Flight and passenger efficiency-fairness trade-off for ATFM delay assignment

https://doi.org/10.1016/j.jairtraman.2019.101758Get rights and content

Highlights

  • Pre-tactical optimisation of performance and equity using flight and passenger metrics.

  • Fairness metrics defined from flight and from passenger perspectives.

  • Definition and analysis of efficiency-fairness trade-offs.

  • Possible improvements for one stakeholder without significant reduction of the other.

Abstract

This paper studies trade-offs between efficiency (performance) and fairness (equity), when assigning ATFM delay pre-tactically (on-ground at origin airport) due to reduced airport capacity at destination. Delay is assigned as the result of the optimisation of a deterministic multi-objective problem considering flight and passenger perspectives when defining objectives of performance and fairness. Two optimisation cases are presented: one where objectives are based on flight metrics, and another one where they are based on passenger metrics. The paper defines and analyses efficiency-fairness trade-offs: the concepts of price of fairness for flights and passengers are defined as the percentage of efficiency loss due to the consideration in the optimisation of fairness; whereas the price of efficiency is considered as the fairness loss relative to the maximum value of the fairness metric, when considering flight or passenger delay in the optimisation. The optimisation model is based on the ground holding problem and uses various objective functions. For performance, total delay for flights (considering reactionary delay), and total delay for passengers (considering outbound connections) are defined. For fairness, the deviation of flight arrivals from a Ration By Schedule solution, and the deviation of delay experienced by passengers with respect to the one obtained in an RBS situation are used. An illustrative application on traffic at Paris Charles de Gaulle airport, a busy European hub airport, and including realistic values of traffic is modelled. A comprehensive trade-off analysis is presented. Results show, how in some cases, gains on one stakeholder can be achieved without implying any detriment on the other one. Passengers are more sensitive to the optimisation and hence, their consideration when assigning delay is recommended. Further research should explore how to combine flight and passenger indicators in the optimisation and consider how the lack of data availability could be mitigated.

Introduction

Airports are limited in capacity by operational constraints (Bazargan et al., 2002; Gilbo, 1993), generating in some cases, a significant imbalance between capacity and demand. Air Traffic Flow and Capacity Management (ATFCM) initiatives are then implemented to smooth traffic arrivals, transferring costly airborne delay, carried out with holdings and/or path stretching, to pre-departure on-ground delay (Carlier et al., 2007).

When a capacity-demand imbalance is detected, a Ration by Schedule (RBS) prioritisation of flights is the current practice for slot assignment (EUROCONTROL, 2015a). RBS policy is considered by the different stakeholders to be the fairest delay assignment, since it respects the original slot planning, but economical optimum cannot be guaranteed and only arrival delay is considered. Nevertheless, while airline reported delay was 14.7 min per flight in Europe in 2018, reactionary delay was responsible for 45% of it, heavily influenced by the impact of first-rotation delays (EUROCONTROL, 2019), and should thus not be neglected.

In the current operational environment, the system is optimised considering a flight-centred perspective, however, different stakeholders might experience ATM system performances differently. In particular, passenger-centric metrics might differ from their equivalent flight-centric ones (Cook et al., 2012). In (Montlaur and Delgado, 2017), performances for flight and passenger delays of an extended arrival manager (E-AMAN) were analysed in conjunction with a pre-tactical optimisation of flights. In that work, the assignment of slots was optimised considering either arrival delay for flights, arrival delay for passengers, total delay for flights (considering reactionary delay) or total delay for passengers (considering outbound connections). Results showed that in the scope of an E-AMAN, the distances and possible delays that can be assigned do not justify the application of a more sophisticated strategy than RBS. Nevertheless, when the scope of optimisation was enlarged to include the pre-tactical phase, benefits (and trade-offs) were obtained by optimising the assignment of delay instead of only considering flight schedules. While minimising the total delay for passengers is, as expected, the best strategy from the passengers perspective, it leads to higher reactionary delay for flights with respect to a flight-centric optimisation. Though optimisation carried out in (Montlaur and Delgado, 2017) focused on only one stakeholder at a time and did not include an explicit consideration of equity, it was shown that if focus is given to flight total delay, the benefit per passenger remains similar to the passenger centred optimisation and the variability with respect to the RBS delay assignment was reduced, improving the fairness of the solution.

Equity from a flight perspective was explicitly included in (Montlaur and Delgado, 2018), where flight and passenger delays and equity were conjointly considered in the optimisation scheme. Preliminary conclusions showed how focusing on one objective would affect the two others. The work presented in (Montlaur and Delgado, 2018) focused on the optimisation of three metrics (flight and passenger delays and flight-fairness) at the same time, allowing a better understanding of the trade-offs between performance from a flight delay, passenger delay and flight fairness perspective. The work focused on allowing decision makers to consider informed a posteriori articulation of preferences when optimising the delay assignment. To that end, the concept of price of performance (understood as price of trade-off) and price of efficiency as the ones defined in (Bertsimas et al., 2012) were used.

Previous research, focusing on three variables at the same time, lacked some applicability and a definition of fairness for passengers was missing. This paper has three main goals: 1) It considers a classical efficiency-equity trade-off at flight level, which would use easily available data. 2) It then studies the impact on passengers metrics. 3) It finally aims at covering the gap of equity for passengers by suggesting a definition of passenger delay fairness and finally analyses the trade-offs existing between flight and passenger performance and fairness metrics. Note that, during real life operations, the impact of uncertainty would be very relevant as the planned optimisation performed pre-tactically is subject to degradation when tactically realised. This might impact the trade-off between indicators and affect the optimal assignment of slots focusing on the expected outcome, as done for example in (Glover and Ball, 2013). In this paper, however, the assignment of slots is optimised in a deterministic manner since the main objective consists in defining and presenting different metrics. Future work should research the impact of the system's stochasticity on the stability of the solutions.

Section 2 details the background on fairness on delay assignment with an analysis of literature when allocating resources from a fairness point of view. Section 3 first explains the optimisation model and objectives used in this work. It then considers how individual objectives are combined in a multi-objective problem and presents the optimisation cases analysed in the paper. Finally, the consideration of efficiency and fairness trade-off with the concepts of price of fairness and price of efficiency are described. Section 4 recalls the main hypothesis of the simulation of traffic that has been used as an example in this paper. The main results (for the multi-objective optimisation and for the performance-fairness trade-offs) and their discussion are reported in Section 5. The paper finalises with the conclusions and further work found in Section 6.

Section snippets

Background: fairness on delay assignment

When assigning delay to flights due to capacity-demand imbalances, even though commonly used, RBS is not the only possibility. Extensive research has been conducted to assign the required delay in a most cost-effective manner (Ball et al., 2007; Gilbo, 1993; Dell’Olmo and Lulli, 2003; Vranas et al., 1994; Montlaur and Delgado, 2017). As described in (Bertsimas et al., 2012), this type of resource allocation problems may be viewed as a utility allocation among different parties, which will lead

Delay and fairness optimisation models

This section introduces the mathematical model used in this paper for the optimisation and for the analysis of the trade-off between efficiency (delay generated in the system) and fairness (based on the delay experienced by stakeholders). A description of price of fairness and price of efficiency is provided as a tool to understand these trade-offs. The different objective functions considered are explored in detail with particular focus on the definition of metrics (and objectives functions)

Problem description

In order to test the previously defined metrics and to provide a testing environment for trade-offs between performance and fairness for flights and passengers, a specific traffic scenario is modelled. The arrival demand at Paris Charles de Gaulle (CDG) airport on September 12th, 2014 has been considered for all simulations. It was a busy Friday without any major disruption. Morning traffic, between 5.00 GMT and 11.00 GMT, is analysed. For the traffic scheduled, data from EUROCONTROL Demand

Results

We calculate the Pareto optimal points needed for computation of the Pareto extreme points required in Equations (11) and (12). These optimal points (lightly shaded boxes in Table 2) are defined as the points minimising each objective, that is, α=1 for ObjFair, α=0 for ObjPerf. The darker boxes indicate the maximum of each function. Note that, as shown in Table 2, optimal and maximum values are calculated for the chosen optimisation case (Case 1 for flight-based, Case 2 for passenger-based).

Conclusions and further work

In this paper, arrival delay due to ATFM regulations has been assigned as the optimisation of a multi-objective problem considering system performance (total delay) and fairness (deviation with respect to RBS solution). This has been done considering the perspective of flights and passengers. Two optimisation cases have been discussed: one where flights metrics are considered and another one focused on passengers indicators. A new definition for passenger fair assignment of delay has been

Acknowledgment

This work has been partly financed thanks to the Agència de Gestió d'Ajuts Universitaris i de Recerca - Generalitat de Catalunya (Grant number 2017-SGR-1278).

References (30)

  • D. Bertsimas et al.

    Fairness and collaboration in network air traffic flow management: an optimization approach

    Transp. Sci.

    (2016)
  • S. Carlier et al.

    Environmental impact of air traffic flow management delays

  • A. Cook et al.

    Passenger-oriented enhanced metrics

  • L. Delgado et al.

    Quantifying resilience in ATM, contrasting the impacts of four mechanism during disturbance

  • P. Dell'Olmo et al.

    A dynamic programming approach for the airport capacity allocation problem

    IMA J. Manag. Math.

    (2003)
  • Cited by (0)

    View full text