An Arch-Bridge Topology-Based Expressway Network Structure and Automatic Generation
Round 1
Reviewer 1 Report
According to a definition from the web, “network topology is used to describe the physical and logical structure of a network.” So, the network is given. Please explain the meaning of topologies (in plural), how and why do you create different topologies for a given network.
In line 79 you write “Specifically, we proposed a novel arch-bridge topology-based expressway network structure”. It is not clear, whether and how do you want to change the existing network structure. Please clarify.
The term “Arch-bridge topology” is a key element of the paper. Its definition appears first only in page 5, through a complicated set of mathematical expressions. Could you give a simple verbal definition of this topology?
Using gantries is one of the several possible ETC systems. Other ETC systems without gantries, e.g. GPS-based car following are also used in several countries. Please explain, whether your results are relevant or not for those systems.
Please explain the meaning of “the user penetration rate exceeds 80%” and its relation to the content of the research.
Author Response
Thank you very much for your hard work.
We have responded to each of your valuable comments, as detailed below.
Point 1: According to a definition from the web, “network topology is used to describe the physical and logical structure of a network.” So, the network is given. Please explain the meaning of topologies (in plural), how and why do you create different topologies for a given network.
In line 79 you write “Specifically, we proposed a novel arch-bridge topology-based expressway network structure”. It is not clear, whether and how do you want to change the existing network structure. Please clarify.
Response1: In the ETC system, OBU (On Board Unit) are mounted on vehicles and RSU (Road Side Unit) are mounted on the roadside. Through DSRC (Dedicated Short Range Communication) technology, they communicate with each other to collect all relevant information of passing vehicles. In other words, the expressway network uses ETC gantries as key nodes that can collect information from all vehicles in the entire area and all samples. However, according to the "ETC gantry system technical requirements for expressways" issued by the Ministry of Transport in 2019, ETC gantries are deployed not only at the entrances and exits of each toll station but also on road sections before changes in traffic flow occur, such as entrance/exit ramps and interchanges. Obviously, this is significantly different from the traditional expressway road network structure, i.e., intersections as nodes. But in the current practical situation, the expressway intersection as a node can not get all the passing vehicle information, which makes it difficult to realize the fine management of expressway.
Therefore, we propose a new type of expressway network structure based on the ETC system. As shown in Fig. 2, this topology shape is similar to an arch-bridge, so it is named "arch-bridge topology".
Point 2: The term “Arch-bridge topology” is a key element of the paper. Its definition appears first only in page 5, through a complicated set of mathematical expressions. Could you give a simple verbal definition of this topology?
Response 2: The term “Arch-bridge topology” is a key element of the paper. It is defined as rigorous and academic. To make it easier to understand, we have marked the topography in different colour(s) over the existing image and added legends with explanations. Meanwhile, the set of road sections in Fig. 1 is abstracted one-by-one correspondingly to the set of edges in Fig. 2.
Please refer to the revised manuscript for the specific details.
Point 3: Using gantries is one of the several possible ETC systems. Other ETC systems without gantries, e.g. GPS-based car following are also used in several countries. Please explain, whether your results are relevant or not for those systems.
Response 3: Since ETC systems are widely used on expressway in many countries such as Europe, Japan, and China, we focus on gantry-based ETC systems in this study. However, although our approach focuses on gantry-based ETC systems, some of the principles and techniques are also of some reference value for these other types of systems. For example, the data mining and topology optimization rules we propose in this study can be applied to GPS-based systems to some extent.
Point 4: Please explain the meaning of “the user penetration rate exceeds 80%” and its relation to the content of the research.
Response 4: In this study, a user penetration rate of over 80% means that over 80% of the vehicles on the expressway are using the ETC service, which covers the vast majority of expressway users. These vehicles generate a large amount of transaction data every day, providing a comprehensive and representative data source for this work. Our study finds that compared to small samples of floating car data, ETC transaction data enables us to obtain information about the whole road network more easily, and also to complete the automatic generation of road network topology more efficiently.
Thank you again for your hard work.
Author Response File: 
Author Response.docx
Reviewer 2 Report
The article presents a relevant topic. The language quality is mostly acceptable. The structure is unconventional, which makes the article less easy to follow.
The Introduction section presents the challenge adequately and lists the references. There are some claims, though, where the refrences are missing (lines 51-59). These should be added.
The Literature review cites relevant sources and thoroughly describes the state-of-the-art in the field.
The sections 3 and 4 titled "Arch-bridge topology" and "Automatic generation of expressway network topology", respectively, should be joined in a "Methodology" section. The 3rd level subsections make the text hard to read and should be abolished in favour of conciseness and comprehensibility. The first paragraph of Section 3 (lines 233-245) fails to provide a sensible description of the current system. The authors should rewrite it to present the current system clearly. If possible, another figure with explanation of the terms (such as locations of ETC gantries, different from those in Fig. 1) should be added. Also the communication between vehicles and infrastructure should be described more precisely, possibly with references to used equipment and transfer protocols. The abbreviations (such as "RSU", "OBU") should be explained on first use. The terminology in the description of Fig. 1 (lines 261-267) is vague and doesn't reflect the actual figure. The terms used in the description (nodes other than gantries, edges, "up and down lanes", "upstream and downstream sections") should be more clearly marked in the figure. Suggestions: mark the topography in different colour(s) over the existing image of the actual road, or combine Fig. 2 and Fig. 1 into a single figure. The sets of nodes and edges in Definition 1 (line 269) should be defined in terms of the definition of the network in Figure 1. In Figure 2, all entities (also the edges) should be marked according to this definition. From the definition of the travel trajectory (in lines 317-322) the meaning of different gantry types ("entrance gantry, provincial border entrance gantry, normal gantry, exit gantry, and provincial border exit gantry") and its implication on the model should be explained. The Data anomaly (3) description (lines 350-353), although accurate, is very awkward - the authors should rephrase this using simpler terms. There is a reference to "gantry gantries" in line 371, which is probably not intended. The reference to the original Dijkstra's algorithm proposal in Numerische Mathemaitk should be added (lines 404-405).
The description of the hardware and software setup used to conduct the experiments in Section 5 (lines 465-468) is vague and possibly inaccurate. The authors should state the number of CPU cores the model ran on and avoid reffering to Jupyter Notebook as merely an "open-source web application" to avoid confusion. A reference to Jupyter home page should be added for completeness.
The quantities in equations 16 and 17 should be explained. The evaluation of the efficiency test (lines 540-543) should be explained in more detail (i.e. how many tests were done with each size of the dataset, how the average runtime has been computed and how it was assured it's correct). It would also be interesting if the actual time complexity was quantified and expressed in terms of O(f(n)) as already hinted with the red line in Fig. 9.
The conclusions in section 6 are accurate but somewhat short and dry. They would make more impact if they were expanded and re-written as free-flowing text rather than a list of numbered points.
Author Response
Thank you very much for your hard work.
We have responded to each of your valuable comments, as detailed below.
Point 1: The Introduction section presents the challenge adequately and lists the references. There are some claims, though, where the refrences are missing (lines 51-59). The reference to the original Dijkstra's algorithm proposal in Numerische Mathemaitk should be added (lines 404-405). These should be added.
Response1: The literature citation here is very necessary. We have added the relevant references.
- Zhu, L, Wen H. Automatic incremental identification and technology update of traffic road network data[J]. Traffic Information and Safety,2009,27(02):22-24+55.
- ZHANG S, GUAN G, ZOU F, ZHU D. Automatic detection of additional roads by combining floating vehicle and spatial grid[J]. Science Technology and Engineering,2012,12(22):5496-5501.
- Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische Mathematik, 1(1), 269-271.
Please refer to the revised manuscript for the specific details.
Point 2: The sections 3 and 4 titled "Arch-bridge topology" and "Automatic generation of expressway network topology", respectively, should be joined in a "Methodology" section. The 3rd level subsections make the text hard to read and should be abolished in favour of conciseness and comprehensibility.
Response 2: We agree that the recommendations are pertinent. Thus, we have restructured the article as follows.
- Methodology
3.1. Arch-bridge topology
3.2.1. Topology set generation and analysis
3.2.2. Candidate topology set optimization
Point 3: The first paragraph of Section 3 (lines 233-245) fails to provide a sensible description of the current system. The authors should rewrite it to present the current system clearly. If possible, another figure with explanation of the terms (such as locations of ETC gantries, different from those in Fig. 1) should be added. Also the communication between vehicles and infrastructure should be described more precisely, possibly with references to used equipment and transfer protocols. The abbreviations (such as "RSU", "OBU") should be explained on first use.
Response 3: We agree that the recommendations made by the reviewer are pertinent. The first paragraph of Section 3 has been rewrited as follows.
In the ETC system, OBU (On Board Unit) are mounted on vehicles and RSU (Road Side Unit) are mounted on the roadside. Through DSRC (Dedicated Short Range Communication) technology, they communicate with each other to collect all relevant information of passing vehicles. In other words, the expressway network uses ETC gantries as key nodes that can collect information from all vehicles in the entire area and all samples. However, according to the "ETC gantry system technical requirements for expressways" issued by the Ministry of Transport in 2019 [59], ETC gantries are de-ployed not only at the entrances and exits of each toll station but also on road sections before changes in traffic flow occur, such as entrance/exit ramps and interchanges. Obviously, this is significantly different from the traditional expressway road network structure, i.e., intersections as nodes. As shown in Fig. 1, red triangles are used to mark expressway intersections. But in the current practical situation, the expressway inter-section as a node can not get all the passing vehicle information, which makes it difficult to realize the fine management of expressway.
In addtion, The abbreviations (such as "RSU", "OBU") have been explained on first use.
Please refer to the revised manuscript for the remaining specific details.
Point 4: The terminology in the description of Fig. 1 (lines 261-267) is vague and doesn't reflect the actual figure. The terms used in the description (nodes other than gantries, edges, "up and down lanes", "upstream and downstream sections") should be more clearly marked in the figure. Suggestions: mark the topography in different colour(s) over the existing image of the actual road, or combine Fig. 2 and Fig. 1 into a single figure. The sets of nodes and edges in Definition 1 (line 269) should be defined in terms of the definition of the network in Figure 1. In Figure 2, all entities (also the edges) should be marked according to this definition. From the definition of the travel trajectory (in lines 317-322) the meaning of different gantry types ("entrance gantry, provincial border entrance gantry, normal gantry, exit gantry, and provincial border exit gantry") and its implication on the model should be explained.
Response 4: We appreciate your suggestions, and have marked the topography in different colour(s) over the existing image and added legends with explanations. Meanwhile, the set of road sections in Fig. 1 is abstracted one-by-one correspondingly to the set of edges in Fig. 2.
Fig. 1. A simple road network.
Fig. 2. Arch-bridge topology.
The meaning of different gantry types has been explained as follows.
The gantry type , representing toll station entrance gantry, provincial border entrance gantry, roadside gantry, toll station exit gantry, and provincial border exit gantry respectively. Specifically, both the toll station entrance gantry and the provincial border entrance gantry are entrance gantries, which are the starting gantry of the ETC trajectory. both the toll station exit gantry and the provincial border exit gantry are entrance gantries, which are the ending gantry of the ETC trajectory. The roadside gantry is installed at expressway roadside, which is the middle gantry of the ETC trajectory.
Please refer to the revised manuscript for the remaining specific details.
Point 5: The Data anomaly (3) description (lines 350-353), although accurate, is very awkward - the authors should rephrase this using simpler terms. There is a reference to "gantry gantries" in line 371, which is probably not intended.
Response 5: We use simpler and more accessible terms to describe data anomaly (lines 350-353), as described below.
Due to wireless interference, the vehicle-mounted OBU driving on the road up-line(downline) successfully communicates with the ETC antenna deployed on the road downline (upline), and the resulting topology is referred to as a opposite topology, as shown in Fig. 4(c).
In addition, the "gantry gantries" in line 371 has been revised in the manuscript.
Point 6: The description of the hardware and software setup used to conduct the experiments in Section 5 (lines 465-468) is vague and possibly inaccurate. The authors should state the number of CPU cores the model ran on and avoid reffering to Jupyter Notebook as merely an "open-source web application" to avoid confusion. A reference to Jupyter home page should be added for completeness.
Response 6: The experimental hardware and software setup for this work has been revised and is described as follows:
The experimental platform utilized an Intel(R) Core(TM) i9-10900K CPU with 10 cores and a base clock of 3.70GHz, along with 64GB RAM. The experiments were per-formed on the CentOS Linux release 7.9.2009 (Core) operating system. Python 3.8.8 was the programming language used for development, while Jupyter Notebook, an open-source interactive computing environment, was employed for conducting and presenting the experiments. For more information about Jupyter Notebook, please refer to its official website (https://jupyter.org/).
Point 7: The quantities in equations 16 and 17 should be explained.
Response 7: We have explained the quantities in equations 16 and 17 in the revised manuscript, as follows.
The precision of the method is determined by the excess topology generated in the reconstructed topology, with fewer excess unmatched topologies indicating higher precision, as shown in the following formula:
| precision = 1- #FP/(#TP+#FP) |
(16) |
Where, and indicates the number of correct topologies identified as correct topologies and the number of incorrect topologies identified as correct topologies in the generated topology set, respectively.
Likewise, the recall is determined by the number of unmatched topologies in the actual topology of the expressway network, with fewer unmatched topologies indicating higher recall, as shown in the following formula:
| recall = 1- #FN/(#TP+#FN) |
(17) |
Where, indicates the number of unmatched topologies in the actual topology of the expressway network.
Point 8: The evaluation of the efficiency test (lines 540-543) should be explained in more detail (i.e. how many tests were done with each size of the dataset, how the average runtime has been computed and how it was assured it's correct). It would also be interesting if the actual time complexity was quantified and expressed in terms of O(f(n)) as already hinted with the red line in Fig. 9.
Response 8: To evaluate the efficiency of the proposed method, we conducted experiments with datasets of varying sizes, which are respectively 1k, 10k, 100k, 1000k(1M) and 10000k(10M). We repeated the experiment 10 times on each order of magnitude and calculated the mean of the running time. Meanwhile, we give the running time linear fitting equation and its fitting curve.
Point 9: The conclusions in section 6 are accurate but somewhat short and dry. They would make more impact if they were expanded and re-written as free-flowing text rather than a list of numbered points.
Response 9: We have expanded and re-written as free-flowing text as follows.
In this work, we proposed an arch-bridge topology-based expressway network structure and automatic generation method. First, we propose a novel arch-bridge topology structure base on the ETC gantry system that redefines the topology of the expressway network, enabling the minimization and structuring of the complex and intertwined topology of the road network into its smallest unit. Second, by utilizing real ETC data with a user penetration rate of over 80%, our approach not only addresses the issue of insufficient data but also allows for the efficient and cost-effective extraction and improvement of the expressway road network. Third, through deep mining and analysis of ETC data characteristics and the design of optimized topology rules, our approach achieves rapid generation of the entire domain of the expressway road network topology. In addition, the expressway road network topology based on the ETC system enables real-time monitoring of all vehicle information, making refined management of the expressway system a possibility. In future work, it can be further developed in various ways, such as building feature engineering and using machine learning for training learning and building 3D road model.
Thank you again for your hard work.
Author Response File: Author Response.docx
Reviewer 3 Report
The proposed paper addresses a physical express road network that has been extensively equiped with Electronic Tool Collection in China. In order to exploit the large amount of data from the registration travel process, a methodology of the representation (modelling) of a such physical network was developed, in an adequate description.
1.However, the methodology is restricted to the particular situation of a dense implementation of ETC on a road network and the paper doesn’t sufficiently address the model generalisation and its validity in case of a lower density of such equipment.
2.Moreover, there is a hypotezys that is not sufficiently explained: the rate of user penetration of over 80% (from what?). It is suppose that the user penetration rate is understood as a road capacity utilisation rate. But, what is happening in case of lower capacity utilisation is not considered in the proposed methodology.
3. Because of such dense equipment for the tool collection, the issue of personal data confidentillity or protection should be also addressed. In other words, is it possible that the data for a specific individual to be manipulated or extracted?
4.The methodology itself, the concepts definition, the model development, the experiments and and results are well addressed.
5.The original coverage of the network with ETC equipment is important for data mining accuracy. The paper proposes also a method for determining the relationship between the coverage and the data size, precision, recall. However, there is no sufficient discussion on the results, especially at the lower coverage levels.
6.Conclusion section is not sufficiently developed.
7. The list of references is relevant and rather recent as long as more then a half of the total number is from last five years.
Author Response
Dear Reviewer,
Thank you very much for your hard work.
We have responded to each of your valuable comments, as detailed below.
Point 1: However, the methodology is restricted to the particular situation of a dense implementation of ETC on a road network and the paper doesn’t sufficiently address the model generalisation and its validity in case of a lower density of such equipment.
Response1: According to the definitions and models in the article, it can be seen that any physical network that fits into a similar ETC gantry topology is applicable, and we can abstract them as nodes and corresponding edges. For example, the camera topology deployed throughout the road network is also applicable.
Meanwhile, according to the "ETC gantry system technical requirements for expressways" issued by the Ministry of Transport in 2019 [59], ETC gantries are deployed not only at the entrances and exits of each toll station but also on road sections before changes in traffic flow occur, such as entrance/exit ramps and interchanges. At present, the total number of ETC gantries deployed on the province's 6,000 km-long expressway is about 1,000. Therefore, the ETC gantry deployment density shows as high as one gantry deployed for 2km when it is dense and as low as one gantry deployed for more than 40km when it is sparse. Theoretical analysis and experimental results show that our method is effective both when the gantries are dense and when they are sparse.
Point 2: Moreover, there is a hypotezys that is not sufficiently explained: the rate of user penetration of over 80% (from what?). It is suppose that the user penetration rate is understood as a road capacity utilisation rate. But, what is happening in case of lower capacity utilisation is not considered in the proposed methodology.
Response 2: In this study, a user penetration rate of over 80% means that over 80% of the vehicles on the expressway are using the ETC service, which covers the vast majority of expressway users. The remaining 20% of vehicles receive a Composite Pass Card (CPC) at the toll station entrance, which will record the vehicle's path for billing purposes and will be recycled at the toll station exit. These vehicles generate a large amount of transaction data every day, providing a comprehensive and representative data source for this work. Our study finds that compared to small samples of floating car data, ETC transaction data enables us to obtain information about the whole road network more easily, and also to complete the automatic generation of road network topology more efficiently.
Please refer to the revised manuscript for the specific details.
Point 3: Because of such dense equipment for the tool collection, the issue of personal data confidentillity or protection should be also addressed. In other words, is it possible that the data for a specific individual to be manipulated or extracted?
Response 3: All of us understand the reviewers' concerns. However, the gantry topology generation is done by obtaining only the passing path of the vehicle in this work. Therefore, vehicle information as well as personal information will be used after desensitization, i.e., all privacy-related information will be removed, so there is no need to worry about manipulation or extraction of data of a specific individual.
Point 4: The methodology itself, the concepts definition, the model development, the experiments and and results are well addressed.
Response 4: Thank you very much for the pertinent comments.
Point 5: The original coverage of the network with ETC equipment is important for data mining accuracy. The paper proposes also a method for determining the relationship between the coverage and the data size, precision, recall. However, there is no sufficient discussion on the results, especially at the lower coverage levels.
Response 5: We further analyzed the results, which are shown below.
We further delved into the evaluation of topological coverage and recall rates. Fig. 8 illustrates that under various levels of topological coverage, the topological recall rate is always greater than or equal to the topological coverage rate. Specifically, the corresponding recall rates are 0.34, 0.57, 0.74, 0.89, and 0.98 when the topological coverage is 0.2, 0.4, 0.6, 0.8, and 0.98, respectively. In other words, the recall is improved by 0.14, 0.17, 0.14, 0.08, and 0 for topological coverage of 0.2, 0.4, 0.6, 0.8, and 0.98, respectively. …
Please refer to the revised manuscript for the specific details.
Fig. 8. The relationship between recall and coverage
Point 6: Conclusion section is not sufficiently developed.
Response 6: We have added and expanded on the conclusion section as shown below.
In this work, we proposed an arch-bridge topology-based expressway network structure and automatic generation method. First, we propose a novel arch-bridge topology structure base on the ETC gantry system that redefines the topology of the ex-pressway network, enabling the minimization and structuring of the complex and in-tertwined topology of the road network into its smallest unit. Second, by utilizing real ETC data with a user penetration rate of over 80%, our approach not only addresses the issue of insufficient data but also allows for the efficient and cost-effective extraction and improvement of the expressway road network. Third, through deep mining and analysis of ETC data characteristics and the design of optimized topology rules, our approach achieves rapid generation of the entire domain of the expressway road network topology. In addition, the expressway road network topology based on the ETC system enables real-time monitoring of all vehicle information, making refined management of the expressway system a possibility. In future work, it can be further developed in various ways, such as building feature engineering and using machine learning for training learning and building 3D road model.
Point 7: The list of references is relevant and rather recent as long as more then a half of the total number is from last five years.
Response 7: Thank you very much for the pertinent comments. We have read and cited a large amount of relevant national and international literature around the article topic.
Thank you again for your hard work.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The authors have adequately addressed the issued raised in the first round of the review. The article is ready for publicaion once it's been thoroughly proofread.
Author Response
Thank you very much for your hard work.
We will proofread the article thoroughly and look forward to public publication.
Reviewer 3 Report
Autorii nu au imbunatatit in fapt secțiunea de concluzii, deoarece, au schimbat doar forma de prezentare, dar nu și conținutul acelei sectiuni. cause, in fact only form of presentation was changed but not the content.
Author Response
Thank you again for your hard work!
We have added and expanded on the conclusion section as shown below. Please offer your valuable advice and wisdom again.
A good and reasonable road network topology has important research significance and application value for traffic management and mining analysis. Therefore, in this work, we proposed an arch-bridge topology-based expressway network structure and automatic generation method for the first time. First, we propose a novel arch-bridge topology structure base on the ETC gantry system that redefines the topology of the expressway network, enabling the minimization and structuring of the complex and intertwined topology of the road network into its smallest unit. Second, by utilizing real ETC data with a user penetration rate of over 80%, our approach not only addresses the issue of insufficient data but also allows for the efficient and cost-effective extraction and improvement of the expressway road network. Third, through deep mining and analysis of ETC data features, optimization rules are designed for loop topology, reverse topology, missing topology, and opposite topology, which achieves rapid generation of the entire domain of the expressway road network topology. In addition, the expressway road network topology based on the ETC system enables real-time monitoring of all vehicle information, making refined management of the expressway system a possibility.
In the future, we believe that automatic incremental recognition and fast dynamic update of road network topology are worthy of further study, and on this basis, we will conduct further application research such as abnormal trajectory detection and recognition, travel path planning, vehicle group travel behavior mining and analysis.
