traffic simulation is beyond the scope of the present work. To address your concerns, we have added a new subsection to the paper’s conclusion. This section offers a theoretical exploration of the method’s potential and inherent advantages in complex traffic environments, providing a foundation for subsequent experimental validation. Please refer to lines 402–410 on page 15 for the revised content.]

 

Comments 7: [ Regarding the results, the differences in performance (e.g., +7.06% average speed, +75% smoothness) are significant, but statistical significance is not stated.]

Response 7: [ Thank you for pointing this out. We agree with this comment. Therefore, we have incorporated your suggestions by adding Section 4.3 on simulation result validation and including statistical tests in the results analysis. An independent samples t-test of the simulation results confirms that the reported performance improvements are statistically significant, thereby supporting the reliability of our conclusions. The revised content can be found on page 15, lines 384–392.]

 

Comments 8: [ The interpretation of the results appears to be more descriptive than analytical. The possible influence of input parameters (initial speed, safety distance, traffic density) is not discussed]

Response 8: [ Thank you for pointing this out. We agree with this comment. Therefore, we have enhanced the results analysis section by providing a stronger mechanistic interpretation of the simulation outcomes and reducing purely descriptive accounts of phenomena. Regarding the impact of input parameters, we agree that systematic sensitivity analysis is crucial. Given that this study focuses on constructing and validating the core framework, such an analysis has been identified as a key priority for future work. We have explicitly acknowledged this as a primary limitation in the conclusions section and placed the development of a parameter sensitivity analysis framework at the forefront of our future research agenda.]

 

Comments 9: [ Tables 1-4 are clear but should include units of measurement and sources.]

Response 9: [ Thank you for pointing this out. We agree with this comment. Therefore, we have systematically reviewed and updated all tables in the document. Measurement units have been explicitly specified for all numerical data. For example, velocity, acceleration, and time-distance are now labeled as (m/s), (m/s²), and (s), respectively.]

 

Comments 10: [ The bibliography is solid but too focused on Chinese works. There is no reference to recent Western studies.]

Response 10: [ Thank you for pointing this out. We agree with this comment. Therefore, we have incorporated recent Western research findings into the references. This revision enhances the structure of the literature review and ensures the comprehensiveness of the research background. The modified content can be found on pages 16-17, lines 436-476.]

 

Comments 11: [ The authors acknowledge some limitations, e.g., the incomplete representation of individual variability and the simplification of traffic, but without proposing operational strategies to overcome them.]

Response 11: [ Thank you for pointing this out. We agree with this comment. Therefore, we have outlined our future operational strategy in the conclusion section. First, we plan to address the issue of insufficient representation of individual differences by developing online driving style recognition technology. Second, we aim to enhance performance in complex traffic environments through large-scale robustness testing and the development of adaptive weighting strategies. Please refer to lines 402–410 on page 15 for the revised content.]

 

Comments 12: [ A bibliography of 20 articles seems a bit sparse to me.]

Response 12: [ Thank you for pointing this out. We agree with this comment. Therefore, we have expanded the references to include recent Western research findings.]

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Accept in present form

Comments on the Quality of English Language

 Read the manuscript carefully to remove grammar and spelling mistakes 

Author Response

Comments 1: [ Accept in present form.]

Response 1: [ We thank the reviewer for their positive and encouraging feedback on our manuscript. We are very pleased to hear that the work is considered acceptable for publication in its present form.]

 

Comments 2: [ Read the manuscript carefully to remove grammar and spelling mistakes.]

Response 2: [ Thank you for pointing this out. We agree with this comment. Therefore, we have thoroughly proofread the entire manuscript to correct any grammatical, spelling, and typographical errors. This meticulous check has significantly improved the clarity and polish of the manuscript.]

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have adequately satisfied my reviews and the paper is now publishable

Author Response

Comments 1: [ The authors have adequately satisfied my reviews and the paper is now publishable.]

Response 1: [ We thank the reviewer for their positive and encouraging feedback on our manuscript. We are pleased that the manuscript is now deemed publishable.]

Author Response File: Author Response.docx