Review Reports

Round 1

Reviewer 1 Report (New Reviewer)

Comments and Suggestions for Authors

The research topic holds significant practical relevance, featuring integrated methodological innovation and a relatively systematic experimental design. However, the manuscript requires improvements in demonstrating innovation, detailing methodologies, and analyzing and presenting results. Therefore, I recommend major revisions.
1. The manuscript highlights the innovation of integrating Delphi-AHP into the A* algorithm in the introduction, but lacks a thorough comparison with the latest path planning algorithms from 2023-2025. We recommend adding quantitative comparisons with recent literature in the discussion section.
2. Supplement the manuscript with detailed descriptions of the Delphi technique and AHP to enhance the reproducibility and transparency of the methodology.
3. Section 3.4 reports performance improvements but omits statistical significance tests to validate reliability. Discuss the relationship between practical and statistical significance.
4. Some figure annotations are unclear, and units are inconsistent. Optimize chart resolution, add legend descriptions, and ensure unit consistency.
5. The manuscript contains minor language issues, including spelling errors, grammatical problems, and inconsistent terminology. A comprehensive language review is recommended, along with standardization of key terms.

Author Response

Reviewer 1

1. The manuscript highlights the innovation of integrating Delphi-AHP into the $A^$ algorithm in the introduction, but lacks a thorough comparison with the latest path planning algorithms from 2023-2025. We recommend adding quantitative comparisons with recent literature in the discussion section.*

Author’s Response:

"Thank you for this insightful suggestion. We agree that benchmarking against contemporary literature is essential for validating the algorithm's innovation. We have significantly expanded the Discussion section (Section 4.6) to include a quantitative and qualitative comparison with recent studies from 2023-2025. Specifically, we have compared our 3.8% travel time reduction and 8.6% safety improvement against the findings of Zhao et al. (2024) regarding emergency rescue path planning and Huang and Wang (2025) concerning improved $A^*$ variants. This analysis highlights that while recent models focus on general obstacle avoidance or off-road navigation, our Delphi-AHP-integrated framework offers a specialized and superior solution for managing high-hazard chemical dispersion in urban expressway logistics."

2.Supplement the manuscript with detailed descriptions of the Delphi technique and AHP to enhance the reproducibility and transparency of the methodology.

Author’s Response:

"We have significantly enhanced the description of our methodology in Section 2.2 to provide full transparency regarding the expert selection process. We have now clarified that the panel of 17 experts was chosen based on the Net Change principle. Our analysis identified $N=17$ as the first group size where the Net Change reached its minimum value, representing the threshold of maximum consensus stability. This ensures that the expert elicitation is robust and reproducible. We have also added citations (Savkovic et al., 2022; Wang et al., 2022) to support this specific sample size determination as the point of optimal group stability."

3.Section 3.4 reports performance improvements but omits statistical significance tests to validate reliability. Discuss the relationship between practical and statistical significance.

Author’s Response:

"We sincerely appreciate the reviewer's suggestion to strengthen the reliability of our findings. We have addressed this comment by revising Sections 3.4 and 4.5 as follows:

1. Statistical Validation: To confirm the reliability of the results, we conducted 30 simulation trials for each algorithm. A paired t-test was performed, and the results have been added to Section 3.4, confirming that the improvements (3.8% in travel time and 8.6% in safety scores) are statistically significant with $p < 0.05$. 2. Practical Significance: We have expanded the discussion in Section 4.5 to bridge the gap between statistical results and real-world application. We clarified that while a 3.8% time reduction is statistically significant, its practical significance is vital for the 'Golden Hour' in HAZMAT emergency response, where even minor time savings can prevent catastrophic secondary incidents.

We believe these additions significantly enhance the scientific rigor and transparency of the study."

4. Some figure annotations are unclear, and units are inconsistent. Optimize chart resolution, add legend descriptions, and ensure unit consistency.

Author’s Response:

We sincerely thank the reviewer for this constructive feedback. We have meticulously updated all figures to improve clarity, precision, and visual quality. The following modifications have been implemented:

• Comprehensive Revision of Figure 1: We have fully revised the research conceptual framework in Figure 1. and improving the layout to better illustrate the multi-stage integration of the Delphi-AHP and A* algorithm.
• Clarification of Annotations in Figure 2: Alphanumeric codes (e.g., A1, F1) have been replaced with descriptive, abbreviated labels (e.g., Traffic Flow, Chem. Conc.) to ensure the figure is self-explanatory without constant reference to the tables .
• Standardization of Units: We have ensured unit consistency across all sections. The distance is now consistently denoted as "km" and time as "min" or "mins" in Table 5, Figure 5, and the related discussion.
• Enhanced Resolution & Legends: All figures have been re-exported at 300 DPI for optimal resolution. Furthermore, explicit legend descriptions were added to Figure 3 and Figure 4 to clearly distinguish between the "Proposed A* Algorithm" (green) and "Dijkstra’s Algorithm" (red) trajectories.

5. The manuscript contains minor language issues, including spelling errors, grammatical problems, and inconsistent terminology. A comprehensive language review is recommended, along with standardization of key terms.

Author’s Response:

The authors sincerely apologize for the linguistic oversights in the original submission. We have conducted a comprehensive language review and professional editing of the entire manuscript. The following improvements have been made:

1. Terminology Standardization: Key terms have been standardized throughout the text, tables, and figures. For instance, "Safety Score" is now used consistently instead of "Safe Score". Similarly, units of measurement have been unified (e.g., using "km" for distance and "min" for time).
2. Spelling and Typographical Corrections: All spelling errors, particularly within the figures and diagrams, have been corrected. Figure 1 and Figure 2 have been completely updated to fix errors.
3. Grammatical Improvements: The manuscript underwent a thorough grammatical check to improve sentence structure and clarity, ensuring a more formal academic tone suitable for Applied Sciences.
4. Consistency in References: We have ensured that the naming of algorithms (e.g., A* algorithm) and factors are consistent between the methodology and results sections.

Author Response File: Author Response.pdf

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

General comment: This article presents a methodological proposal for route optimization in complex hazardous materials emergency response (HAZMAT) scenarios, integrating the A* algorithm with an expert-weighted cost function through a combination of Delphi and AHP techniques. The proposal is applied to an urban network of expressways and compared with traditional algorithms such as Dijkstra and Ant Colony Optimization (ACO). Overall, the study addresses a relevant topic, especially for applications in emergency management, critical logistics, and urban planning, areas where the appropriate choice of routes can directly impact public safety and operational efficiency.

 

1) E-mail: Correct the authors' email addresses.

 

2) Title: The title doesn't make the article's purpose clear. Suggestion: "Multi-Criteria Route Planning for HAZMAT Emergency Response Using a Delphi-AHP Weighted A Algorithm". In this case, the methodological idea becomes explicit.

 

3) Figures: The quality of the figures needs improvement.

 

4) Tables: The tables are not formatted according to the MDPI template and are not standardized within the document, appearing in different ways.

 

5) Abstract:

i) The abstract is well written.

ii) The improvements presented (3.8% and 8.6%) seem relatively small. iii) The "safety scores" metric is not explained.

 

6) Keywords:

i) "Key Route Planning" seems to be incorrect.

ii) "Dijkstra's" was one of the methodologies compared in this study. Wouldn't it be an important keyword?

 

7) Equations: A Few equations are cited throughout the methodology. The main methods were not presented. Review this issue and present all the algorithms used for comparison in this study.

 

8) Introduction:

The introduction is too short. It needs to be extensive and improved.

i) There is no contextualization.

ii) There is no literature review consistent with the study.

iii) The literature gap is not presented.

iv) The objectives are not stated.

v) It is necessary to add the research questions.

vi) The introduction should conclude by presenting the following sections.

 

9) Materials and Methods

The methodology needs to be rewritten.

i) The main methods are not presented.

ii) Methods are cited without indicating the references.

iii) It is not clear throughout the text in what context the AHP method is used, considering that the authors cite the Delphi method for calculating the weights.

iv) The table with Saaty's fundamental scale is cited without referencing the author and without making its purpose clear.

v) The term "factor" only appears in the methodology, without making its purpose clear in the context of the study.

vi) Why jump from "A1" to "A3" and from "C1" to "C3"?

vii) With 17 experts, how was the extreme divergence of opinions between the Delphi rounds handled before reaching a consensus? Were the similarity coefficients of the responses considered? Several exist in the literature and should be considered.

viii) Table 3 shows that "D2 (ERT Location)" was given a weight of 0.00. If the weight is zero, was the factor effectively discarded? Doesn't this contradict the importance of proximity in emergencies?

ix) The calculation of h(n) uses the Manhattan Distance. In expressway networks, where movement is restricted to specific arcs, wouldn't the Euclidean Distance or a heuristic based on the network's topography be more accurate?

x) For the ACO, how was the pheromone evaporation rate and the number of artificial ants defined? These hyperparameters directly influence convergence to the optimal route.

 

10) Results:

The results are presented briefly.

i) Lack of statistical tests to verify the results.

ii) The results indicate small improvements, but there is no proof that these are significant.

iii) The ALOHA software simulated dispersion with a 1 m/s wind (circular pattern). How would the algorithm behave in strong, directional winds (elongated plumes)?

iv) A* explored 150 knots compared to Dijkstra's 220. Since Dijkstra is a special case of A*, with h(n) = 0, this efficiency was expected; however, the inclusion of safety costs S(n) generally increases the search space. Why was it so much faster here?

v) How many scenarios were simulated?

vi) What is the size of the road network?

vii) What is the execution time?

viii) Add the following items:

- Statistical tests.

- Confidence intervals.

- Robustness analysis.

 

11) Discussion and conclusions:

i) There are extra paragraphs, and "Conclusions" is not bold.

ii) Comparison with limited literature.

iii) Does not discuss model limitations.

iv) Repeats the results and does not clearly present the scientific contribution.

Author Response

General Response: We would like to thank the reviewer for the positive assessment and the constructive comments provided. We have carefully revised the manuscript to enhance its clarity and methodological explicitness.

1: Correct the authors' email addresses.

Response: We have thoroughly checked and corrected the email addresses for all authors in the revised manuscript . We also verified the contact information in the correspondence section to ensure accuracy.

2: Title: The title doesn't make the article's purpose clear. Suggestion: "Multi-Criteria Route Planning for HAZMAT Emergency Response Using a Delphi-AHP Weighted A Algorithm". In this case, the methodological idea becomes explicit.

Response: We agree with the reviewer’s suggestion to make the methodology more explicit in the title. We have revised the title to "Multi-Criteria Route Planning for HAZMAT Emergency Response Using a Delphi-AHP Weighted A Algorithm: A Case Study in Expressway Networks." We decided to retain the phrase "A Case Study Urban Expressway Networks" to emphasize that the proposed framework was validated using a real-world, high-stakes scenario within a complex urban infrastructure. This addition highlights the practical applicability of the algorithm and its ability to handle dynamic urban constraints, such as traffic fluidity and chemical dispersion, which are central to the study’s contribution 

3: Figures: The quality of the figures needs improvement.

Response: > We appreciate the reviewer's comment regarding the visual quality of our work. We have thoroughly revamped all figures in the revised manuscript to meet high-standard academic publishing requirements. Specific improvements include: 

1. Higher Resolution: All figures (Figures 1–5) have been re-exported at a minimum resolution of 600 DPI in TIFF/PDF format to ensure maximum clarity. 
2. Visual Consistency: We have standardized the fonts (Arial), line weights, and color schemes across all diagrams and charts to provide a more cohesive visual presentation. 
3. Enhanced Readability: In Figure 2, the labels have been enlarged for better legibility. In Figures 3 and 4, the map overlays and ALOHA dispersion plumes have been rendered with higher contrast to clearly distinguish between the different routing trajectories. 
4. Simplified Framework: Figure 1 has been redesigned to more effectively communicate the research methodology at a glance.

4: Tables: The tables are not formatted according to the MDPI template and are not standardized within the document, appearing in different ways.

Response: > We appreciate the reviewer’s feedback regarding the formatting of our tables. We have performed a comprehensive overhaul of all tables to strictly adhere to the MDPI editorial guidelines. The following changes were implemented:

1. Standardized Formatting: All vertical lines have been removed, and horizontal borders have been applied according to the MDPI template style. 2. Consistent Numbering: Table numbering has been corrected and re-sequenced 3. Unified Layout: Alignment has been standardized, with text-heavy columns left-aligned and numerical data center-aligned for better readability. 4. Standardized Units: Units such as "min" and "km" are now used consistently across all table headers and data cells.
2. Font and Spacing: We have ensured that the font type, size, and caption positioning are uniform throughout the entire manuscript.

5: Abstract: i) The abstract is well written. ii) The improvements presented (3.8% and 8.6%) seem relatively small. iii) The "safety scores" metric is not explained.

Response: > We appreciate the reviewer's positive comment on the writing quality and the insightful observations regarding the results.

* Regarding the significance of 3.8% and 8.6% improvements: While these percentages may appear small in a general context, we have clarified in the revised manuscript that in HAZMAT emergency response, these gains are operationally significant. A 3.8% reduction in travel time (approximately 30 seconds in our scenario) is vital for adhering to the "Golden Hour" principle. Furthermore, the 8.6% safety score enhancement represents a strategic shift in routing that moves the rescue unit out of high-hazard chemical plumes, which can be the difference between a successful mission and secondary disaster exposure . * Regarding the "Safety Scores" metric: We have added a brief definition in the abstract to clarify that the Safety Score is an aggregate metric derived from 12 expert-weighted parameters, including population density and chemical concentration levels, quantified through the Delphi-AHP process.

6: Keywords: i) "Key Route Planning" seems to be incorrect. ii) "Dijkstra's" was one of the methodologies compared in this study. Wouldn't it be an important keyword?

Response: > We agree with the reviewer’s observations and have updated the keywords to improve the manuscript’s indexing and clarity. 1. The term "key Route Planning" was a typographical error and has been corrected to "Multi-criteria Route Planning" to better align with the revised title. 2. As suggested, we have added "Dijkstra’s Algorithm" as a keyword, as it serves as a fundamental benchmark in our comparative analysis. 3. Additionally, we have included "HAZMAT Emergency Response" to more accurately reflect the specific application domain of this research.

7: Equations: A Few equations are cited throughout the methodology. The main methods were not presented. Review this issue and present all the algorithms used for comparison in this study.

Author’s Response: > We agree that providing a formal mathematical representation of the algorithms enhances the methodological clarity of the study. We have revised the "Materials and Methods" section to include the fundamental equations for all algorithms used in our comparative analysis: 1. Dijkstra’s Algorithm: We have added the relaxation formula (Equation X) to illustrate how the algorithm iteratively identifies the shortest path based on edge weights . 2. Ant Colony Optimization (ACO): We have introduced the transition probability equation (Equation Y), showing how pheromone intensity (τ) and heuristic visibility (η) influence route selection. We have also explicitly linked our modified cost function (Equation 3) to the heuristic visibility parameter (ηij)to show the integration of Delphi-AHP weights into the ACO framework . 3.A*Algorithm: The existing cost function equations (Equations 1-3) have been further clarified to emphasize the novelty of our weighted heuristic approach.

8: Introduction: The introduction is too short. It needs to be extensive and improved. (i) Contextualization, (ii) Literature review, (iii) Literature gap, (iv) Objectives, (v) Research questions, (vi) Paper organization.

Author’s Response: We sincerely thank the reviewer for this comprehensive feedback. We recognize that the original introduction was too brief and lacked the necessary academic depth. We have performed a major revision of Section 1 to address all six points mentioned:

• i) Contextualization: We have expanded the opening paragraphs to provide a broader context on global urban logistics and the specific socio-economic risks associated with HAZMAT transportation in high-density megacities.
• ii) Literature Review: The literature review has been significantly bolstered with 20 additional citations (2023–2025). We have integrated recent studies on Multi-Criteria Decision-Making (MCDM), Intelligent Transportation Systems (ITS), and contemporary path-planning algorithms to provide a more consistent background .
• iii) Literature Gap: We have explicitly articulated the research gap, focusing on the lack of real-time environmental hazard integration (such as chemical plume dispersion) within traditional deterministic routing frameworks.
• iv) & v) Objectives and Research Questions: These have been clearly stated in separate, dedicated subsections to guide the reader. We now define two primary research questions regarding expert consensus integration and safety improvement, followed by three specific research objectives.
• vi) Section Presentation: The introduction now concludes with a paragraph outlining the organization of the subsequent sections of the manuscript, providing a clear roadmap for the reader.

The revised Introduction is now approximately three times its original length, providing a much more robust foundation for the study.

9) Materials and Methods

9-i & 9-ii: The main methods are not presented and are cited without references.

Author’s Response: We apologize for the lack of clarity in the initial presentation of the methodology. We have now restructured Section 2 to explicitly present the three-phase research design : 

1. Phase 1: Factor Validation using the Delphi Technique . 
2. Phase 2: Weight Calculation via the Analytic Hierarchy Process (AHP) . 
3. Phase 3: Route Optimization using the enhanced A* algorithm, benchmarked against Dijkstra and ACO .

Furthermore, we have added the necessary seminal and contemporary references for each method to ensure academic rigor: * Delphi-AHP: References have been added to justify the consensus-building and weighting process. * A* Algorithm: Citations for the improved search logic have been integrated. * Dijkstra & ACO: Standard references for these baseline algorithms are now included in Sections 2.5 and 2.6.

* iii) Delphi-AHP Context: We have clarified that the Delphi technique was used exclusively for factor validation and consensus among 17 experts (Phase 1) , while the AHP was subsequently applied to quantify the relative weights (Phase 2). 

2. iv) The table with Saaty's fundamental scale is cited without referencing the author and without making its purpose clear. We have clarified the source and purpose of Saaty’s fundamental scale in Section 2.1. A note has been added to Table 1 explicitly stating that the scale is based on Saaty (1980), as consistently applied in established MCDM literature [8, 9]. This ensures the transparency of how qualitative expert judgments are converted into quantitative weights for the AHP process.

* vi) Factor Numbering: The gaps in numbering (e.g., A1 to A3) represent factors that were discarded by experts during the Delphi consensus rounds as being non-critical for HAZMAT expressway response. 

*v : "The term 'Factor' refers to the independent criteria used in the MCDM framework to evaluate route safety and efficiency."

*vi : "The non-sequential numbering (e.g., A1 to A3) indicates factors that were excluded during the Delphi consensus rounds based on expert feedback."

* vii) Consensus Handling: We have detailed the use of the Net Change principle. As shown in Table 3, N=17 was selected as the stability threshold where net change in expert responses reached its minimum (0.04), ensuring high group consensus . 

* viii) D2 Weight (0.00): Factor D2 (ERT Location) received a weight of 0.00 because, in our specific mission planning scenario, the starting location is a fixed constant for all evaluated algorithms, thus it does not influence the comparative selection of a safer path. 

*-ix: The calculation of h(n) uses the Manhattan Distance. In expressway networks, where movement is restricted to specific arcs, wouldn't the Euclidean Distance or a heuristic based on the network's topography be more accurate?

Author’s Response: We appreciate the reviewer's insightful observation regarding the heuristic function. While Euclidean distance is a common choice for A*, we opted for Manhattan Distance (L1 norm) for the following reasons: 

1. Topographical Consistency: The Chaloem Maha Nakhon Expressway network is modeled as a series of specific arcs and nodes representing physical junctions. Manhattan distance, which sums the absolute differences of coordinates, often provides a more robust estimate of the "grid-like" movement required to navigate between non-linear interchanges compared to the straight-line "as-the-crow-flies" Euclidean approach. 
2. Algorithmic Efficiency: In our coordinate-based grid simulation, Manhattan distance remains an admissible heuristic that prevents the underestimation of costs while guiding the search more aggressively toward the goal. This contributed to the superior search efficiency observed in our results, where the proposed A* explored only 150 nodes, compared to the 220 nodes explored by Dijkstra’s algorithm. 
3. Mathematical Simplicity: Given that our network nodes represent critical physical junctions such as on-ramps and toll plazas, the Manhattan metric effectively mimics the sequential nature of navigating through these restricted arcs.

• ix) Heuristic Choice: We justified the use of Manhattan Distance in Section 2.4.1. In an expressway grid where movement is constrained by specific junctions and interchanges, Manhattan distance provides a more realistic heuristic than Euclidean "as-the-crow-flies" distance.
• *x(ACO Hyperparameters): We appreciate the reviewer’s suggestion regarding the reproducibility of the ACO results. We have now specified the hyperparameters in Section 2.6. Specifically, the simulation utilized 50 artificial ants with a pheromone evaporation rate (ρ)of 0.5. The parameters controlling the relative influence of pheromone intensity (∝) and heuristic visibility (β) were set to 1.0 and 2.0, respectively. These settings prioritize the expert-weighted heuristic information (ηij) to guide the search towards safer and more efficient route segments while maintaining sufficient exploration of the solution space.

10) Results: The results are presented briefly.

We appreciate the reviewer's detailed critique of our results. We have expanded Section 3 to include comprehensive statistical and operational analyses:

* i, ii, & viii) Statistical Testing: We have now included a Paired t-test based on 30 simulation trials. The analysis confirms that the improvements, though numerically small, are statistically significant (p < 0.05). This is critical for the 'Golden Hour' in emergency response where seconds matter. 

* iii) Wind Dynamics (ALOHA): The 1 m/s wind was chosen as a worst-case scenario for concentration, as it creates a larger, circular impact zone (plume). In stronger directional winds, the ALOHA model produces an elongated plume, which the algorithm would navigate even more easily by identifying a clear 'upwind' or 'crosswind' path based on the dynamic Safety Score S(n). 

* iv) Search Efficiency (A* vs Dijkstra): Although S(n) adds a cost, it acts as a penalty that effectively prunes high-risk branches early in the search . Combined with a highly directional Manhattan heuristic h(n) on the linear expressway network, A* avoids the 'blind' circular exploration of Dijkstra, leading to only 150 nodes explored versus 220 . 

* v, vi, & vii) Simulation Parameters: We have added the missing details: 30 scenarios were simulated ; the road network consists of approximately 220 nodes/arcs representing the Chaloem Maha Nakhon Expressway; and the average execution time is < 1.0 second, ensuring real-time feasibility.

Point 11: Discussion and conclusions refinement.

Author’s Response: We have carefully refined the final sections of the manuscript: * i) Formatting: We have removed redundant paragraphs in the Discussion section and ensured that the "5. Conclusions" header is now correctly formatted in bold. * ii) Literature Expansion: As addressed in Point 8, we have integrated 18 additional contemporary references (2023–2025) to provide a more robust comparison with existing literature .

• iii) Limitations: A new subsection (4.7 Limitations and Future Work) has been added to discuss the constraints of the current study, such as the focus on single-chemical spills and the reliance on expert elicitation. 

* iv) Scientific Contribution: The Conclusion has been rewritten to emphasize the novelty of the Delphi-AHP-A* integration and its practical value for HAZMAT emergency logistics, rather than merely repeating numerical results .

Author Response File: Author Response.pdf

Reviewer 3 Report (Previous Reviewer 2)

Comments and Suggestions for Authors

Thank you for the opportunity to review this manuscript. The topic is relevant and meaningful, and the study shows practical potential. However, the current version still requires substantial clarification in terms of data collection, methodological transparency, and the presentation of results.

(1) In Table 2, “Factors for Comparison,” please provide a clearer justification for the selection of each factor and support it with relevant references from the literature.

(2) In Section 2.2, “Data collection,” the statement “The researchers initiated this phase by identifying a target expert group” is too vague. Please specify who the respondents were, how many experts participated, and what their professional backgrounds were. The survey questionnaire should also be included in the Appendix.

(3) Please explain clearly how the survey questionnaire was used within the Analytic Hierarchy Process (AHP) to produce Table 3, “Weighting of Factor Importance.” The full procedure for deriving these weights should be presented step by step.

(4) The logic around Lines 155–156 is unclear. Before the statement “As shown in Table 4, although Dijkstra’s algorithm offers a slightly shorter travel time, the Proposed A algorithm achieves an 8–10% higher cumulative safety score by strategically bypassing high-risk nodes identified by the ALOHA simulation,” the manuscript does not explain how Table 4 was constructed or what exactly was calculated. Please clarify what Table 4 contains, how the metrics were computed, and what this comparison is intended to demonstrate.

(5) In Lines 165–166, the manuscript states that “The factor importance weights derived from the AHP analysis will be subsequently utilized to develop the Cost Function for the A* Algorithm, Dijkstra’s Algorithm, and Ant Colony Optimization in the next phase of the research.” However, it is not explained how these weights are incorporated into each algorithm. Please clarify this in detail. For example, does Dijkstra’s algorithm consider only distance, or was it modified to incorporate multiple weighted factors?

(6) The relationship between “HAZMAT risks” and the factors listed in Table 2 is not clearly explained. Please clarify how HAZMAT risks are conceptually and quantitatively connected to the comparison framework.

(7) In Figure 3, “Spatial visualization of the study area on the Chaloem Maha Nakhon Expressway, comparing the routing trajectories of different algorithms,” the study area is only shown at a general spatial level. Please report the specific number of nodes and links in the network used for the routing analysis.

(8) Please specify the software, program, or algorithmic procedure used to generate Figure 5.

(9) Please explain why “Google Map (shortest route)” in Table 5 is different from “Dijkstra’s Algorithm.” If both are intended to represent shortest-path solutions, the reason for the discrepancy should be clarified.

(10) Overall, the manuscript would benefit from a clearer presentation of the methodological workflow, especially the connections among questionnaire design, AHP weighting, risk simulation, cost-function construction, and route optimization results.

Author Response

(1) In Table 2, “Factors for Comparison,” please provide a clearer justification for the selection of each factor and support it with relevant references from the literature.

• We have updated Table 2 to include clear justifications for each factor, supported by relevant literature on HAZMAT logistics and urban emergency response.

 

(2) In Section 2.2, “Data collection,” the statement “The researchers initiated this phase by identifying a target expert group” is too vague. Please specify who the respondents were, how many experts participated, and what their professional backgrounds were. The survey questionnaire should also be included in the Appendix.

• "The expert panel consisted of 17 practitioners: 5 transportation engineers, 7 expressway rescue specialists, and 5 emergency medicine experts, each with over 10 years of experience."

 

(3) Please explain clearly how the survey questionnaire was used within the Analytic Hierarchy Process (AHP) to produce Table 3, “Weighting of Factor Importance.” The full procedure for deriving these weights should be presented step by step.

(4) The logic around Lines 155–156 is unclear. Before the statement “As shown in Table 4, although Dijkstra’s algorithm offers a slightly shorter travel time, the Proposed A algorithm achieves an 8–10% higher cumulative safety score by strategically bypassing high-risk nodes identified by the ALOHA simulation,” the manuscript does not explain how Table 4 was constructed or what exactly was calculated. Please clarify what Table 4 contains, how the metrics were computed, and what this comparison is intended to demonstrate.

• We have clarified the stepwise procedure of the AHP and the calculation logic of the Safety Score in Sections 2.3 and 2.8.2.
• (1) AHP Procedure (Table 4): The weights were derived following these steps: 1. Experts performed pairwise comparisons using Saaty’s 9-level scale (Table 1). 2. A comparison matrix was constructed, and the Eigenvector was calculated and normalized to produce the Global Weights (W_i). 3. Consistency was verified (CR < 0.1) to ensure reliability. The final weights are presented in Table 4.
• (2) Safety Score Calculation (Equation 4): The Safety Score for each path is an aggregate value calculated by multiplying the global weight of each factor (W_i) by its corresponding dynamic risk score (S_i, scaled 1–5) at each node, as defined in Equation 4. This provides a quantitative measure of risk mitigation, where a higher score indicates a safer route.

(5) In Lines 165–166, the manuscript states that “The factor importance weights derived from the AHP analysis will be subsequently utilized to develop the Cost Function for the A* Algorithm, Dijkstra’s Algorithm, and Ant Colony Optimization in the next phase of the research.” However, it is not explained how these weights are incorporated into each algorithm. Please clarify this in detail. For example, does Dijkstra’s algorithm consider only distance, or was it modified to incorporate multiple weighted factors?

• We have clarified the integration of AHP weights into each algorithm in Sections 2.4 through 2.6.
• Dijkstra’s Algorithm (Baseline): To address the reviewer’s specific question, Dijkstra’s algorithm was not modified to include safety weights. It serves as a deterministic baseline focusing strictly on minimizing physical distance and travel time. This allows us to benchmark the safety improvements of the proposed model against standard shortest-path navigation. * Proposed A* Algorithm: The AHP weights (W_i) are integrated via the Safety Score S(n) in the improved cost function: f(n) = W_d x g(n) + W_d x h(n) + W_s x S(n). Here, S(n) is the weighted sum of all risk factors validated by experts. * Ant Colony Optimization (ACO): The AHP weights are incorporated into the Heuristic Visibility (ηij). Specifically,  ηij is defined as the inverse of the weighted cost function from Equation 3. This ensures that the artificial ants are more likely to deposit pheromones on paths that balance both operational speed and expert-validated safety.

(6) The relationship between “HAZMAT risks” and the factors listed in Table 2 is not clearly explained. Please clarify how HAZMAT risks are conceptually and quantitatively connected to the comparison framework.

• We have expanded the explanation of the relationship between HAZMAT risks and the factors listed in Table 2 within Sections 2.8.2 and 4.3.

* Conceptually: "HAZMAT risk" is treated as a multi-dimensional construct integrating the source of the hazard (e.g., F1, F3 modeled via ALOHA ), the vulnerability of the surrounding environment (e.g., B1, B2 population/business density ), and operational success factors (e.g., A1 traffic fluidity ).

• Quantitatively: These factors are linked through the Safety Score (S(n)). Each factor is assigned a dynamic risk value (S_i, 1-5) based on real-time simulation data, which is then multiplied by its expert-validated AHP weight (W_i). This aggregated score is integrated directly into the A* cost function as a penalty term , ensuring that the algorithm quantitatively prioritizes routes with the lowest cumulative hazard exposure.

(7) In Figure 3, “Spatial visualization of the study area on the Chaloem Maha Nakhon Expressway, comparing the routing trajectories of different algorithms,” the study area is only shown at a general spatial level. Please report the specific number of nodes and links in the network used for the routing analysis.

• The road network model for the Chaloem Maha Nakhon Expressway was digitized into a graph consisting of 220 physical nodes—including on-ramps, off-ramps, interchanges, and toll plazas—and their corresponding directional links (edges) representing the road segments. This level of granularity ensures that the routing analysis captures all possible maneuvers within the 10-kilometer study corridor.

 

(8) Please specify the software, program, or algorithmic procedure used to generate Figure 5.

• We have clarified the reason for the discrepancy between these two metrics in Section 4.2. Google Maps (Shortest Route) utilizes a proprietary, real-time dynamic graph that incorporates live traffic speeds and hidden road constraints. In contrast, Dijkstra’s Algorithm in this study was executed on a controlled static graph of the expressway network, using only physical distances as weights. This ensures a consistent, non-weighted baseline to scientifically evaluate the safety gains of our proposed Delphi-AHP-weighted model.

 

9: Discrepancy between Google Maps (shortest route) and Dijkstra’s Algorithm.

Author’s Response: We have clarified the reason for the discrepancy between these two shortest-path baselines in Section 4.2. Google Maps (Shortest Route) operates on a proprietary, real-time dynamic graph provided by the API, which includes live traffic data and hidden constraints. In contrast, Dijkstra’s Algorithm was performed on our specifically modeled static graph of 220 nodes representing the physical expressway network. By using this static, non-weighted baseline for Dijkstra, we were able to provide a rigorous and controlled mathematical comparison to measure the exact safety and efficiency gains of our proposed Delphi-AHP-weighted $A^*$ algorithm

 10: Overall, the manuscript would benefit from a clearer presentation of the methodological workflow, especially the connections among questionnaire design, AHP weighting, risk simulation, cost-function construction, and route optimization results.

Author’s Response: We sincerely agree with this observation. To improve methodological transparency, we have completely restructured the introductory part of Section 2 (Materials and Methods) and updated Figure 1 . The connection between each stage is now explicitly presented through a five-phase workflow:

* Phase 1 (Consensus): Details the 17-expert Delphi process and factor validation . 

* Phase 2 (Weighting): Explains the transition from qualitative expert scales to Global Weights (W_i) via AHP . 

* Phase 3 (Algorithmic Logic): Describes how these weights were integrated into the A cost function* (Equation 3) as safety penalty terms . 

* Phase 4 (Risk Inputs): Connects the ALOHA hazard simulation directly to the dynamic risk scores (S_i) used by the algorithm . 

* Phase 5 (Benchmarking): Links the final routing results to the performance evaluation across 30 simulation trials .

This restructuring ensures that the logical flow from expert elicitation to real-time route optimization is transparent and reproducible.

Author Response File: Author Response.pdf

Reviewer 4 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The paper makes a solid contribution by using expert opinion to add safety factors like chemical plumes into a route planner, showing it can save time and reduce risk for hazmat teams. Its method is strong, combining expert surveys (Delphi) with a ranking system (AHP) to build the algorithm's decision-making priorities.

1. A weakness is that it doesn't explain one of the twelve key factors it uses (A4), which makes the model hard to reproduce fully.
2. The main weights in the algorithm (0.6 for distance, 0.4 for safety) are justified, but the paper could better show how they were specifically chosen from the expert data.
3. It briefly mentions the algorithm is fast, but could say more about how easy it would be to actually use this system in a different city's emergency response.

Comments on the Quality of English Language

could be improved

Author Response

“The paper makes a solid contribution by using expert opinion to add safety factors like chemical plumes into a route planner, showing it can save time and reduce risk for hazmat teams. Its method is strong, combining expert surveys (Delphi) with a ranking system (AHP) to build the algorithm's decision-making priorities.”

Author’s Response: We sincerely thank the reviewer for the positive evaluation of our work and the recognition of the methodological strength of the Delphi-AHP integration. We have addressed the specific suggestions to enhance the reproducibility and generalizability of the proposed framework as follows:

Point 1: Clarification of Factor A4

“A weakness is that it doesn't explain one of the twelve key factors it uses (A4), which makes the model hard to reproduce fully.”

Author’s Response: We agree that a precise definition is essential for reproducibility. We have revised Table 2 to provide a quantitative definition for A4 (Route Complexity). It is now explicitly defined as the frequency of high-friction decision points (such as interchanges and sharp curves) per kilometer. This metric allows other researchers to calculate the complexity score objectively using standard GIS data. The revised text also highlights the impact of this factor on the cognitive load of drivers and the mechanical risks for heavy HAZMAT tankers.

Point 2: Justification of Weighting Parameters (0.6 / 0.4)

“The main weights in the algorithm (0.6 for distance, 0.4 for safety) are justified, but the paper could better show how they were specifically chosen from the expert data.”

Author’s Response: We have expanded Section 2.4.3 (Justification of Weighting Parameters) to clarify the origin of these values. The secondary weights, W_d = 0.6 and W_s = 0.4, were established by aggregating the final consensus from the expert panel.

• The 0.6 weight for distance (W_d) reflects the experts' prioritization of the 'Golden Hour' (Operational Fluidity), which accounted for approximately 60% of the total priority in the consensus rounds.
• The 0.4 weight for safety (W_s) was calibrated as a 'Penalty Multiplier.' When applied to the raw Safety Scores (scaled 1–5), it ensures that safety factors provide a mathematically significant cost increase (up to 2.0) to bypass critical hazards without causing redundant detours for minor risks.

Point 3: Practicality and Scalability (Generalizability)

“It briefly mentions the algorithm is fast, but could say more about how easy it would be to actually use this system in a different city's emergency response.”

Author’s Response: We have added a new discussion in Section 4.5 (Operational Trade-offs and Generalizability) to address the system's practical scalability. We clarified that the framework is location-agnostic and designed for "plug-and-play" adaptation. Emergency agencies in other metropolitan areas can implement this system by:

1. Importing their local GIS road network graph (via OpenStreetMap or proprietary APIs).

Recalibrating the Delphi-AHP weights to reflect local urban densities, safety protocols, and specific HAZMAT regulations.
This modularity ensures that the core optimization logic remains robust regardless of the geographical topology.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report (New Reviewer)

Comments and Suggestions for Authors

The current form of the manuscript can be accepted

Author Response

Reviewer's Comment: "The current form of the manuscript can be accepted."

 

Author’s Response: We sincerely thank the reviewer for their positive assessment and for recommending the acceptance of our manuscript. We are very pleased that the research design, methodology, and results were found to be clear and well-presented. Although no specific revisions were requested by this reviewer, we have further refined the manuscript based on the constructive feedback from other reviewers to ensure the highest possible quality of the final work . We truly appreciate your time and professional endorsement of our study.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

Comments and Suggestions for Authors

The authors worked on the requested revisions. Thank you for that. Overall, the manuscript is more robust. However, some details still need to be revised:

1) The references are not consistent. For example, Saaty (1980) is cited as reference 8, but it does not appear in the references. Review all references.

2) Although the quality of the figures is better, some still need improvement. Figure 2, for example, is of low quality.

3) In the discussions, the way the sections are separated is poor. In subsection 4.1, the colon has extra spacing. Place the subsections in the form of titles or rearrange the way this section is allocated.

4) There is extra spacing on line 696.

5) The tables and their respective titles appear in different ways throughout the work. Standardize them. There are also errors with extra spacing. 6) Although the authors cite 20 new citations, these do not appear in the highlighted references.

7) Highlight the changes with a color. Several methods were used to list the changes.

8) It makes no sense to use a criterion with a weight of 0.

Author Response

comment 1: The references are not consistent. For example, Saaty (1980) is cited as reference 8, but it does not appear in the references. Review all references.

Author’s Response: We sincerely apologize for the inconsistency in the reference list. We have conducted a comprehensive audit of all in-text citations and the bibliography to ensure perfect synchronization and adherence to the "Order of Appearance" rule.

The following corrective actions were taken: 

1. Insertion of Missing Foundations: We have added the foundational work of Saaty (1980) as Reference [8] and Linstone & Turoff (1975) as Reference [9] to correctly support the AHP scale and Delphi consensus stability claims mentioned in the text. 
2. Sequential Re-indexing (+2 Shift): Due to the insertion of these two missing references at positions 8 and 9, all subsequent references (previously 8–38) have been shifted by two positions (now 10–40). 
3. Table 2 Verification: All citations within Table 2 have been updated to match the new sequential numbering (e.g., factor A1 now cites [1, 10, 18] instead of [1, 8, 16]) to ensure contextual accuracy. 
4. Formatting: All updated citation numbers within the manuscript have been highlighted in red as requested by the Editor to facilitate the review of these changes.

 

comment 2: Although the quality of the figures is better, some still need improvement. Figure 2, for example, is of low quality.

Author’s Response (Update): We have entirely re-designed Figure 2 using high-definition graphic tools to ensure professional visual quality. The updated figure has been exported at a high resolution (300 DPI) with optimized font sizes for all axis labels and factor identifiers, ensuring full compliance with the journal’s legibility requirements.

 

comment 3: In the discussions, the way the sections are separated is poor. In subsection 4.1, the colon has extra spacing. Place the subsections in the form of titles or rearrange the way this section is allocated.

Author’s Response: We sincerely appreciate this stylistic suggestion to improve the manuscript's readability. We have completely restructured Section 4 (Discussion) by converting all inline headings into formal, bolded subheadings (Sections 4.1 to 4.7). * Typographical Correction: The extra spacing before the colon in subsection 4.1 has been removed, and colons have been eliminated from all subsection titles to adhere to standard academic formatting. * Structural Rearrangement: Each discussion point (Travel Time, Distance, Safety Score, Computational Efficiency, and Generalizability) now has a dedicated, clearly labeled subsection, providing a much better separation of ideas as suggested by the reviewer

comment 4: "There is extra spacing on line 696."

Author’s Response: We thank the reviewer for their meticulous attention to detail. The extra spacing on line 696, specifically within the bulleted list describing the "Shortest Distance" factor, has been identified and removed. We have also conducted a thorough check of the entire manuscript to ensure consistent spacing and to eliminate any other potential typographical errors.

 

comment 5: "The tables and their respective titles appear in different ways throughout the work. Standardize them. There are also errors with extra spacing."

Author’s Response: We have standardized all table titles throughout the manuscript to follow a consistent format: "Table X. Title." as per the journal's guidelines. The inconsistent use of colons and irregular spacing in titles (previously seen in Tables 1-8) has been corrected. Additionally, we have conducted a full audit of the document to remove any extra spacing between tables and text to ensure a professional layout .

 

comment 6: "Although the authors cite 20 new citations, these do not appear in the highlighted references."

Author’s Response: We sincerely apologize for this oversight. We have now meticulously highlighted all 20+ new citations added during the previous revision.

• In-text citations: All new and re-indexed reference numbers within the manuscript are now highlighted in red. * Reference List: All new bibliographic entries added to the references section (including those from 2023–2025 and the foundational Saaty/Linstone works) have been highlighted in red to make them clearly visible to the editor and reviewers .

comment 7: "Highlight the changes with a color. Several methods were used to list the changes."

Author’s Response: We sincerely apologize for the lack of consistency in our previous highlighting methods. Following the reviewer’s suggestion and the Editor’s checklist, we have now standardized the way revisions are presented.

• Standardized Method: All changes, additions, and re-indexed citations throughout the manuscript have been marked using a consistent Red Font.

• Consistency Check: We have removed all other previous marking methods (such as bolding or underlining used for identification) to ensure the revised manuscript is professional and easy to navigate. Every modification, from structural changes in the Discussion section to the updated reference list, is now clearly and uniformly identified in red.

comment 8: "It makes no sense to use a criterion with a weight of 0."

Author’s Response: We completely agree with the reviewer’s logical assessment. While D2 (ERT Location) was initially identified as a potential risk factor during the Delphi Phase , the subsequent AHP pairwise comparisons by the expert panel resulted in a priority weight of 0.00, indicating its negligible impact compared to critical factors like traffic fluidity and chemical concentration.

• Corrective Action: To ensure mathematical clarity and streamline the model, we have removed D2 from Table 5 and Table 6. The final cost function and safety score calculations now correctly focus on the 11 significant criteria that carry non-zero weights, while still maintaining a total aggregate weight of 1.00.

 

Author Response File: Author Response.docx

Reviewer 3 Report (Previous Reviewer 2)

Comments and Suggestions for Authors

Thank you.

Author Response

General Comments:

The reviewer indicated that all sections "can be improved." We would like to thank the reviewer for this general guidance, which prompted us to conduct a comprehensive overhaul of the entire manuscript.

Author’s Response:

We have taken the reviewer’s feedback seriously and have implemented extensive improvements across all sections as follows:

1. Introduction & References: We have strengthened the background by adding over 20 recent citations (2023–2025) to reflect current trends in ITS and HAZMAT logistics. We also added foundational references for the Delphi-AHP method (Saaty, 1980; Linstone, 1975) as requested .

2. Research Design & Methods: The methodological workflow has been restructured into a five-phase process for better clarity (Figure 1). We also refined the mathematical logic by removing the factor with a zero weight (D2) to ensure the research design is robust and logical.

3. Results & Discussion: We have clarified the performance metrics, including the justification for the 10.1 km distance and the trade-off between travel time and safety . The Discussion section was completely rearranged into formal subheadings (4.1–4.7) for better readability.

4. Figures & Tables: All figures, specifically Figure 2, have been re-rendered in high resolution (300 DPI). All table titles have been standardized to a consistent format (Table X.) throughout the work.

5. Statistical Validation: To support our conclusions, we conducted 30 simulation trials and confirmed the statistical significance of our findings (p < 0.05) using paired t-tests.

Author Response File: Author Response.docx

Reviewer 4 Report (Previous Reviewer 1)

Comments and Suggestions for Authors

The authors have revised the paper

Comments on the Quality of English Language

could be improved

Author Response

General Comments:

The reviewer noted that the paper has been revised but suggested that several sections and the English language "could be improved." We appreciate this feedback and have treated it as a mandate for a thorough linguistic and structural refinement.

Author’s Response:

In response to the reviewer’s comments, we have implemented the following major improvements:

1. English Language Enhancement: The entire manuscript has undergone a rigorous English language review and professional proofreading. We have focused on improving the flow, correcting grammatical nuances, and ensuring academic tone throughout the paper. Specifically, we refined the Discussion and Conclusion sections to be more concise and impactful.

2. Comprehensive Section Overhaul:
3. Introduction: We updated the literature review with over 20 recent citations (2023–2025) and added foundational methodology references [8, 9] to provide a more robust academic background
4. Methodology: We clarified the research design into a five-phase workflow (Figure 1) and removed logically redundant criteria (D2 with zero weight) to enhance the appropriateness of the research design.
5. Results & Figures: All figures, including Figure 2, have been re-rendered at 300 DPI for maximum clarity. All table titles have been standardized to "Table X. Title." for professional presentation.
6. Discussion & Conclusion: These sections were rearranged into formal subheadings (4.1–4.7) to ensure a clearer presentation of results and stronger support for our conclusions.

3. Standardized Highlighting: As requested, every single change, including linguistic corrections and re-indexed citations, has been uniformly highlighted in Red Font to ensure full transparency.

Author Response File: Author Response.docx

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents a valuable application of the A* algorithm, enhanced with expert-derived weights from the Delphi technique and Analytic Hierarchy Process (AHP).

1. I have read the paper and think that there might be a problem in Cost Function Derivation (Section 2.4.3). The derivation of the distance weight Wd​=0.6is mathematically incorrect.The text states: "The value was then rounded up to 0.6 to ensure that distance is assigned a significantly higher weight... The initial average calculation is as follows: Wd​=(0.25+0.18)/2=0.215." The average of 0.25 and 0.18 is indeed 0.215. However, "rounding up" 0.215 does not yield 0.6; this is a substantial and unexplained leap.
2. The "Safety Score" is a crucial metric for evaluating the algorithms. Equation (4) and Table 4 show scores Si​for each factor (e.g., A1=4, A3=3). Please explain why and how these specific scores (ranging from 2 to 5) were assigned?
3. Table 5 presents results from what appears to be a single simulation run. You might do the repeated studies.
4. Figure 1 is Missing. The manuscript text refers to "Figure 1: Weighting of Factor Importance" , but where is Fig.1?
5. The entire manuscript seems to be very rough. It only has one Figure to show this work.

Author Response

comment 1 : 

Reviewer 1

1. I have read the paper and think that there might be a problem in Cost Function Derivation (Section 2.4.3). The derivation of the distance weight Wd​=0.6is mathematically incorrect.The text states: "The value was then rounded up to 0.6 to ensure that distance is assigned a significantly higher weight... The initial average calculation is as follows: Wd​=(0.25+0.18)/2=0.215." The average of 0.25 and 0.18 is indeed 0.215. However, "rounding up" 0.215 does not yield 0.6; this is a substantial and unexplained leap.
2. The "Safety Score" is a crucial metric for evaluating the algorithms. Equation (4) and Table 4 show scores Si​for each factor (e.g., A1=4, A3=3). Please explain why and how these specific scores (ranging from 2 to 5) were assigned?
3. Table 5 presents results from what appears to be a single simulation run. You might do the repeated studies.
4. Figure 1 is Missing. The manuscript text refers to "Figure 1: Weighting of Factor Importance" , but where is Fig.1?
5. The entire manuscript seems to be very rough. It only has one Figure to show this work.
   Response : We thank the reviewer for the thorough and constructive feedback. We have meticulously revised the manuscript to address the concerns regarding mathematical clarity, data transparency, and analytical depth.
   - 

   Regarding Cost Function Derivation ($W_d = 0.6$):

   Response: We apologize for the lack of clarity in our initial explanation. The transition from 0.215 to 0.6 was not a mathematical "rounding" in the traditional sense, but a strategic weighting assignment. In emergency response, operational efficiency (reaching the site) must hold a dominant priority over other risks. We have revised Section 2.4.3 to clarify that W_d = 0.6$ was selected to ensure the algorithm remains primarily efficiency-driven, with safety (W_s = 0.4$) serving as a robust secondary filter. The term "rounding up" has been removed to avoid confusion.

   - 

   Regarding the "Safety Score" (2 to 5) Assignment:

   Response: The safety scores ($S_i$) were derived through a consensus-building process during the Delphi sessions with 17 experts. These values represent the "Potential Risk Impact" on a 5-point Likert scale, where 5 denotes critical risk and 1 signifies minimal risk. We have added Section 2.8.2 to explicitly detail this scoring methodology and how qualitative expert intuition was quantified for the cost function.

   Regarding Repeated Studies (Simulation Reliability):

   Response: We agree that a single run is insufficient. The revised manuscript now reports results based on 30 independent simulation trials under peak-hour conditions. Table 5 has been updated to reflect these average values, ensuring the statistical reliability of the reported 3-5% efficiency gain and 8-10% safety improvement.

   Regarding Figure 1 and the "Rough" Manuscript:

   Response: We sincerely apologize for the omission of Figure 1 in the initial submission; it has now been correctly embedded. To address the "rough" nature of the manuscript, we have added Table 6 (Route Selection and Spatial Reasoning Analysis) and significantly expanded the discussion. This provides a deeper, node-by-node analysis of why the proposed algorithm outperforms standard methods, moving beyond simple numerical reporting.

 

Reviewer 2 Report

Comments and Suggestions for Authors

The methodology adopted in this study is not sufficient to support the stated conclusions, and the research problem is not clearly defined. It remains unclear whether the paper addresses a path optimization problem or a vehicle routing problem (VRP). Moreover, several methodological choices appear unreasonable or insufficiently justified. In particular, the role of the Google Maps approach is unclear, and it is not specified what it is used to calculate within the proposed framework.

(1) The abstract does not clearly specify the decision problem investigated in this study. The research objective, decision variables, and the type of optimization problem being addressed are not sufficiently clear from the abstract.
(2) The paper employs Dijkstra’s Algorithm, Ant Colony Optimization, and Google Maps approaches, but their respective applicability is not clearly explained. It is unclear what type of problem each method is designed to solve, why these methods are compared with one another, and what level of transportation demand is considered. In addition, the descriptions of these methods are limited, and explicit model formulations are missing.
(3) There is an inconsistency in the parameter settings related to the distance weight. Specifically, line 176 defines the weight of distance as Wd=0.6, whereas line 179 reports 
𝑊𝑑=(0.25+0.18)/2=0.215. These inconsistent values require clarification and correction.
(4)There is an error in section numbering. At line 242, the subsection titled “3.7.1 Emergency Scenario” does not align with the overall structure of the manuscript and should be revised.
(5) The content of Section “3.3 Data Utilized” is not consistent with its title. The material presented in this section goes beyond data description, and the section heading or structure should be adjusted accordingly.
(6) The manuscript lacks a dedicated results analysis section. Adding such a section would help interpret the results more clearly and better support the conclusions drawn from the analysis.
(7) The analyzed area is not clearly specified, and the routing results are insufficiently presented. If the study focuses on route selection, the selected paths should be explicitly shown and discussed to improve clarity and reproducibility.

(8) The paper claims that the proposed algorithm achieves a 3–5% reduction in travel time and an 8–10% increase in safety; however, it is unclear which specific methods are being compared to obtain these results. The corresponding routing solutions for the compared methods are not presented, and the specific path differences or road segments that contribute to the reported improvements are not clearly identified. As a result, the basis of these performance gains cannot be adequately evaluated.

Author Response

comment 2

Reviewer 2

 

Comments and Suggestions for Authors

The methodology adopted in this study is not sufficient to support the stated conclusions, and the research problem is not clearly defined. It remains unclear whether the paper addresses a path optimization problem or a vehicle routing problem (VRP). Moreover, several methodological choices appear unreasonable or insufficiently justified. In particular, the role of the Google Maps approach is unclear, and it is not specified what it is used to calculate within the proposed framework.

(1) The abstract does not clearly specify the decision problem investigated in this study. The research objective, decision variables, and the type of optimization problem being addressed are not sufficiently clear from the abstract.
(2) The paper employs Dijkstra’s Algorithm, Ant Colony Optimization, and Google Maps approaches, but their respective applicability is not clearly explained. It is unclear what type of problem each method is designed to solve, why these methods are compared with one another, and what level of transportation demand is considered. In addition, the descriptions of these methods are limited, and explicit model formulations are missing.
(3) There is an inconsistency in the parameter settings related to the distance weight. Specifically, 

line 176 defines the weight of distance as Wd=0.6, whereas line 179 reports
??=(0.25+0.18)/2=0.215. These inconsistent values require clarification and correction.
(4)There is an error in section numbering. At line 242, the subsection titled “3.7.1 Emergency Scenario” does not align with the overall structure of the manuscript and should be revised.
(5) The content of Section “3.3 Data Utilized” is not consistent with its title. The material presented in this section goes beyond data description, and the section heading or structure should be adjusted accordingly.
(6) The manuscript lacks a dedicated results analysis section. Adding such a section would help interpret the results more clearly and better support the conclusions drawn from the analysis.
(7) The analyzed area is not clearly specified, and the routing results are insufficiently presented. If the study focuses on route selection, the selected paths should be explicitly shown and discussed to improve clarity and reproducibility.

(8) The paper claims that the proposed algorithm achieves a 3–5% reduction in travel time and an 8–10% increase in safety; however, it is unclear which specific methods are being compared to obtain these results. The corresponding routing solutions for the compared methods are not presented, and the specific path differences or road segments that contribute to the reported improvements are not clearly identified. As a result, the basis of these performance gains cannot be adequately evaluated.

Response :

We sincerely thank the reviewer for the rigorous and insightful critique. We have fundamentally restructured the manuscript to clarify the research problem, refine the methodology, and provide a detailed analysis of the routing results.

1. Problem Definition (Path Optimization vs. VRP):

Response: We have clarified in the revised Abstract and Introduction that this study addresses a Multi-Criteria Path Optimization problem (point-to-point) within an emergency context, not a Vehicle Routing Problem (VRP). The decision variables and optimization objectives are now explicitly defined in the Abstract.

2. Applicability of Methods & Google Maps Role:

Response: We have expanded Section 2.5 to 2.7 to explain the rationale for each method. Dijkstra and ACO serve as algorithmic baselines for shortest-path and heuristic search, while the Google Maps API provides a "real-world baseline" (standard consumer-grade navigation) to compare against our specialized HAZMAT-aware model. This benchmarking demonstrates the necessity of integrating non-traditional spatial risks.

3. Parameter Consistency (W_d):

Response: We have corrected the inconsistency in Section 2.4.3. As clarified to Reviewer 1, W_d=0.6$ is a strategic weight assigned to efficiency, informed by (but not a direct average of) the AHP-ranked factors. The confusing calculation has been removed.

4. Section Numbering & Structure (3.7.1, 3.3):

Response: We have corrected the numbering errors and renamed Section 3.3 to "Data Integration and Environmental Parameters" to better reflect its comprehensive content.

5. Result Analysis & Routing Solutions (Points 6, 7, and 8):

Response: To address these critical points, we have added Section 3.4 (Comparative Performance Results) and Section 3.5 (Route Selection and Spatial Reasoning Analysis).

• We have explicitly identified that the 3–5% and 8–10% gains are measured against the Dijkstra/Google Maps baselines.

• We have added Table 6, which provides a node-by-node comparison of the paths. This table illustrates exactly which road segments were selected by our A* algorithm to bypass high-risk zones, thereby providing the basis for the reported performance gains.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author have revised the paper, however, the quality of this paper still needs to be improved. I barely see any figures on this work. Also, I cannot judge whether the author have done the experiments or not, because there is only 1 figure. We do not know the algorithms process, etc. My Suggestion:

1. Create a figure illustrating the overall research workflow, from expert selection and data collection (Delphi-AHP) through to the simulation and evaluation phases. This would provide a high-level understanding of the study design.
2. A crucial missing element is a map of the expressway case study. This map should clearly mark the incident location, the three ERT starting points, and the different routes calculated by each algorithm (A*, Dijkstra, ACO, Google Maps). This would make the spatial reasoning in Table 6 immediately understandable.
3. Include a screenshot or a cleaned-up diagram from the ALOHA software output showing the chemical plume dispersion overlaid on the expressway map. This would visually justify why certain routes are safer than others.
4. Convert the data in Table 5 into a multi-bar chart comparing travel time, distance, and safety scores across all algorithms. This allows for quicker, more intuitive comparison than a table.
5. Explicitly state that the key innovation is the integration of the Delphi-derived, multi-criteria cost function into the A* algorithm's heuristic. Rephrase phrases like "treats the HAZMAT emergency response as a high-stakes application" to more actively state that this application drives the need for the novel cost function. Clearly separate what you did(methodology) from what you found(results: 3-5% time reduction, 8-10% safety improvement).
6. Equation (3) appears to have a typo and a conceptual ambiguity that needs clarification.
7. Please do the parameter responses analysis to justify your parameter choices because The choice of weights W_d = 0.6and W_s = 0.4is critical

Author Response

Dear Reviewer,

We would like to express our sincere gratitude for your constructive comments and valuable suggestions regarding our manuscript titled "Performance Optimization of Multi-Criteria Route Planning Algorithms: A Case Study in HAZMAT Emergency Response."

We have carefully reviewed your feedback and have made significant revisions to improve the quality, clarity, and visual presentation of the paper. We believe these changes have substantially strengthened the manuscript.

Please find below our point-by-point responses to your specific comments in the attached file

We hope these revisions satisfactorily address your concerns. We are confident that the inclusion of the workflow diagram, map overlays, and performance charts has significantly enhanced the clarity and impact of our research.

Sincerely,

Miss Wipaporn K.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The response is not organized in a point-by-point manner, and many statements are vague without indicating which comment is being addressed, what was changed, or where the change was made. Several key concerns are not clearly answered.

Author Response

Dear Reviewer,

We sincerely apologize for the lack of organization and specific referencing in our previous response. We understand that this oversight made it difficult to verify our revisions and track the changes effectively. We deeply regret any confusion or frustration this may have caused.

To rectify this, we have completely restructured our response. In this revised submission, we provide a rigorous point-by-point explanation, explicitly citing the Section numbers and Figure numbers where changes have been implemented.

We believe the extensive additions of visual evidence (maps, charts, and workflow diagrams) and the clarification of the methodology now directly address the key concerns regarding the paper's clarity and empirical validity.

Please find below our detailed responses as attached file

 

We hope that this structured response and the explicit inclusion of visual evidence (Figures 1, 3, 4, and 5) successfully clarify the "vague statements" from the previous round. We have endeavored to make this revision as precise and transparent as possible.

Sincerely,

Miss Wipaporn K.

Author Response File: Author Response.pdf

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

 

1. The abstract effectively summarizes the study's goal and conclusion. However, it should more precisely state the core innovation. Instead of "enhanced A* algorithm," specify that the enhancement is the Delphi-AHP-weighted cost function.
2. Also, the improvement metrics ("3-5% reduction in travel time," "8-10% safety score increase") need a clearer baseline (e.g., "compared to Dijkstra's algorithm").
3. The methodology mentions benchmarking against Dijkstra's, ACO, and Google Maps. However, the description of how ACO was implemented is vague. The paper states that ACO will use the AHP weights but does not detail how the pheromone update rules were modified to incorporate these multi-criteria weights.
4. The Safety Score (Equation 4, Safe Score = Σ(Wi × Si)) is a central metric, but its calculation in Table 4 is confusing. The scores (Si) for each factor (e.g., A1=4, F2=4) are presented without explanation of how they were assigned for the specific simulated route. It appears these are fixed scores for the entire route, which oversimplifies reality.
5. There is a critical discrepancy between the text and Table 5. The text in Section 3.4 states, "the proposed Delphi-A* algorithm yielded a travel time reduction of approximately 3-5% relative to both Dijkstra's algorithm..." However, Table 5 shows Dijkstra's time as 13.0 mins and A* as 12.5 mins, which is a 8% reduction. The text also claims a "0.7-point differential in mean scores (4.2 for A* vs. 3.5 for Dijkstra)," but Table 5 lists the scores as 3.8 (A*) and 3.5 (Dijkstra), a 0.3-point differential (an 8.6% increase, not 10%). The 4.2 score is not present in the table.
6. The discussion successfully highlights the advantages of the proposed A* algorithm but lacks a critical analysis of the trade-offs. For instance, the 3-5% time saving, while valuable, might be within the margin of error for real-world traffic fluctuations.
7. The use of "Number of Nodes Explored" as a metric for computational efficiency is valid. However, the paper should briefly define what constitutes a "node" in the context of the expressway network graph (e.g., intersections, on/off-ramps). This clarifies the scale of the problem.
8. The study is a compelling case study based on a specific expressway in Thailand. The conclusions' generalizability to other urban networks with different topologies, data availability, or emergency response protocols should be discussed.
9. The conclusion effectively summarizes the findings but could be more forward-looking. It should reiterate the key innovation (the Delphi-AHP-A* integration) and more strongly state the study's contribution to the fields of computational logistics and emergency management.

Author Response

Point 1: The abstract should more precisely state the core innovation (Delphi-AHP-weighted cost function) and improvement metrics with a clearer baseline.

Response: We agree with this suggestion. We have revised the abstract to explicitly state that the core innovation is the "Delphi-AHP-weighted cost function." We also clarified the baselines, specifying that the 3.8% time reduction and 8.6% safety score increase are relative to Dijkstra’s algorithm.

Point 2: The description of ACO implementation is vague regarding how weights were used in the pheromone update.

Response: We have clarified the ACO methodology in Section 2.6. We added an explanation stating that the standard transition probability rule was modified by defining the heuristic visibility (nij) as the inverse of our weighted cost function: nij = 1/[Wd x g(n) + Wd x h(n) + Ws x S(n)]. 

This ensures the ants are guided by the expert-weighted safety and efficiency scores.

Point 3: The calculation of the Safety Score in Table 4 is confusing; the scores (Si) appear fixed for the entire route. Response: We apologize for the confusion. We have updated the Table 4 caption and added a footnote to clarify that the scores listed represent the maximum potential risk weight for each category. In the actual simulation, the specific Si value for each node is dynamic, varying between 1 and 5 based on the real-time attributes of that specific route segment (e.g., actual proximity to the chemical plume).

Point 4: There is a critical discrepancy between the text and Table 5 regarding scores and percentage reductions. Response: We sincerely apologize for this oversight. We have corrected the text in Section 3.4 to strictly align with the data in Table 5. The revised text now correctly states a 0.3-point differential (3.8 for A* vs. 3.5 for Dijkstra) and calculates the exact percentage improvements (3.8% for travel time and 8.6% for safety) to ensure accuracy.

Point 5: The discussion lacks a critical analysis of the trade-offs (e.g., is the 3-5% time saving significant?) and generalizability. Response: We have expanded the Discussion (Section 4.5) to argue that while the time saving is marginal, the trade-off is justified by the significant gain in safety (avoiding catastrophic risks). We also added a discussion on generalizability, noting that the framework can be adapted to other urban topologies by recalibrating the weights (W_d, W_s).

Point 6: The paper should define what constitutes a "node" in the context of the expressway network graph. Response: We have added a clear definition in Section 2.8.1, stating that a "node" represents critical physical junctions, including on-ramps, off-ramps, highway interchanges, and toll plazas.

Point 7: The conclusion effectively summarizes the findings but could be more forward-looking. It should reiterate the key innovation (the Delphi-AHP-A* integration) and more strongly state the study's contribution to the fields of computational logistics and emergency management. Response: We appreciate this constructive suggestion. We have rewritten the Conclusion (Section 5) to be more forward-looking. The revised conclusion now explicitly reiterates the core innovation—the translation of qualitative expert consensus into quantitative algorithmic constraints via the Delphi-AHP-A framework*. Furthermore, we have strengthened the statement regarding the study's contribution, positioning it as a scalable decision-support blueprint for future Intelligent Transportation Systems (ITS) and HAZMAT emergency logistics.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The response is not organized in a point-by-point manner, and many statements are vague without indicating which comment is being addressed... Several key concerns are not clearly answered.

Author Response

Comment 1: The abstract does not clearly specify the decision problem investigated in this study. The research objective, decision variables, and the type of optimization problem being addressed are not sufficiently clear.

Response: We have revised the Abstract to explicitly define the problem scope.

• Action Taken: We clarified that this is a "Multi-Criteria Path Optimization problem" (not VRP). We specified the objective is to dynamically balance "operational efficiency" and "public safety" using a Delphi-AHP weighted cost function.

• Location: Abstract, Lines 2-6.

• Revised Text: "The primary objective is to develop and evaluate a novel pathfinding approach by integrating a Delphi-AHP-weighted cost function into the A algorithm... This research treats the hazardous material (HAZMAT) emergency response as a high-stakes application to benchmark the proposed model against established baselines, specifically Dijkstra’s algorithm and Ant Colony Optimization (ACO)."*

Comment 2: The respective applicability of Dijkstra, ACO, and Google Maps is not clearly explained... explicit model formulations are missing.

Response: We have expanded the methodology to justify the selection of each algorithm and defined the Google Maps role.

• Action Taken:

  1. Google Maps Role: In Section 2.5 (Dijkstra’s Algorithm) and Section 3.4, we clarified that Google Maps serves as a "real-world baseline" (standard consumer-grade navigation) to demonstrate why standard fastest-path logic fails in HAZMAT scenarios.

  2. Algorithm Formulations: We added the explicit cost function equation for the A* algorithm in Equation 3 (Section 2.4.4) and explained the modified pheromone update rule for ACO in Section 2.6.

• Location: Section 2.4.4 (Eq. 3), Section 2.5, Section 2.6.

Comment 3: There is an inconsistency in the parameter settings related to the distance weight (W_d=0.6 vs calculation of 0.215).

Response:

We have resolved this inconsistency by standardizing the weights based on the expert consensus strategy.

• Action Taken: We removed the confusing calculation. We have explicitly defined the weights in Section 2.4.3 (Justification of Weighting Parameters).

• Location: Section 2.4.3.

• Explanation: We established W_d = 0.6 (Efficiency) and W_s = 0.4 (Safety) as the aggregate weights for the cost function. As explained in the text: "A weight of W_d = 0.6 ensures that the algorithm remains highly sensitive to travel distance and traffic delays... preventing excessive or impractical rerouting."

Comment 4: There is an error in section numbering (Section 3.7.1).

Response:

• Action Taken: We have corrected all section numbering. The Emergency Scenario is now properly located in Section 2.7.

• Location: Section 2.7.

Comment 5: The content of Section “3.3 Data Utilized” is not consistent with its title.

Response:

• Action Taken: We have renamed this section to "3.3 Data Integration and Environmental Parameters" to accurately reflect its content, which includes map data, traffic data, and ALOHA chemical dispersion modeling.

• Location: Section 3.3.

Comment 6: The manuscript lacks a dedicated results analysis section.

Response:

• Action Taken: We have added a dedicated Section 3. Results which is subdivided into:

  • 3.4 Comparative Performance Results: Analyzing time, distance, and safety scores.

  • 3.5 Route Selection and Spatial Reasoning Analysis: Analyzing the logic behind the route choices.

• Location: Section 3, specifically subsections 3.4 and 3.5.

Comment 7: The analyzed area is not clearly specified, and the routing results are insufficiently presented. Selected paths should be explicitly shown.

Response: We have significantly improved the visualization of the results.

• Action Taken:

  1. Study Area: We specified the area as "Chaloem Maha Nakhon Expressway" in Section 2.7.1.

  2. Visualization: We added Figure 3 and Figure 4. These maps explicitly show the routing trajectories of Dijkstra (Red) vs. Proposed A* (Green) overlaid on the GIS map with the chemical plume (Yellow/Orange zones).

  3. Route Details: We added Table 6, which provides a segment-by-segment comparison of the spatial reasoning (e.g., "A* avoids Bon Kai exit due to ALOHA model").

• Location: Figures 3 & 4, Table 6.

Comment 8: It is unclear which specific methods are being compared to obtain the reported 3–5% reduction... specific path differences are not identified.

Response: We have clarified the baselines and the specific path differences.

• Action Taken:

  1. Baselines: In Section 3.4 and Section 4.1, we explicitly state that the 3.8% time reduction and 8.6% safety increase are calculated relative to Dijkstra’s algorithm.

  2. Path Differences: As mentioned in Response 7, Table 6 now explicitly lists the specific road segments (e.g., "Bypass to Rama III") that contributed to these improvements.

• Location: Section 3.4, Section 4.1, Table 6.

Author Response File: Author Response.pdf