How Do Humanlike Behaviors of Connected Autonomous Vehicles Affect Traffic Conditions in Mixed Traffic?
Round 1
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsDear Authors,
the manuscript “How Human alike Connected Autonomous Vehicles affect traffic conditions in urban Environment?” is an interesting an important topic for the research community. The manuscript is improved very much compared to the earlier version.
However, it has still several issues, which need to be addressed for the acceptance of the manuscript.
Introduction: „three different human driving patterns“ -> where are these discussed? The reviewer just looked for driving pattern, and the authors did not introduce any of them! Why do the authors address these?
- Update: Okay, they are called “different driving modules”. But in that case please be consistent. That would help the reader
The last paragraph in the introduction is still not clear enough:
- What exactly was investigated in this research work? The reviewer can somehow understand the core idea (which is good, and relevant!), but the presentation is not good. It may help if the authors use bullet points to highlight the novelty of the paper. Furthermore, the usage of CAV/AV should be more consistent: Are the authors considering the benefits of CAV compared to AV? (Yes, the reviewer does know, after reading the paper, which is the case, however, the reader should know at the very beginning, what the research work will present!)
- Please try to describe this paragraph more thoughtful: 1) what is given, 2) what are the challenges, 3) what this paper will address and 4) how (e.g. with sensitivity analysis)
Section 2 is nice to have; however, it does not point out the necessity of this research work:
- What ‘can’ the state of the art (SOA)?
- Which aspects have been addressed by the SOA?
- Which not? (Why)
- How this research is novel compared to the SOA?
- What are the main benefits compared to SOA?
These aspects should be answered in the related work section!
Figure 3 does not help a better understanding, for the reader, how the scenario looks like.
Why does the word speed in Section 3.3.?.
Figure 4 is hard to read, pixel image with bad quality. Please replace it.
Result images: non-vector graphic images – hard to read
It is not clear whether the authors propose/show the benefits of CAV compared to AV or something else. It should be clearer in the related work section. The novelty of the paper should also be clearer in relation with the SOA.
The expression “higher penetration of CAV” is not specified in detail
The authors use not consistently what AV function and what CAV functions are, see “due to various benefits of AV technologies such as platooning”-> is here platooning AV or CAV?
Section 4.3. -> how can a function be calibrated? Tuned, parameter setting would suit better. What is the goal of the calibration? That will not clear from the text.
Penetration should be defined in the introduction -> the authors could state, this is the main objective of the study-> helps the reader
Fig 8 and Fig 10 -> Numbers are still cut down.
The terms in Fig 14 are not defined, so it is not clear for the reader what they exactly mean.
Please note the earlier comment of this reviewer, which are not addressed:
Old feedback
“The State of the art (SOA) should be restructured and extended by the following way
- Simulation Frameworks/concepts of pedestrian-automated vehicle interactions.
- Simulation Frameworks/concepts of automated and human-driven vehicles interactions
o Pure simulation frameworks/concepts
o Frameworks with possible human interactions
- Studies with real vehicles: concepts of automated and human-driven vehicles interactions
Then please add some more preliminaries, in which the PTV Vissim is presented in more detail. Further potentials and limitation of Vissim should be discussed. Furthermore, the Wiedermann model should be presented shortly.
Some of the works in literature focus on human (pedestrian) machine (automated vehicles). These are also simulations frameworks, which could be adapted for the study presented in the work.” See, for more details, the earlier comment of the reviewer!
Summary:
The manuscript presents an interesting and relevant research topic. For a publication, a further revision is necessary.
Comments on the Quality of English Language-
Author Response
Dear Reviewer,
Thank you for the intensive review. I have followed your directions step by step. Please note that all of them are taken into consideration. Please find my answers below:
Inquiry 1
Introduction: „three different human driving patterns“ -> where are these discussed? The reviewer just looked for driving pattern, and the authors did not introduce any of them! Why do the authors address these?
- Update: Okay, they are called “different driving modules”. But in that case please be consistent. That would help the reader
Feedback:
After replying to another reviewer, I fixed it to driving behaviour. This is more understandable for people with different backgrounds.
Inquiry 2
The last paragraph in the introduction is still not clear enough:
- What exactly was investigated in this research work? The reviewer can somehow understand the core idea (which is good, and relevant!), but the presentation is not good. It may help if the authors use bullet points to highlight the novelty of the paper. Furthermore, the usage of CAV/AV should be more consistent: Are the authors considering the benefits of CAV compared to AV? (Yes, the reviewer does know, after reading the paper, which is the case, however, the reader should know at the very beginning, what the research work will present!)
Feedback:
I have made significant change in introduction and conclusion after considering your and others opinion. The part in introduction and conclusion clarify the novelity of this research:
Even today, it is still not possible to obtain publicly available real traffic data from autonomous vehicles. This gap hinders the calibration and validation of the Connected Autonomous Vehicle (CAV), two important elements of microscopic traffic simulation. To overcome this gap, this study assumes that both CAVs and AVs would behave similarly to human-driven vehicles. The characteristics of driving behavior for connected and/or autonomous vehicles (C/AVs) need to be determined according to distinct driving behaviors, such as aggressive, normal, and safe (Sukennik 2018), reflecting three different human driving patterns makes it more human-like than other C/AVs. Vehicles will decide which parameters to follow from their permissible range in accordance with their regulating driving behavior s. Direct user selection of these driving behavior s is also possible through external connections, such as V2V, V2I, and emergency states. These behavior s exhibit wildly disparate responses to roads and traffic. A safe driving behavior is not permitted to drop the safe distance below the authorized limits, however an aggressive driving behavior may have a smaller safe distance with increased acceleration, resulting in C/AVs that resemble humans (Zeidler et al. 2018; Atkins 2016; Sukennik 2018).
The research findings show how traffic, and safety aspects are influenced by C/AVs that resemble humans. Additionally, it will use a sensitivity analysis platform to visualize how various AV driving behavior parameters interact with traffic performance. Speed, delay time, and travel time have all been used to examine traffic performance. In this process, only the interaction among motor vehicles is taken into account. For these types of studies, PTV VISSIM and Surrogate Safety Assessment Model (SSAM) must work together to identify the number of potential conflicts from the simulated data and vehicle trajectories to understand the safety implications of C/AVs. Examining how mixed traffic in an urban corridor will be impacted by human-like C/AVs is the primary objective of this study. To do this, the HV-AV-CAV ratio for three traffic flows—peak hour traffic demand, 20% below peak hour traffic demand, and 20% over peak hour traffic demand—will be studied. For C/AV, three different driving behaviors—aggressive, normal, and safe—will be evaluated. The second objective is to look at how the parameters of the AVs' driving behavior connect with traffic performance. To determine which car-following parameter has the greatest influence, this will be researched and studied. Following the identification of influential car following parameters, sensitivity analysis for a chosen traffic performance indicator will be carried out on them.
Conclusion:
The lack of real or natural data can be compensated by the similarity of the driving behavior of HV and CAV. It can be a good data input for calibration and validation until we get natural data from CAVs in real traffic. If CAVs are assumed to behave like human-driven vehicles and this is evaluated against natural data, this could be a benchmark for future research. This will reduce the dependency on test bed data for the assessment of CAVs in mixed traffic to get more realistic influence in urban areas.
Along with this consideration in calibration and validation, this study also defined three driving behaviors and human alike car following method. From the analysis, it is clear that Introduction of CAV in higher number in mixed traffic could be beneficial if they have interaction with many other CAVs. The pattern starts changing from 50-70% of penetration. Below 50% penetration, the system works inefficiently because of lack of available connecting vehicles. After 70%, the system would work better if interacting vehicles are higher in number.
After analyzing eight parameters from driving behavior using a sensitivity analysis, it becomes obvious that some of them play an influential role in traffic performance. Minimum clearance, look back distance and number of interacting vehicles play dominating role than look ahead distance, minimum collision time gain, accepted deceleration, standstill distance and gap time distribution.
Inquiry 3
Please try to describe this paragraph more thoughtful: 1) what is given, 2) what are the challenges, 3) what this paper will address and 4) how (e.g. with sensitivity analysis)
Feedback:
Your this concern is addressed here in introduction and method section.
Even today, it is still not possible to obtain publicly available real traffic data from autonomous vehicles. This gap hinders the calibration and validation of the Connected Autonomous Vehicle (CAV), two important elements of microscopic traffic simulation. To overcome this gap, this study assumes that both CAVs and AVs would behave similarly to human-driven vehicles. There are traffic volume and speed from human driven vehicles which could be used for C/AVs as well from the assumption in similarity. The characteristics of driving behavior for connected and/or autonomous vehicles (C/AVs) need to be determined according to distinct driving behaviors, such as aggressive, normal, and safe (Sukennik 2018), reflecting three different human driving patterns makes it more human-like than other C/AVs. Vehicles will decide which parameters to follow from their permissible range in accordance with their regulating driving behavior s. Direct user selection of these driving behavior s is also possible through external connections, such as V2V, V2I, and emergency states. These behaviors exhibit wildly disparate responses to roads and traffic. A safe driving behavior is not permitted to drop the safe distance below the authorized limits, however an aggressive driving behavior may have a smaller safe distance with increased acceleration, resulting in C/AVs that resemble humans (Zeidler et al. 2018; Atkins 2016; Sukennik 2018).
The research findings show how traffic, and safety aspects are influenced by C/AVs that resemble humans. Additionally, it will use a sensitivity analysis platform to visualize how various AV driving behavior parameters interact with traffic performance. Speed, delay time, and travel time have all been used to examine traffic performance. In this process, only the interaction among motor vehicles is taken into account. For these types of studies, PTV VISSIM and Surrogate Safety Assessment Model (SSAM) must work together to identify the number of potential conflicts from the simulated data and vehicle trajectories to understand the safety implications of C/AVs. Examining how mixed traffic in an urban corridor will be impacted by human-like C/AVs is the primary objective of this study. To do this, the HV-AV-CAV ratio for three traffic flows—peak hour traffic demand, 20% below peak hour traffic demand, and 20% over peak hour traffic demand—will be studied. For C/AV, three different driving behaviors—aggressive, normal, and safe—will be evaluated. The second objective is to look at how the parameters of the AVs' driving behavior connect with traffic performance. To determine which car-following parameter has the greatest influence, this will be researched and studied. Following the identification of influential car following parameters, sensitivity analysis for a chosen traffic performance indicator will be carried out on them.
Inquiry 4
Section 2 is nice to have; however, it does not point out the necessity of this research work:
- What ‘can’ the state of the art (SOA)?
- Which aspects have been addressed by the SOA?
- Which not? (Why)
- How this research is novel compared to the SOA?
- What are the main benefits compared to SOA?
These aspects should be answered in the related work section!
Feedback:
SOA- What do you mean by this? I have explained the process to deal with the issue of having no real world data for calibration and validation of CAVs. Please refer to introduction and method section which carry the idea of my methodology. There are very few works (only 2 work- PTV CoExist and Makridis et. al.) worked in similar approach so i do not have much to create a set of comparisons.
Inquiry 5
Figure 3 does not help a better understanding, for the reader, how the scenario looks like.
Feedback:
Figure 3 is not for scenario. It shows the study area and its different intersections in the PTV VISSIM interface. I am following other paper for presentation. In my case, the combination in experimental setup is still visible from visualization in the result section. I was successful to express it to other reviewers. The explanation of ratio of AV-CAV-HV will take more space and it will be still repeated in the result section.
Inquiry 6
Why does the word speed in Section 3.3.?.
Feedback:
The speed is a traffic performance indicator so it is also added in section 3.3 for explanation.
Inquiry 7
Figure 4 is hard to read, pixel image with bad quality. Please replace it. Result images: non-vector graphic images – hard to read
Feedback:
It is changed as per your suggestion. As i have obtained it from another research work, the quality is not in its best condition.
Inquiry 8
It is not clear whether the authors propose/show the benefits of CAV compared to AV or something else. It should be clearer in the related work section. The novelty of the paper should also be clearer in relation with the SOA.
The expression “higher penetration of CAV” is not specified in detail
Feedback:
Related work has stated the CAV has better performance than the AV and HV. The figure 1 shows the speed progress with higher penetration of CAV in the traffic. The triangular graphs show how % of any HV, AV or CAV influence the participation of each other.
Portion of related work is shown below:
Previous studies showed that higher penetration of CAV in the mixed traffic causes higher quality of traffic condition. It is to be expected that in the first stage of the implementation of C/AVs, the road capacity and thus the speed of vehicles will drop. With more of these kinds of vehicles on the road, the status of mobility, road safety and environmental conditions will improve due to various benefits of AV technologies such as platooning (Ahmed et al. 2022; Tettamanti et al. 2016). Figure 1 shows three ternary plots which depicts three different traffic demands, where the color level indicates the harmonic average speeds over the network. High traffic demand leads to a deterioration of traffic conditions as the number of interacting vehicles gradually increases and reaches technical limits (Makridis et al. 2018b; Mattas et al. 2018).
Higher penetration means more CAV comparatively to other vehicles.
Inquiry 9
The authors use not consistently what AV function and what CAV functions are, see “due to various benefits of AV technologies such as platooning”-> is here platooning AV or CAV
Feedback:
Platoon is for CAV. The 3.2 contains information about it.
3.2. Driving Behavior
Unlike the CAVs, the AVs do not act on platoon and do not slowing down in the traffic through connectivity. AVs are totally isolated and acts for its surrounding vehicles using its sensors.
Inquiry 10
Section 4.3. -> how can a function be calibrated? Tuned, parameter setting would suit better. What is the goal of the calibration? That will not clear from the text.
Feedback:
The function is not calibrated. SPSA is a calibration algorithm which takes decision to choose chosen parameter of car following behaviours. Please check
Qurashi, M. (2018): An Alternative Online Calibration approach for Dynamic Traffic Assignement Systems. Master. Technical University of Munich.
The calibration is used in microscopic transportation simulation to make it matchable to real world. We use real world data (also known as natural data) to compare it whatever we get as output from simulation. The matching is done in 95% confidential interval.
Inquiry 11
Penetration should be defined in the introduction -> the authors could state, this is the main objective of the study-> helps the reader
Feedback:
The penetration is not the main objective of this study. The main study is studying the C/AVs considering it as Human driven vehicles. We have defined the driving behaviours and had different penetration rate for HV, AV and CAV.
Inquiry 12
Fig 8 and Fig 10 -> Numbers are still cut down.
The terms in Fig 14 are not defined, so it is not clear for the reader what they exactly mean.
Feedback:
Figure 8 and 10 had format issue. Now, it is checked and is edited.
In the explanation of the figure 14, following description is added
The first value and the last value from each graphs (in total 8) from the figure 12 and 13 show how much variation has taken place in travel time. Figure 14 shows the percentage of variations between the initial and final value of individual driving parameters.
Inquiry 13
Simulation Frameworks/concepts of pedestrian-automated vehicle interactions.
- Simulation Frameworks/concepts of automated and human-driven vehicles interactions
o Pure simulation frameworks/concepts
o Frameworks with possible human interactions
- Studies with real vehicles: concepts of automated and human-driven vehicles interactions
Then please add some more preliminaries, in which the PTV Vissim is presented in more detail. Further potentials and limitation of Vissim should be discussed. Furthermore, the Wiedermann model should be presented shortly.
Some of the works in literature focus on human (pedestrian) machine (automated vehicles). These are also simulations frameworks, which could be adapted for the study presented in the work.” See, for more details, the earlier comment of the reviewer!
Feedback:
There is only one focus which is in the interaction in motorway. There is no separate plan for pedestrian, wheel chair.
So, Simulation Frameworks/concepts of automated and human-driven vehicles interactions are concern. The framework is defined in 3. Method. There is no other framework except this. Rest is mentioned in 4. Experimental setup.
After introducing HV, AV and CAV in the simulation environment with different ratio, they have been studied.
Again, i am following 2 research papers which are published. There is no limitation of PTV VISSIM which is connected in our research.
This is the details which we have seen other to use for discussing microscopic transportation simulation. We do not go to very deep. Even in the VISSIM manual, the defination is short and limitation is missing. The discussion regarding parameters are relatively bigger and that would be fundamental. When i was preparing the work, i had to follow reference of papers and go to help of VISSIM to learn and implement them.
The VISSIM, a microscopic traffic simulation software from PTV, has been used to perform the simulation. Moreover, SSAM have been used for additional inquires such as impact over accident. The parameters of car following model such as Wiedemann 74 and 99, are chosen in the VISSIM to represent the human driven vehicles for manoeuvre and lane behaviors.
Regarding the literature, i have used them connectedly. They were as per need. I am unable to extend it any longer.
Reviewer 2 Report (Previous Reviewer 3)
Comments and Suggestions for AuthorsAccept in present form.
Author Response
Dear reviewer,
Thank you for reviewing my work. It has helped me to improve our work.
Reviewer 3 Report (Previous Reviewer 4)
Comments and Suggestions for Authors The suggestions given earlier in the article have been sufficiently corrected. However, there are still some minor issues that can be enhanced. 1. It is suggested to highlight the novelty and contribution of the paper in the introduction section and abstract. 2. the future work section can be integrated into the conclusion section. In addition, the conclusion should be more detailed. 3. Besides the car following the model of C/AV and HV, the lane-changing model should be explained. In addition, the model used in the simulation should be highlighted.Author Response
Dear reviewer,
Thank you very much for your review. It has helped me to write a better paper. Please find the feedback for individual inquiries.
1. It is suggested to highlight the novelty and contribution of the paper in the introduction section and abstract.
Feedback:
The problem is mentioned in this line of the abstract section: Lack of reliable real-world data (also known as natural data) to calibrate and to evaluate the connected autonomous vehicles (CAV) simulation model is a major challenge.
The novelty is mentioned in this line of the abstract: To deal with this situation, one interesting methodology could be dealing the CAVs as conventional human driven vehicles and predict its possible characteristics based on the simulation inputs. The conventional human driven vehicles from real world, in this methodology, come to aid as benchmark to offer the measure of effectiveness (MoE) for the calibration and validation.
For the three most common driving behaviors, a sensitivity analysis of the behaviors of AVs and an effective assessment of CAVs in a mixed traffic environment were done to explore the human alike autonomous technology. The findings show that, up to a point, which is directly related to the quantity of interacting vehicles, the impact of CAVs is typically favorable. This study validates the approach and supports past studies by showing that CAVs perform better in traffic than AVs for traffic performance and safety aspects. On top of that, the sensitivity analysis has shown that enhancements in technology are required for obtaining the maximum advantages.
The novelty is mentioned in this line of the introduction:
Even today, it is still not possible to obtain publicly available real traffic data from autonomous vehicles. This gap hinders the calibration and validation of the Connected Autonomous Vehicle (CAV), two important elements of microscopic traffic simulation. To overcome this gap, this study assumes that both CAVs and AVs would behave similarly to human-driven vehicles. The characteristics of driving behavior for connected and/or autonomous vehicles (C/AVs) need to be determined according to distinct driving modules, such as aggressive, normal, and safe (Sukennik 2018), reflecting three different human driving patterns makes it more human-like than other C/AVs. Vehicles will decide which parameters to follow from their permissible range in accordance with their regulating driving modules. Direct user selection of these driving modules is also possible through external connections, such as V2V, V2I, and emergency states. These modules exhibit wildly disparate responses to roads and traffic. A safe driving module is not permitted to drop the safe distance below the authorized limits, however an aggressive driving module may have a smaller safe distance with increased acceleration, resulting in C/AVs that resemble humans (Zeidler et al. 2018; Atkins 2016; Sukennik 2018).
The novelty is mentioned in this line of the conclusion:
The lack of real or natural data can be compensated by the similarity of the driving behavior of HV and CAV. It can be a good data input for calibration and validation until we get natural data from CAVs in real traffic. If CAVs are assumed to behave like human-driven vehicles and this is evaluated against natural data, this could be a benchmark for future research. This will reduce the dependency on test bed data for the assessment of CAVs in mixed traffic to get more realistic influence in urban areas.
Along with this consideration in calibration and validation, this study also defined three driving behaviors and human alike car following method. From the analysis, it is clear that Introduction of CAV in higher number in mixed traffic could be beneficial if they have interaction with many other CAVs. The pattern starts changing from 50-70% of penetration. Below 50% penetration, the system works inefficiently because of lack of available connecting vehicles. After 70%, the system would work better if interacting vehicles are higher in number.
After analyzing eight parameters from driving behavior using a sensitivity analysis, it becomes obvious that some of them play an influential role in traffic performance. Minimum clearance, look back distance and number of interacting vehicles play dominating role than look ahead distance, minimum collision time gain, accepted deceleration, standstill distance and gap time distribution.
Only platoon building and traffic signal slowing are implemented from V2V and V2I features in this study. There are numerous other CAV advancements that can be studied, including vehicle-to-pedestrian (V2P), controlled mobility pattern, and early congestion warning (Iranmanesh et al. 2022). Research into C/AVs that resemble humans can be done using other car-following models besides Wiedermann's psycho-physical model. Implementing the Gipps car following model, which offers a safety distance concept, would be one interesting approach. This study also shown that a few car following behaviors get higher degree of influence in the performance. These can be used to study the CAVs and AVs in future researches.
2. The future work section can be integrated into the conclusion section. In addition, the conclusion should be more detailed.
Feedback:
The conclusion is extended with future work. As per your suggestion, i have included more details in the conclusion section.
Conclusion:
The lack of real or natural data can be compensated by the similarity of the driving behavior of HV and CAV. It can be a good data input for calibration and validation until we get natural data from CAVs in real traffic. If CAVs are assumed to behave like human-driven vehicles and this is evaluated against natural data, this could be a benchmark for future research. This will reduce the dependency on test bed data for the assessment of CAVs in mixed traffic to get more realistic influence in urban areas.
Along with this consideration in calibration and validation, this study also defined three driving behaviors and human alike car following method. From the analysis, it is clear that Introduction of CAV in higher number in mixed traffic could be beneficial if they have interaction with many other CAVs. The pattern starts changing from 50-70% of penetration. Below 50% penetration, the system works inefficiently because of lack of available connecting vehicles. After 70%, the system would work better if interacting vehicles are higher in number.
After analyzing eight parameters from driving behavior using a sensitivity analysis, it becomes obvious that some of them play an influential role in traffic performance. Minimum clearance, look back distance and number of interacting vehicles play dominating role than look ahead distance, minimum collision time gain, accepted deceleration, standstill distance and gap time distribution.
Only platoon building and traffic signal slowing are implemented from V2V and V2I features in this study. There are numerous other CAV advancements that can be studied, including vehicle-to-pedestrian (V2P), controlled mobility pattern, and early congestion warning (Iranmanesh et al. 2022). Research into C/AVs that resemble humans can be done using other car-following models besides Wiedermann's psycho-physical model. Implementing the Gipps car following model, which offers a safety distance concept, would be one interesting approach. This study also shown that a few car following behaviors get higher degree of influence in the performance. These can be used to study the CAVs and AVs in future researches.
3. Besides the car following the model of C/AV and HV, the lane-changing model should be explained. In addition, the model used in the simulation should be highlighted.
Feedback: Car following model of C/AV and HV as per PTV has been set to Widemann 99. It is mentioned in page 5. In the Table 1 and 2, all the parameters for the car following behaviour has been planned for 3 different behaviours: Aggressive, Normal and Cautious. I have followed a few papers which have similar kinds of presentation:
Makridis, M.; Mattas, K.; Ciuffo, B.; Raposo, M. A.; Toledo, T.; Thiel, C. (2018a): Connected and Automated Vehicles on a freeway scenario. Effect on traffic congestion and network capacity. Vienna, Austria: Proceedings of 7th Transport Research Arena TRA 2018.
Makridis, M.; Mattas, K.; Ciuffo, B.; Raposo, M. A.; Toledo, T.; Thiel, C. (2018b): Connected and Automated Vehicles on a freeway scenario. Effect on traffic congestion and network capacity. Vienna, Austria: Proceedings of 7th Transport Research Arena TRA 2018.
Round 2
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsDear Authors,
SOA means "state of the art".
Please address all the comments of the reviewer.
Author Response
Dear Reviewer,
I have same opinion like the last review. Please have a look. I have prepared the current manuscript as per the instruction of other reviewers and you.
First of all, thank you for the intensive review. I have followed your directions step by step. Please note that all of them are taken into consideration. Please find my answers below:
Inquiry 1
Introduction: „three different human driving patterns“ -> where are these discussed? The reviewer just looked for driving pattern, and the authors did not introduce any of them! Why do the authors address these?
- Update: Okay, they are called “different driving modules”. But in that case please be consistent. That would help the reader
Feedback:
After replying to another reviewer, I fixed it to driving behaviour. This is more understandable for people with different backgrounds.
Inquiry 2
The last paragraph in the introduction is still not clear enough:
- What exactly was investigated in this research work? The reviewer can somehow understand the core idea (which is good, and relevant!), but the presentation is not good. It may help if the authors use bullet points to highlight the novelty of the paper. Furthermore, the usage of CAV/AV should be more consistent: Are the authors considering the benefits of CAV compared to AV? (Yes, the reviewer does know, after reading the paper, which is the case, however, the reader should know at the very beginning, what the research work will present!)
Feedback:
I have made significant change in introduction and conclusion after considering your and others opinion. The part in introduction and conclusion clarify the novelity of this research:
Even today, it is still not possible to obtain publicly available real traffic data from autonomous vehicles. This gap hinders the calibration and validation of the Connected Autonomous Vehicle (CAV), two important elements of microscopic traffic simulation. To overcome this gap, this study assumes that both CAVs and AVs would behave similarly to human-driven vehicles. The characteristics of driving behavior for connected and/or autonomous vehicles (C/AVs) need to be determined according to distinct driving behaviors, such as aggressive, normal, and safe (Sukennik 2018), reflecting three different human driving patterns makes it more human-like than other C/AVs. Vehicles will decide which parameters to follow from their permissible range in accordance with their regulating driving behavior s. Direct user selection of these driving behavior s is also possible through external connections, such as V2V, V2I, and emergency states. These behavior s exhibit wildly disparate responses to roads and traffic. A safe driving behavior is not permitted to drop the safe distance below the authorized limits, however an aggressive driving behavior may have a smaller safe distance with increased acceleration, resulting in C/AVs that resemble humans (Zeidler et al. 2018; Atkins 2016; Sukennik 2018).
The research findings show how traffic, and safety aspects are influenced by C/AVs that resemble humans. Additionally, it will use a sensitivity analysis platform to visualize how various AV driving behavior parameters interact with traffic performance. Speed, delay time, and travel time have all been used to examine traffic performance. In this process, only the interaction among motor vehicles is taken into account. For these types of studies, PTV VISSIM and Surrogate Safety Assessment Model (SSAM) must work together to identify the number of potential conflicts from the simulated data and vehicle trajectories to understand the safety implications of C/AVs. Examining how mixed traffic in an urban corridor will be impacted by human-like C/AVs is the primary objective of this study. To do this, the HV-AV-CAV ratio for three traffic flows—peak hour traffic demand, 20% below peak hour traffic demand, and 20% over peak hour traffic demand—will be studied. For C/AV, three different driving behaviors—aggressive, normal, and safe—will be evaluated. The second objective is to look at how the parameters of the AVs' driving behavior connect with traffic performance. To determine which car-following parameter has the greatest influence, this will be researched and studied. Following the identification of influential car following parameters, sensitivity analysis for a chosen traffic performance indicator will be carried out on them.
Conclusion:
The lack of real or natural data can be compensated by the similarity of the driving behavior of HV and CAV. It can be a good data input for calibration and validation until we get natural data from CAVs in real traffic. If CAVs are assumed to behave like human-driven vehicles and this is evaluated against natural data, this could be a benchmark for future research. This will reduce the dependency on test bed data for the assessment of CAVs in mixed traffic to get more realistic influence in urban areas.
Along with this consideration in calibration and validation, this study also defined three driving behaviors and human alike car following method. From the analysis, it is clear that Introduction of CAV in higher number in mixed traffic could be beneficial if they have interaction with many other CAVs. The pattern starts changing from 50-70% of penetration. Below 50% penetration, the system works inefficiently because of lack of available connecting vehicles. After 70%, the system would work better if interacting vehicles are higher in number.
After analyzing eight parameters from driving behavior using a sensitivity analysis, it becomes obvious that some of them play an influential role in traffic performance. Minimum clearance, look back distance and number of interacting vehicles play dominating role than look ahead distance, minimum collision time gain, accepted deceleration, standstill distance and gap time distribution.
Inquiry 3
Please try to describe this paragraph more thoughtful: 1) what is given, 2) what are the challenges, 3) what this paper will address and 4) how (e.g. with sensitivity analysis)
Feedback:
Your this concern is addressed here in introduction and method section.
Even today, it is still not possible to obtain publicly available real traffic data from autonomous vehicles. This gap hinders the calibration and validation of the Connected Autonomous Vehicle (CAV), two important elements of microscopic traffic simulation. To overcome this gap, this study assumes that both CAVs and AVs would behave similarly to human-driven vehicles. There are traffic volume and speed from human driven vehicles which could be used for C/AVs as well from the assumption in similarity. The characteristics of driving behavior for connected and/or autonomous vehicles (C/AVs) need to be determined according to distinct driving behaviors, such as aggressive, normal, and safe (Sukennik 2018), reflecting three different human driving patterns makes it more human-like than other C/AVs. Vehicles will decide which parameters to follow from their permissible range in accordance with their regulating driving behavior s. Direct user selection of these driving behavior s is also possible through external connections, such as V2V, V2I, and emergency states. These behaviors exhibit wildly disparate responses to roads and traffic. A safe driving behavior is not permitted to drop the safe distance below the authorized limits, however an aggressive driving behavior may have a smaller safe distance with increased acceleration, resulting in C/AVs that resemble humans (Zeidler et al. 2018; Atkins 2016; Sukennik 2018).
The research findings show how traffic, and safety aspects are influenced by C/AVs that resemble humans. Additionally, it will use a sensitivity analysis platform to visualize how various AV driving behavior parameters interact with traffic performance. Speed, delay time, and travel time have all been used to examine traffic performance. In this process, only the interaction among motor vehicles is taken into account. For these types of studies, PTV VISSIM and Surrogate Safety Assessment Model (SSAM) must work together to identify the number of potential conflicts from the simulated data and vehicle trajectories to understand the safety implications of C/AVs. Examining how mixed traffic in an urban corridor will be impacted by human-like C/AVs is the primary objective of this study. To do this, the HV-AV-CAV ratio for three traffic flows—peak hour traffic demand, 20% below peak hour traffic demand, and 20% over peak hour traffic demand—will be studied. For C/AV, three different driving behaviors—aggressive, normal, and safe—will be evaluated. The second objective is to look at how the parameters of the AVs' driving behavior connect with traffic performance. To determine which car-following parameter has the greatest influence, this will be researched and studied. Following the identification of influential car following parameters, sensitivity analysis for a chosen traffic performance indicator will be carried out on them.
Inquiry 4
Section 2 is nice to have; however, it does not point out the necessity of this research work:
- What ‘can’ the state of the art (SOA)?
- Which aspects have been addressed by the SOA?
- Which not? (Why)
- How this research is novel compared to the SOA?
- What are the main benefits compared to SOA?
These aspects should be answered in the related work section!
Feedback:
Thank you for answering the exact idea of SOA. I have explained the process to deal with the issue of having no real world data for calibration and validation of CAVs. Please refer to introduction and method section which carry the idea of my methodology. There are very few works (only 2 work- PTV CoExist and Makridis et. al.) worked in similar approach so i do not have much to create a set of comparisons.
Inquiry 5
Figure 3 does not help a better understanding, for the reader, how the scenario looks like.
Feedback:
Figure 3 is not for scenario. It shows the study area and its different intersections in the PTV VISSIM interface. I am following other paper for presentation. In my case, the combination in experimental setup is still visible from visualization in the result section. I was successful to express it to other reviewers. The explanation of ratio of AV-CAV-HV will take more space and it will be still repeated in the result section.
Inquiry 6
Why does the word speed in Section 3.3.?.
Feedback:
The speed is a traffic performance indicator so it is also added in section 3.3 for explanation.
Inquiry 7
Figure 4 is hard to read, pixel image with bad quality. Please replace it. Result images: non-vector graphic images – hard to read
Feedback:
It is changed as per your suggestion. As i have obtained it from another research work, the quality is not in its best condition.
Inquiry 8
It is not clear whether the authors propose/show the benefits of CAV compared to AV or something else. It should be clearer in the related work section. The novelty of the paper should also be clearer in relation with the SOA.
The expression “higher penetration of CAV” is not specified in detail
Feedback:
Related work has stated the CAV has better performance than the AV and HV. The figure 1 shows the speed progress with higher penetration of CAV in the traffic. The triangular graphs show how % of any HV, AV or CAV influence the participation of each other.
Portion of related work is shown below:
Previous studies showed that higher penetration of CAV in the mixed traffic causes higher quality of traffic condition. It is to be expected that in the first stage of the implementation of C/AVs, the road capacity and thus the speed of vehicles will drop. With more of these kinds of vehicles on the road, the status of mobility, road safety and environmental conditions will improve due to various benefits of AV technologies such as platooning (Ahmed et al. 2022; Tettamanti et al. 2016). Figure 1 shows three ternary plots which depicts three different traffic demands, where the color level indicates the harmonic average speeds over the network. High traffic demand leads to a deterioration of traffic conditions as the number of interacting vehicles gradually increases and reaches technical limits (Makridis et al. 2018b; Mattas et al. 2018).
Higher penetration means more CAV comparatively to other vehicles.
Inquiry 9
The authors use not consistently what AV function and what CAV functions are, see “due to various benefits of AV technologies such as platooning”-> is here platooning AV or CAV
Feedback:
Platoon is for CAV. The 3.2 contains information about it.
3.2. Driving Behavior
Unlike the CAVs, the AVs do not act on platoon and do not slowing down in the traffic through connectivity. AVs are totally isolated and acts for its surrounding vehicles using its sensors.
Inquiry 10
Section 4.3. -> how can a function be calibrated? Tuned, parameter setting would suit better. What is the goal of the calibration? That will not clear from the text.
Feedback:
The function is not calibrated. SPSA is a calibration algorithm which takes decision to choose chosen parameter of car following behaviours. Please check
Qurashi, M. (2018): An Alternative Online Calibration approach for Dynamic Traffic Assignement Systems. Master. Technical University of Munich.
The calibration is used in microscopic transportation simulation to make it matchable to real world. We use real world data (also known as natural data) to compare it whatever we get as output from simulation. The matching is done in 95% confidential interval.
Inquiry 11
Penetration should be defined in the introduction -> the authors could state, this is the main objective of the study-> helps the reader
Feedback:
The penetration is not the main objective of this study. The main study is studying the C/AVs considering it as Human driven vehicles. We have defined the driving behaviours and had different penetration rate for HV, AV and CAV.
Inquiry 12
Fig 8 and Fig 10 -> Numbers are still cut down.
The terms in Fig 14 are not defined, so it is not clear for the reader what they exactly mean.
Feedback:
Figure 8 and 10 had format issue. Now, it is checked and is edited.
In the explanation of the figure 14, following description is added
The first value and the last value from each graphs (in total 8) from the figure 12 and 13 show how much variation has taken place in travel time. Figure 14 shows the percentage of variations between the initial and final value of individual driving parameters.
Inquiry 13
Simulation Frameworks/concepts of pedestrian-automated vehicle interactions.
- Simulation Frameworks/concepts of automated and human-driven vehicles interactions
o Pure simulation frameworks/concepts
o Frameworks with possible human interactions
- Studies with real vehicles: concepts of automated and human-driven vehicles interactions
Then please add some more preliminaries, in which the PTV Vissim is presented in more detail. Further potentials and limitation of Vissim should be discussed. Furthermore, the Wiedermann model should be presented shortly.
Some of the works in literature focus on human (pedestrian) machine (automated vehicles). These are also simulations frameworks, which could be adapted for the study presented in the work.” See, for more details, the earlier comment of the reviewer!
Feedback:
There is only one focus which is in the interaction in motorway. There is no separate plan for pedestrian, wheel chair.
So, Simulation Frameworks/concepts of automated and human-driven vehicles interactions are concern. The framework is defined in 3. Method. There is no other framework except this. Rest is mentioned in 4. Experimental setup.
After introducing HV, AV and CAV in the simulation environment with different ratio, they have been studied.
Again, i am following 2 research papers which are published. There is no limitation of PTV VISSIM which is connected in our research.
This is the details which we have seen other to use for discussing microscopic transportation simulation. We do not go to very deep. Even in the VISSIM manual, the defination is short and limitation is missing. The discussion regarding parameters are relatively bigger and that would be fundamental. When i was preparing the work, i had to follow reference of papers and go to help of VISSIM to learn and implement them.
The VISSIM, a microscopic traffic simulation software from PTV, has been used to perform the simulation. Moreover, SSAM have been used for additional inquires such as impact over accident. The parameters of car following model such as Wiedemann 74 and 99, are chosen in the VISSIM to represent the human driven vehicles for manoeuvre and lane behaviors.
Regarding the literature, i have used them connectedly. They were as per need. I am unable to extend it any longer.
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 AuthorsThe Authors find the following points of the reviewer:
2)
Feedback: It is clear in the introduction chapter that we are focusing on the study with account of interaction among motor vehicles; not interaction between pedestrians and motor vehicles. To keep the discussion brief in literature review, we skipped another perspective of simulation out of writing. This sentence identified the plan: In this process, only the interaction among motor vehicles is taken into account. As I am following a standard approach (not particularly innovative approach) of microscopic simulation, I focused more into latter chapters: method, experimental setup and result. As it is a microscopic simulation approach, which is clear from discussion and Figure 2, we stayed focused on the framework of microscopic simulation and relevant previous works. Other references will make it excessively lengthy and hence incomplete while we attempt to keep it brief. The references which were given in the review, I have taken time to understand them. They take another direction of methodology.
Reviewer:
In its current form, the related work is not suitable for a good journal. The reviewer suggested that the related works should include the "related works". Otherwise, it is only a "Technical overview".
3)
Feedback: In 3.1. Simulation planning and Assumptions i have discussed this matter. If i add more about it, it will become a lengthy discussion.
Reviewer:
Please, the authors have to manage to summarize in 2-3 sentences without becoming too lengthy. This is part of paper writing.
Feedback: 3.2 Performance indicators talk about the traffic and safety result after studying the impact of the implementation of connected autonomous vehicles. The penetration is part of experimental setup; not the performance indicator.
Reviewer:
The reviewer does not understand the answer. Please provide a reasoning for the use of the performance indicators
5)
Feedback: The calibration, connectivity code and simulation code have more than 1000 lines which are not straightforward to express in short discussion. I have changed the quality of the image and explained it.
Reviewer:
Please, try to summarize in a short manner. It is important for a good understanding of the work
8)
Feedback: These mentioned sections are written in separate sections organizedly. The goal of the study is discussed in the introduction and presentation of results are done in the result chapter. The conclusion recorded the findings.
Reviewer:
The answer of the Authors is not sufficient. They should make the structure much clearer and use better wording to obtain a better understanding. The placement of the images makes also it harder to follow this section.
Comments on the Quality of English Language
No comment
Reviewer 2 Report
Comments and Suggestions for AuthorsThe comments are not addressed properly.
What is CAV? not defined!!! How about C/AV? How about AV?
"This study is predicated on the idea .... " what does this mean?
The novelty of this work is not clear at all. Expecting to see the contributions emphasised.
How does the system model look like?
I believe the readability of the manuscript is very low. some sentences are not following each others. some paragraphs are not following each other.
what is the tool used for simulation studies?
the presentation of section 3.2 should be improved
Fig.2 seems to be the architecture of the method but why is it in simulation section?
I do not see pure analysis of the results. For example, "These additional strategies offered by C/AVs work better for peak hour demand and 20% below the peak hour than 20% above the peak hour demand case for the same study corridor." why?
The authors said "feedback answered" but would be great to discuss it further for more clarity. Use a standard rebuttal format is recommended.
Comments on the Quality of English Language
presentation of the paper need substantial improvement to increase the readability
Reviewer 3 Report
Comments and Suggestions for AuthorsDear authors!
I think that most of the recommendations from the previous review have been taken into account.
However, I recommend that Figure 11 and Figure 10 be shortened and put together, for example, there are no values greater than 314 travel time.
I recommend adding the technique of the microscopic traffic simulation. How to use microscopic traffic simulation?
Reviewer 4 Report
Comments and Suggestions for AuthorsThe topic is interesting and expected. As autonomous vehicles appear on the road, how the connected autonomous vehicle affects the traffic is crucial for traffic agents in traffic planning and control. However, there are still some issues that should be enhanced.
1. The novelty of this paper is unclear. It is suggested to highlight the novelty, contribution, and potential application of this paper. In addition, the research gap should be identified clearly.
2. It is suggested to tie this research to the literature on the related topic of this paper, such as the effects of AV in mixed traffic flow and the driving behavior of AV. Therefore, the related work section should be updated.
3. Safe and cautious for AV is easy to understand. However, aggressive as the driving module of AV should be explained in more detail.
4. The method to evaluate the effect of AV in traffic conditions is unclear. It is suggested to provide a flowchart to describe the procedure.
5. It seems that the title of this paper and the context and method described are out of line. It is suggested to refine the title of this paper.
6. It seems that the paper used VISSIM data to evaluate the effect of AV. It is suggested to use the natural data to analysis in the experiment section.
