Periodic Reporting for period 4 - ARCTIC (Air Transport as Information and Computation)
Período documentado: 2024-09-01 hasta 2025-08-31
Air transport has mostly been studied as a transportation process, in which different elements, e.g. aircraft or passengers, move within the system. While intuitive, this approach entails several drawbacks, including the need for large-scale simulations and the difficulty of extracting macro-scale conclusions. The lack of a better approach is in part responsible for our inability to fully understand delay propagation, one of the most important phenomena in air transport.
Delays are unavoidable consequences of the limitedness of resources in air transport, and also of the need to ensure the highest level of safety and security. To illustrate, a flight landing can be delayed because of saturated airspaces, and of the need of maintaining a safe distance between aircraft. If delays ought to be accepted, a different problem is their propagation, i.e. when the late arrival of one aircraft results in the delay of several other flights. These, also known as secondary or knock-off delays, account for about half of the delays observed in the European system, and have huge social and economic impact.
ARCTIC proposes to change the conceptual framework used to analyse air transport, inspired by the way the brain is studied in neuroscience. It is based on understanding air transport as an information processing system. Airports then become computational units, receiving information from their neighbours through inbound flights under the form of delays; processing it in a potentially non-linear way; and redistributing the result to the system as outbound delays. Such computation can be made explicit by using a combination of information sciences and statistical physics techniques: from the detection of information movements through causality metrics, up to the representation of the resulting transfer structures through complex networks. The approach also entails important challenges, like the translation of the obtained insights into implementable policies.
ARCTIC aims at developing the first model of air transport as an information processing system, by leveraging on existing neuroscience and statistical physics techniques. Objectives include: to understand, from a theoretical perspective, what it means to describe the phenomenon of delay propagation as information processing; to characterise the real delay propagation phenomenon, under both normal and abnormal conditions; and to propose strategies to optimise the dynamics of the system.
Delays are unavoidable consequences of the limitedness of resources in air transport, and also of the need to ensure the highest level of safety and security. To illustrate, a flight landing can be delayed because of saturated airspaces, and of the need of maintaining a safe distance between aircraft. If delays ought to be accepted, a different problem is their propagation, i.e. when the late arrival of one aircraft results in the delay of several other flights. These, also known as secondary or knock-off delays, account for about half of the delays observed in the European system, and have huge social and economic impact.
ARCTIC proposes to change the conceptual framework used to analyse air transport, inspired by the way the brain is studied in neuroscience. It is based on understanding air transport as an information processing system. Airports then become computational units, receiving information from their neighbours through inbound flights under the form of delays; processing it in a potentially non-linear way; and redistributing the result to the system as outbound delays. Such computation can be made explicit by using a combination of information sciences and statistical physics techniques: from the detection of information movements through causality metrics, up to the representation of the resulting transfer structures through complex networks. The approach also entails important challenges, like the translation of the obtained insights into implementable policies.
ARCTIC aims at developing the first model of air transport as an information processing system, by leveraging on existing neuroscience and statistical physics techniques. Objectives include: to understand, from a theoretical perspective, what it means to describe the phenomenon of delay propagation as information processing; to characterise the real delay propagation phenomenon, under both normal and abnormal conditions; and to propose strategies to optimise the dynamics of the system.
Throughout the project, the team has focused on two complementary topics.
On the one hand, an important effort has been devoted to lay down the foundation of the analysis of air transport from an information processing viewpoint. This included, among others: the analysis of the performance of existing causality and information theory metrics, under different conditions and in real-world data; the validation of these metrics, using synthetic models of air transport; the pre-processing of real delay data, to minimise confounding effects; the development of new techniques for reconstructing networks of delay propagations, based on Deep Learning and other dynamical models; and the simplification of obtained networks, in order to highlight their most salient characteristics. In order to foster the use of these ideas, several open-source libraries have been made available to the community, covering topics like the calculation of information theory metrics, the creation of synthetic time series of delays, and a complete pipeline for the analysis of delay propagation.
On the other hand, these theoretical and computational tools have been deployed to the analysis of multiple air transport systems – chiefly the European one, but also considering US, China, and Africa. Results have highlighted relevant facts about the propagation of delays, both on average and through time; the efficiency of individual airports; the emergence of local patterns of delay propagation centred at specific airports, and how these are consequence of the presence of different types of airlines; and the evolution of the role of airports throughout the day. These have led to the identification of key airports in Europe, i.e. those mostly responsible for the propagation of delays; the quantification of the costs associated to such inefficiencies; and the proposal of operational solutions to mitigate them. Additionally, the project leveraged the widespread impacts of the COVID-19 pandemic, as an instance of a profound and long-lasting perturbation that has disrupted the normal dynamics of the system. Analyses focused on how different systems and airports therein have recovered, the specificities of such recovery, and how changes in dynamics have lingered even years after the peak of the disruption.
On the one hand, an important effort has been devoted to lay down the foundation of the analysis of air transport from an information processing viewpoint. This included, among others: the analysis of the performance of existing causality and information theory metrics, under different conditions and in real-world data; the validation of these metrics, using synthetic models of air transport; the pre-processing of real delay data, to minimise confounding effects; the development of new techniques for reconstructing networks of delay propagations, based on Deep Learning and other dynamical models; and the simplification of obtained networks, in order to highlight their most salient characteristics. In order to foster the use of these ideas, several open-source libraries have been made available to the community, covering topics like the calculation of information theory metrics, the creation of synthetic time series of delays, and a complete pipeline for the analysis of delay propagation.
On the other hand, these theoretical and computational tools have been deployed to the analysis of multiple air transport systems – chiefly the European one, but also considering US, China, and Africa. Results have highlighted relevant facts about the propagation of delays, both on average and through time; the efficiency of individual airports; the emergence of local patterns of delay propagation centred at specific airports, and how these are consequence of the presence of different types of airlines; and the evolution of the role of airports throughout the day. These have led to the identification of key airports in Europe, i.e. those mostly responsible for the propagation of delays; the quantification of the costs associated to such inefficiencies; and the proposal of operational solutions to mitigate them. Additionally, the project leveraged the widespread impacts of the COVID-19 pandemic, as an instance of a profound and long-lasting perturbation that has disrupted the normal dynamics of the system. Analyses focused on how different systems and airports therein have recovered, the specificities of such recovery, and how changes in dynamics have lingered even years after the peak of the disruption.
ARCTIC has provided unique benefits to European science and society, especially in the context of air transport. On the one hand, we have provided quantitative tools for assessing and tackling the propagation of delays, and for identifying which airports are most responsible for such propagation. This can be at the foundation of more informed decisions regarding the introduction of new policies, with potentially large benefits for this economic sector, for society at large, and with a reduction of the environmental impact of air transport. This has been complemented by making available multiple software libraries, aimed both at fostering reproducibility and at supporting future studies on similar topics.
From a more theoretical perspective, ARCTIC has been the first systematic attempt at importing neuroscience, statistical physics and information theory techniques into air transport. While some research programmes stressed the urgency of importing and adapting ideas from other research fields, much progress is still required, due to the complexity of such adaptation. Notably, the air transport community is starting to accept the use of these concepts, beyond the standard solutions based on micro-scale simulations. This entailed overcoming barriers, as the use of statistical physics techniques was unthinkable only a few years ago. While the path is a long one, this project has made a significant progress in the correct direction.
From a more theoretical perspective, ARCTIC has been the first systematic attempt at importing neuroscience, statistical physics and information theory techniques into air transport. While some research programmes stressed the urgency of importing and adapting ideas from other research fields, much progress is still required, due to the complexity of such adaptation. Notably, the air transport community is starting to accept the use of these concepts, beyond the standard solutions based on micro-scale simulations. This entailed overcoming barriers, as the use of statistical physics techniques was unthinkable only a few years ago. While the path is a long one, this project has made a significant progress in the correct direction.