Queue-Based Modeling of the Aircraft Arrival Process at a Single Airport
Round 1
Reviewer 1 Report
This paper studies an interesting problem in aerospace traffic management with the aim of reducing aircraft delays due to saturated airport traffic flows. The authors developed a mathematical model based on queue modelling techniques. A practical assessment of the model is proposed based on realistic data of Tokyo international airport. Overall, the paper is well-written and presents interesting results. However, the reviewer has serious concerns that the authors should address when revising their paper.
Introduction should clarify why queue-based modelling is needed to address the practical problem investigated in this paper, and the novelty of the mathematical modelling proposed in the paper.
The literature review should be extended by revising the following recent stream of papers on optimal scheduling and routing of aircraft in busy airports:
Samà, A. D’Ariano, K. Palagachev, M. Gerdts (2019) Integration methods for aircraft scheduling and trajectory optimization at a busy terminal manoeuvring area, OR Spectrum, https://doi.org/10.1007/s00291-019-00560-1
Samà, A. D’Ariano, F. Corman, D. Pacciarelli (2018) Coordination of scheduling decisions in the management of airport airspace and taxiway operations, Transportation Research, Part A, 114(B) 398–411
Samà, A. D’Ariano, F. Corman, D. Pacciarelli (2017) Metaheuristics for efficient aircraft scheduling and re-routing at busy terminal control areas, Transportation Research, Part C, 80 (1) 485–511
D’Ariano, D. Pacciarelli, M. Pistelli and M. Pranzo (2015) Real-time scheduling of aircraft arrivals and departures in a terminal maneuvering area. Networks, 65(3) 212–227
Please provide a discussion of the novelty of your paper compared to the state-of-the-art in the field.
The study of different levels of automation in modelling the minimum aircraft separation technology is very nice and has a meaningful practical value.
Please provide a motivation of the assumptions made in your paper to model the queuing model.
The computational results on the different levels of automation are very interesting. Maybe further research can be proposed to investigate the combination of different levels of automation with optimal aircraft scheduling and dynamic trajectory management techniques.
Author Response
Comment: Introduction should clarify why queue-based modelling is needed to address the practical problem investigated in this paper, and the novelty of the mathematical modelling proposed in the paper.
Response: Thank you for pointing this out. We have added the following arguments in the Introduction section, lines 36-37 on page 1 and lines 56-62 on page 2. Currently, the Tokyo International Airport is considering re-designing the structure of the airspace sectors. In doing so, one of the challenges is to evaluate potential capacity bottlenecks and delays. Our proposed queuing approach provides closed-form analytical results about flight delay bottlenecks in the airspace area around the airport.
In literature, several queuing models have been used to study air transport operations, such as M/M/1 and M/Ek/1 queues. These models assume that arrivals and/or service times follow exponential distributions. However, from the data analysis of flight arrivals at Tokyo International Airport, these exponentially assumptions do not hold. Also, to model the fact that several flights can be present at the same time in an airspace area, we assume a queue with several servers. As such, we consider a G/G/c queuing model that fits best the available traffic data. As far as we know, G/G/c queuing models have not been used before to analyze air traffic data.
Comment: The literature review should be extended by revising the following recent stream of papers on optimal scheduling and routing of aircraft in busy airports:
Samà, A. D’Ariano, K. Palagachev, M. Gerdts (2019) Integration methods for aircraft scheduling and trajectory optimization at a busy terminal manoeuvring area, OR Spectrum, https://doi.org/10.1007/s00291-019-00560-1
Samà, A. D’Ariano, F. Corman, D. Pacciarelli (2018) Coordination of scheduling decisions in the management of airport airspace and taxiway operations, Transportation Research, Part A, 114(B) 398–411
Samà, A. D’Ariano, F. Corman, D. Pacciarelli (2017) Metaheuristics for efficient aircraft scheduling and re-routing at busy terminal control areas, Transportation Research, Part C, 80 (1) 485–511
D’Ariano, D. Pacciarelli, M. Pistelli and M. Pranzo (2015) Real-time scheduling of aircraft arrivals and departures in a terminal maneuvering area. Networks, 65(3) 212–227
Response: Thank you for introducing insightful works. We read through the suggested references and referred the selected one depending on the context in this paper (Samà, A. D’Ariano, F. Corman, D. Pacciarelli (2017)) in 1.Introduction section, line 27-29 on page 1.
Comment: Please provide a discussion of the novelty of your paper compared to the state-of-the-art in the field.
Response: The novelty of this paper is provided in 6.Discussion section, lines 302-312 on page 16, comparing to the state-of-the-art in the field. Conventionally, as shown in section 4.2, ICAO and FAA have been discussing global standards on the interval management, which harmonizes aircraft self-separation and advanced AMAN, for increasing runway throughputs. Increasing runway throughputs may bring traffic jams in arrival traffic flow, but these impacts have not been discussed. Contribution of this paper is two-hold: 1) this paper provided a quantitative approach which estimates bottlenecks and arrival delay time in the traffic flow as impacts of increasing runway throughput. 2) Feasible solutions were clarified to achieve increasing arrival rate while reducing arrival delay time at a case study airport. These results support on-going process of developing global standards on the automation system design in air transport at ICAO AIRB (Airborne Surveillance) working group. They have started discussion on the global standards of advanced AMAN on the ground. Our future works will develop the proposed data-driven queuing approach and apply the latest air traffic data in 2019 (because new airspace and air route configurations are applied to Tokyo International Airport in July 2019).
Comment: The study of different levels of automation in modelling the minimum aircraft separation technology is very nice and has a meaningful practical value.
Response: Thank you for your constructive feedback. Your comments contributed to improving our discussion in 6.Discussion section, lines 302-312 on page 16.
Comment: Please provide a motivation of the assumptions made in your paper to model the queuing model.
Response: Thank you for pointing this out. To model the traffic in a given airspace area, we assumed a G/G/c model. The reason for doing so are now explained in detail in Section 3.1, lines 147-151 on page 5. By analyzing the historical traffic data, the arrival and service times could not be fitted to specific distributions such as exponential, normal, etc. Thus, we assumed general, data-based distributions to characterize the arrival process of aircraft at an airspace and the time to fly in a given airspace. Also, to model the fact that several aircraft can be present at the same time in a given airspace, we assumed a multi-server queue.
Comment: The computational results on the different levels of automation are very interesting. Maybe further research can be proposed to investigate the combination of different levels of automation with optimal aircraft scheduling and dynamic trajectory management techniques.
Response: Thank you for your insightful comments. Unfortunately, optimal aircraft scheduling and dynamic trajectory management are not currently in the scope of our project (see Acknowledgement), but certainly very interesting topics to investigate as combinations with different levels of automation in the future.
Reviewer 2 Report
Generic comments and recommendation
In the literature review (1. Introduction), several uses of queuing models for arrival metering are mentioned, however, advantages and disadvantages of each model and application are not mentioned. That would be useful to demonstrate the research motivation and methods proposed in the paper.
In Section 2.2., the authors state that data from 71 days were randomly selected from the odd months of 2016 and 2017. It is not clear whether the data were prone to:
Seasonality: airlines schedules and total traffic volumes may vary significantly depending on winter or summer season The effect of weather conditions: VFR and IFR conditions, wind conditions that govern the active airport runway configuration and thunderstorms may significantly affect operations and the total operations.
Not accounting for the above effects may have a significant impact on means and any other statistics. While Gaussian distributions used later in the paper may account for such events as outliers, the complex interrelationships between such events may have an impact in the analysis. I believe that it would be preferred to select days of similar weather conditions (VFR vs IFR) not affected by extreme weather, and also for a specific time of year (summer, winter). Time of day may also be of relevance, as potential departure or arrival push strategies in place may lead to very different total arrival numbers during specific periods of day. The authors should elaborate more on how their statistical approach accounts (or not) for the above effects.
The authors do not explain the rationale behind selecting a G / G / c queuing model as opposed to other models. Again, this could be justified by describing models used in existing literature, their advantages and disadvantages.
Specific comments
Page 4. Line 101: “an average of 569 flights arrival”, should be “an average of 569 arrival flights”
Figure 5: Is this for all the days or for the selected day?
Page 6, Line 130: “with a different in the radius of 10NM” please correct grammar.
Page 6, Lines 126-132 and Lines 133-136: information is repeated, please revise.
Page 8, Line 174: The capacity constraint c is modeled as a constant. However, in real operations airspace sector capacity may vary. Please clarify assumptions.
Page 9, Line 186: The letter c is used for servers. The same letter c was used before to refer to capacity of each sector. This can be confusing, please revise notation.
Page 11, Line 211: “…which is exactly the minimum aircraft separation currently needed for radar separation in the en-route airspace”: the radar separation needed may vary depending on IFR / VFR conditions and aircraft sequence (aircraft type). Moreover, the model considers airspace adjacent to the airport and typically within 40-50NM from the airport, this is terminal airspace. Please elaborate more on these assumptions.
Page 11, Line 214: “This number will be increased in the future to accommodate an increasing runway throughput” please rephrase to “This number can be adjusted to account for increased runway throughput”
Page 12, Line 216: This is in contradiction with the constant 5NM separation mentioned before. What separation standards were ultimately applied: 5NM or the RECAT? If RECAT, please elaborate on specific separation standards considered.
Page 12, Line 220: What do GIM and IM stand for?
Page 15, Line 249: This really depends on aircraft sequence, e.g. small before Large, Heavy before Heavy, etc. Applying the RECAT, meaning aircraft wake categories A-F would still have equivalent implications regarding minimum separation and aircraft arrival sequence. Please elaborate more on this.
Page 15, Line 276: “We leave fully accurate” please rephrase.
Author Response
In the literature review (1. Introduction), several uses of queuing models for arrival metering are mentioned, however, advantages and disadvantages of each model and application are not mentioned. That would be useful to demonstrate the research motivation and methods proposed in the paper.
Response: Thank you for your feedback. We have extended the discussion in the Introduction section lines 56-62 on page 2, the advantages and disadvantages of different models and applications. Overall, several queuing models have been used to study air transport operations, such as M/M/1 and M/Ek/1 queues. These models assume that arrivals and/or service times follow exponential distributions. However, from the data analysis of flight arrivals at Tokyo International Airport, these exponentially assumptions do not hold. Also, to model the fact that several flights can be present at the same time in an airspace area, we assume a queue with several servers. As such, we consider a G/G/c queuing model that fits best the available traffic data.
In Section 2.2., the authors state that data from 71 days were randomly selected from the odd months of 2016 and 2017. It is not clear whether the data were prone to:
Seasonality: airlines schedules and total traffic volumes may vary significantly depending on winter or summer season The effect of weather conditions: VFR and IFR conditions, wind conditions that govern the active airport runway configuration and thunderstorms may significantly affect operations and the total operations.
Not accounting for the above effects may have a significant impact on means and any other statistics. While Gaussian distributions used later in the paper may account for such events as outliers, the complex interrelationships between such events may have an impact in the analysis. I believe that it would be preferred to select days of similar weather conditions (VFR vs IFR) not affected by extreme weather, and also for a specific time of year (summer, winter). Time of day may also be of relevance, as potential departure or arrival push strategies in place may lead to very different total arrival numbers during specific periods of day. The authors should elaborate more on how their statistical approach accounts (or not) for the above effects.
Response: Thank you for your feedbacks. All data were provided by JCAB (Japan Civil Aviation Bureau) restricted for only research use under conditions of discloses specific dates in any publications. Although we cannot reveal the dates, we elaborate data characteristics in section 2.2, lines 98-99 on page 3. All data corresponds to nominal conditions at Tokyo International Airport under IFR conditions excluding weather impacts and other rare events. Airlines schedules (domestic flights are the majority at the airport) and total traffic volumes are not significantly affected by seasonal reasons at Tokyo International Airport. North wind operation is the majority because they keep the operation until exceeding 10knot tail wind. Aircraft flight time is changed by wind velocity; however, it was difficult to distinguish operations in Japan FIR clearly in summer and winter seasons by a specific time of the year in our limited data. Future study will develop the analysis using larger amount of data referring this paper. As described in section 3, line 128 on page 5, 5:00PA to 10:00PM time period was selected in the data-driven modeling approach.
The authors do not explain the rationale behind selecting a G / G / c queuing model as opposed to other models. Again, this could be justified by describing models used in existing literature, their advantages and disadvantages.
Response: Thank you for pointing this out. To model the traffic in a given airspace area, we assumed a G/G/c model. The reason for doing so is explained now in detail in section Introduction, lines 56-62 on page 2. By analyzing the historical traffic data, the arrival and service times could not be fitted to specific distributions such as exponential, normal, etc. Thus, we assumed general, data-based distributions to characterize the arrival process of aircraft at an airspace and the time to fly a given airspace. Also, to model the fact that several aircraft can be present at the same time in a given airspace, we assumed a multi-server queue.
Specific comments
Page 4. Line 101: “an average of 569 flights arrival”, should be “an average of 569 arrival flights”
Response: Thank you for pointing this out. We have corrected this in the current manuscript.
Figure 5: Is this for all the days or for the selected day?
Response: It is for all the days (71 days). We have elaborated this in section 2.2 line 123 on page 4.
Page 6, Line 130: “with a different in the radius of 10NM” please correct grammar.
Response: Thank you for pointing this out. We have corrected this in the current manuscript.
Page 6, Lines 126-132 and Lines 133-136: information is repeated, please revise.
Response: Thank you for pointing this out. We have revised this in the current manuscript.
Page 8, Line 174: The capacity constraint c is modeled as a constant. However, in real operations airspace sector capacity may vary. Please clarify assumptions.
Response: Thank you for your feedbacks. The reason why this paper models the capacity constraint as a constant (c=2) is elaborated in section 4.2 lines 223-230 on page 12 and 13. As shown in Fig.11, the number of servers c, which we define as a capacity in the assigned airspace i, was estimated based on data-driven analysis using all 71 days’ radar data. As a result, the median and mean in the assigned airspace i in the terminal area, approximately within 50NM where i<5, takes c=2. The median and mean in the en-routes takes c=1 or 0. The reason why this paper takes c=2 not only for i<5 airspace but also for i>4 airspace is explained in lines 227-230. Currently en-route radar separation is 5NM, thus we assume that c=2 is relevant in the assigned 10NM area from the perspective of safety in air traffic control also in en-route air spaces.
Page 9, Line 186: The letter c is used for servers. The same letter c was used before to refer to capacity of each sector. This can be confusing, please revise notation.
Response: According to your suggestion, the notation was revised in section 3.2, lines 190 and 191 on page 7.
Page 11, Line 211: “…which is exactly the minimum aircraft separation currently needed for radar separation in the en-route airspace”: the radar separation needed may vary depending on IFR / VFR conditions and aircraft sequence (aircraft type). Moreover, the model considers airspace adjacent to the airport and typically within 40-50NM from the airport, this is terminal airspace. Please elaborate more on these assumptions.
Response: Thank you for pointing this out. We elaborated the assumptions more in section 4.2 lines 223-230 on page 12 and 13.
Page 11, Line 214: “This number will be increased in the future to accommodate an increasing runway throughput” please rephrase to “This number can be adjusted to account for increased runway throughput”
Response: Thank you for pointing this out. We have corrected this in the current manuscript.
Page 12, Line 216: This is in contradiction with the constant 5NM separation mentioned before. What separation standards were ultimately applied: 5NM or the RECAT? If RECAT, please elaborate on specific separation standards considered.
Response: The current operation at Tokyo International Airport has not applied RECAT yet. They take conservative operations: average 5NM separations for all arrival aircraft in the terminal area and at runway threshold (see Figure 11). We have elaborated this in section 4.2 line 235 on page 14.
Page 12, Line 220: What do GIM and IM stand for?
Response: GIM stands for “Ground-based Interval Management”, and IM stands for “Interval Management”. We have inserted both words in section 4.2 lines 240 and 241 on page 15.
Page 15, Line 249: This really depends on aircraft sequence, e.g. small before Large, Heavy before Heavy, etc. Applying the RECAT, meaning aircraft wake categories A-F would still have equivalent implications regarding minimum separation and aircraft arrival sequence. Please elaborate more on this.
Response: Thank you for pointing this out. We have elaborated this in section 5 lines 269-271 on page 15.
Page 15, Line 276: “We leave fully accurate” please rephrase.
Response: Thank you for pointing this out. We have corrected this in the current manuscript.
Round 2
Reviewer 1 Report
The revised paper addresses my major concerns satisfactory. I support the paper publication.
