Community Detection for Air Traffic Networks and Its Application in Strategic Flight Planning

Silvia Zaoli,Lorenzo Castelli,Giovanni Scaini

doi:10.3390/su13168924

Abstract

An environmentally and economically sustainable air traffic management system must rely on fast models to assess and compare various alternatives and decisions at the different flight planning levels. Due to the numerous interactions between flights, mathematical models to manage the traffic can be computationally time-consuming when considering a large number of flights to be optimised at the same time. Focusing on demand–capacity imbalances, this paper proposes an approach that permits to quickly obtain an approximate but acceptable solution of this problem. The approach consists in partitioning flights into subgroups that influence each other only weakly, solving the problem independently in each subgroup, and then aggregating the solutions. The core of the approach is a method to build a network representing the interactions among flights, and several options for the definition of an interaction are tested. The network is then partitioned with existing community detection algorithms. The results show that applying a strategic flight planning optimisation algorithm on each subgroup independently reduces significantly the computational time with respect to its application on the entire European air traffic network, at the cost of few and small violations of sector capacity constraints, much smaller than those actually observed on the day of operations.

Highlights

We propose and compare different ways of defining flight interactions, which lead to different networks representing one day of flights across Europe; we show that it is possible to divide the set of all European flights of one entire day into components that are temporally and/or spatially distinct, such that different components interact weakly; we verify, for a relevant problem of strategic flight planning optimisation, that solving the problem on smaller components and aggregating the solutions can be significantly faster than solving the whole network, at the only expense of a slight deterioration of the optimal solution
We apply this approach to the set of flights, aiming to identify subsets of flights that interact as little as possible with other subgroups, where interactions are defined within the context of the specific problem at hand
We partition the set of flights by applying community detection algorithms

Summary

Introduction

The air transport system is formed of many highly interconnected components. In such a complex system, delays and inefficiencies arise for many different reasons, such as tight aircraft and crew rotations or imbalances between demand and capacity of the airspace. The constant growth in air traffic in recent years has exacerbated the situation, making the efficient management of air traffic a necessary and challenging endeavour. The COVID-19 pandemic has dramatically interrupted this growth (at the time of writing, in Europe, the traffic level is reduced by −65% with respect to 2019), a recovery is expected (+3% in 2024 with respect to 2019, according to some optimistic scenarios [1]). The inefficiencies and problems of the past will certainly recur if not adequately addressed

Objectives

Methods

Results

Conclusion