Complexity Science in Air Traffic Management

Andrew Cook, Damián Rivas, Andrew Cook, Damián Rivas

• Publisher: Routledge
• Year: 2016
• ISBN:
  9781317162728
• Edition: 1
• Topic: Tecnologia e ingegneria
• Subtopic: Ingegneria aerospaziale

Air traffic management (ATM) comprises a highly complex socio-technical system that keeps air traffic flowing safely and efficiently, worldwide, every minute of the year. Over the last few decades, several ambitious ATM performance improvement programmes have been undertaken. Such programmes have mostly delivered local technological solutions, whilst corresponding ATM performance improvements have fallen short of stakeholder expectations. In hindsight, this can be substantially explained from a complexity science perspective: ATM is simply too complex to address through classical approaches such as system engineering and human factors. In order to change this, complexity science has to be embraced as ATM's 'best friend'. The applicability of complexity science paradigms to the analysis and modelling of future operations is driven by the need to accommodate long-term air traffic growth within an already-saturated ATM infrastructure.

Complexity Science in Air Traffic Management is written particularly, but not exclusively, for transport researchers, though it also has a complementary appeal to practitioners, supported through the frequent references made to practical examples and operational themes such as performance, airline strategy, passenger mobility, delay propagation and free-flight safety. The book should also have significant appeal beyond the transport domain, due to its intrinsic value as an exposition of applied complexity science and applied research, drawing on examples of simulations and modelling throughout, with corresponding insights into the design of new concepts and policies, and the understanding of complex phenomena that are invisible to classical techniques.

1 Introduction

DOI: 10.4324/9781315573205-1

Marc Bourgois

This book seeks to pull together research at the interface between complexity science and air traffic management (ATM) to provide a useful resource to facilitate and inspire future work in this area. The idea for this book originated with ComplexWorld, a network of excellence funded by the Single European Sky (SES) air transport policy initiative of the European Union, in collaboration with EURO-CONTROL, the European Organisation for the Safety of Air Navigation.

This chapter introduces the field of complexity science as it relates to ATM, outlines the origins of the ComplexWorld network and the diverse expertise of the participants who contributed to the book, and sets out the topics that are addressed.

1.1 Complexity in Air Traffic Management

ATM combines human actors with technical subsystems and operational procedures in a complex system characterised by the myriad concurrent interactions between its many components. Complex systems cannot be fully understood by the traditional approach of studying their individual components and subsequently summing their characteristics. This is where complex systems science is required: complex systems science provides formal modelling paradigms and analysis techniques that help in uncovering fundamental properties and suggesting design principles to sustain high performance for large non-linear systems.

It is important to differentiate complex systems science from the term ‘traffic complexity’ used in traditional ATM research. Traffic complexity, often abbreviated to ‘complexity’, is the study of factors and metrics that quantify the relationship between traffic patterns, air traffic control (ATC) sectors and associated controller workload. Traffic complexity is not the subject of this book, which instead focuses on the application of complex systems analysis approaches developed in a range of disciplines to the study of the socio-technical air transport system.

Multi-disciplinary research in complex systems and ATM is in its early days, but publications have recently increased in number and promising lines of research can now be identified. The most straightforward studies concern the statistical properties of air transport networks, essentially in search of characteristic power-law distributions (the relevance of this is clarified in Chapter 2). The first review in this field clearly shows this to be a popular application of complex network theory (CNT), but it also shows that the values found for even the most basic network metrics may differ widely, due to differing modelling choices and possibly dataset quality or completeness issues (Zanin and Lillo, 2013). Rather than leaving the modelling entirely to complex network theorists, a multi-disciplinary approach informing the models with operational and engineering expertise from the air traffic community might lead to more robust results.

More sophisticated modelling projects have uncovered interesting properties describing the coexistence of traditional airlines operating hub-and-spoke networks with low-cost airlines operating point-to-point networks. Results from the Single European Sky ATM Research (SESAR) project ELSA, for example, show that capacity can be optimally exploited with a balance of the two airline populations; airspace exclusively populated by either traditional or low-cost airlines would provide for less optimal trajectories, with more delay or rerouting (Gurtner et al., 2015). On the same subject, multi-layered network models can determine which of those two business models contributes more to individual properties of the overall network (Cardillo et al., 2013a). For example, low-cost airlines are shown to be the main contributor to path redundancy, which increases the flight options passengers have between a given origin and destination airport.

Another popular topic of study with complex systems scientists is the phase transition, indicating a sudden regime change, for example, between free-flow and congested states of a transport network. In road transport, this has led to the Nagel-Schreckenberg model, which identifies when jamming transitions emerge by modelling the microscopic behaviour of numerous car drivers on multi-lane roads. The jamming transitions are replicated with just a few simple driving rules and moderate statistical variability. These seminal models were implemented on efficient cellular automata simulators, which have also been exploited in air traffic, for example, to model ground congestion of the taxiway system at Tokyo International airport (Mori, 2012). The Nagel-Schreckenberg model is not directly portable to airborne traffic, as planes that slowed down too much would simply fall out of the sky; nonetheless an analogous problem called ‘traffic bunching’ exists in en-route and terminal air traffic sectors. A reduction in unsafe traffic bunching is considered to be one of the potential benefits of a major conceptual change to ATM under development: four-dimensional (4D) trajectory planning. Improved understanding of traffic bunching is, therefore, a potentially useful line of research (Chalon et. al., 2007).

The core chapters of this book discuss many more examples of relevant research results and opportunities, often in considerably more detail. The references at the end of the book provide numerous further pointers into the relevant literature for the interested reader. The extent of the references list shows that something substantial is clearly growing in this field of research.

1.2 The ComplexWorld Network

Over the last ten years, the European Union has launched a number of initiatives to increase the competitiveness of the European air transport sector and to prepare the sector for sustained growth in the face of increasing external pressures. One of these initiatives is the SES, for which four high-level goals were identified: coping with a threefold increase in traffic; improving safety performance by a factor of ten; reducing the cost of ATM by a factor of two; and reducing the environmental impact of an individual flight by 10 per cent. Much of the burden of these goals lies with a small but pivotal branch of the air transport sector: ATM.

The organisation of ATM within Europe is unfortunately highly fragmented. Each member state has its own air navigation service provider (operating its ATC centres and airport control towers), its own safety regulator and military authorities claiming priority access or exclusive use of airspace. EUROCONTROL performs centralised network management and the European Union has its own safety regulator. Also significant are the many weather information providers, the aircraft and equipment manufacturers and the airlines operating in European airspace.

To achieve the high-level goals of the SES within such a complex institutional environment, a public–private partnership, the SESAR Joint Undertaking (SJU), was established to identify, coordinate and advance the applied research and development needed for the modernisation of European ATM. As part of the research programme, ComplexWorld was established as a network of excellence to explore the potential of complex systems science to increase the understanding of the performance of the European ATM system.

Until recently, complex systems science had hardly tackled the problems of ATM; key experts on complex systems science were to be found outside of the ATM domain, most prominently in statistical physics. The original membership of the ComplexWorld network was carefully constructed to bring together experienced researchers in complex systems science with the established ATM research community.

The authors of this book are core members of the ComplexWorld network. The academic side of the ATM research community is represented by the department of aerospace engineering from the University of Seville. Two members of this group contribute to the book: Damián Rivas and Rafael Vazquez. Dr Rivas is professor, teaching flight mechanics and air navigation, and is also the scientific coordinator of the ComplexWorld network. Dr Vazquez is associate professor and teaches orbital mechanics and navigation systems. Their ATM research focus is on conflict detection and resolution, trajectory prediction and optimisation from a control theoretic perspective, and uncertainty propagation.

In addition to academic institutions, the European air traffic research community includes several national research centres; these research centres are represented in the network by AT-ONE, a joint venture of the institute of flight guidance of the German Aerospace Centre (DLR) and the air transport safety institute of the Dutch National Aerospace Laboratory (NLR). Their contributions to this book stem from Henk Blom and his co-authors Dr Mariken Everdij and PhD candidate Soufiane Bouarfa. Dr Blom and Dr Everdij have been pioneers in using complexity science approaches to research the socio-technical ATM system. Dr Blom also holds a chair in Air Traffic Management Safety at Delft University of Technology and teaches agent-based safety risk analysis.

From the statistical physics-based complexity science community, a prominent group of researchers from the University of Palermo joined the network. This group is represented here by Rosario Mantegna, Salvatore Miccichè and Fabrizio Lillo. Dr Mantegna, who holds a combined professorship at the Central European University in economics and at the University of Palermo in physics, was among the first to analyse and model social and economic systems with the concepts and tools of statistical physics as far back as 1990 (Mantegna and Stanley, 2000). Dr Miccichè is an associate professor at the faculty of medicine and researches on the characterisation of long-range correlations in bioinformatics and econophysics. Dr Lillo recently moved to the Scuola Normale Superiore of Pisa where he is a professor in quantitative finance; he is also a member of the external faculty of the Sante Fe Institute.

Additional complexity science expertise is provided by Innaxis, a Spanish foundation specialising in complexity research with extensive project experience in the energy and transport sectors. Three staff members from the Innaxis Research Institute contribute to this book: David Perez, its director, Massimiliano Zanin and Seddik Belkoura. Before joining Innaxis David Perez had a long career in the aeronautical industry, having worked for Iberia Airlines and Boeing ATM in both engineering and business development roles. He is ideally placed to identify the main contributions of complex systems research so far and to sketch an outline of the challenges ahead for complex systems scientists dedicated to the air transport domain. Dr Massimiliano Zanin is a prolific researcher in data science. Seddik Belkoura is currently pursuing his PhD on complex systems and data science applied to ATM, under the supervision of Dr Zanin.

Drawing on 20 years’ experience in research on the operational aspects of ATM, Andrew Cook provides a most useful bridge as a transport economist with much experience in the wider air transport domain. Dr Cook is a Principal Researcher in the Department of Planning and Transport at the University of Westminster and has made seminal contributions to the issues of delays and their costs from the perspective of airlines, air navigation service providers and, crucially, passengers. Andrew Cook also took on the editing responsibilities for this book, jointly with Damián Rivas.

Over the last three years these authors and several of their colleagues have joined forces in organising thematic workshops, annual conferences and in guiding and tutoring a cohort of PhD students. They have been collaborating on joint research and compiling a state-of-the-art review of complex systems science approaches, results and open questions relevant to ATM. All of it is documented in an open, collaborative authoring platform at www.complexworld.eu. This book is based not only on the research work of the authors themselves but also on the larger body of reviewed complex systems research.

1.3 Topics Covered in this Book

This book can be seen as a compilation of selected topics in complex systems science applied to air transport and ATM. The selected topics are those for which traditional ATM research has not provided satisfactory solutions and for which complex systems science techniques hold particular promise. This section provides an overview on which concepts and examples are covered by each topic and how these topics relate to each other.

The first topic is uncertainty, in Chapter 4. At the micro scale, Damián Rivas and Rafael Vazquez consider individual flights. One aspect of the uncertainty of individual flights is captured by delays: either in observed delay statistics (often separately for the different phases of flight) or by delay propagation models. Delay cost is crucially important; authoritative studies on delays and their costs are referenced across several chapters. Another aspect of the uncertainty of an individual flight is its trajectory uncertainty, that is, the uncertainty in knowing its position at any given time. Solutions are generally based on probabilistic frameworks that model the causes of trajectory uncertainty, which include the initial conditions, aircraft performance, navigational errors and wind conditions. A classification for the most relevant research contributions is outlined.

The mesoscale corresponds to flights close enough to interact, and to weather systems. This is the realm of tactical ATC and flow management. In addition to trajectory uncertainties, causes of traffic uncertainty include human actions (by controllers and pilots), management procedures and weather uncertainty. Whereas early studies on ground holding for flow management were reliant on deterministic trajectories, recent work exploits Markov models, queuing models, stochastic dynamic programming, etc. An alternative to probabilistic models is to calculate worst-case envelopes. This approach is occasionally followed in conflict resolution strategies or in robust optimisation techniques for adverse weather avoidance. A last example at the mesoscale concerns safety assessments based on rare-event Monte Carlo simulations, which is further explored in Chapter 6, which discusses emergent behaviour. Finally, the macroscale comprises the entire air transport network and although there is a growing body of complex systems studies at this level, few have actually dealt adequately with the issue of uncertainty.

A second topic is resilience. Chapter 5 investigates the added value of complexity science for studying the impact of various disruptions. In general, a socio-technical system may absorb a disruption and adapt to it, or may first change its performance and subsequently restore and adapt. These three aspects of resilience, that is, its absorptive, restorative and adaptive capacities, allow us to relate resilience to well-known systems engineering concepts such as dependability and robustness. A further distinction is that resilience aims to address socio-technical systems, rather than technical systems.

Resilience metrics are reviewed from a wide range of scientific disciplines in which resilience has been studied, largely independently. These include ecosystems, critical infrastructures and even psychology. Though the reviewed metrics provide initial guidance, the complexity induced by the combination of social and technical components in the air transport system and the difficulty of measuring the depth and time-evolution of the impact of disturbances leave the issue of deriving a resilience metric as an open challenge.

As complexity stems from the interaction between a large number of subsystems, modelling techniques from complexity science which focus on subsystems interaction are better suited to capture the resilience phenomenon. Agent-based models, network flow-based models and the lesser known stochastic reachability analysis and viability analysis are obvious candidate modelling techniques. In the final section of this chapter, disturbances and how they can be dealt with by human operators are illustrated by several examples of disruption management policies in a typical airline operations centre. The authors show that an agent-based model building on state-of-the-art coordination theory leads to better solutions than the heuristics currently deployed at airline operations centres.

A final topic is emergent behaviour, which is explored in Chapter 6. In complexity science a property or behaviour of a system is called emergent if it is not a property or behaviour of the individual, constituting elements of the system, but results from the interactions between its constituting elements. The further study of this broad spectrum of emergence has occupied philosophers and scholars from complexity science. This has led to a constructive definition of three categories of emergence, based on whether it can be predicted through mathematical analysis, only through simulation, or cannot be predicted even through simulation. For each of these emergence categories illustrative ATM examples are given. Emergent behaviours that can be predicted are of use in the design of a future socio-technical ATM system. A natural fit for predicting emergent behaviour is again provided by agent-based modelling and simulation (ABMS), which is a well-established complexity science paradigm. However, the use of ABMS for the purpose of identifying the capacity and safety properties of a novel socio-technical ATM design poses challenges. Pilots and controllers play a key role in safely handling incidents and avoiding their escalation. Therefore accidents are very rare events, occurring once in every 10 million or even in every billion runs, posing challenges to ABMS. This can only be managed using advanced mathematical techniques, such as compositional modelling and importance sampling of stochastic hybrid systems. An extensive model analysis of a novel self-separation concept for autonomous aircraft, based on previous work from NASA, shows that surprising levels of safety can emerge under extreme high-density operations.

The complexity science approach to these three topics suggests design principles for a high-performance socio-technical ATM system. For the discussion on these topics to be accessible to the reader unfamiliar with complexity science, several modelling paradigms and analysis techniques need to be introduced. Application of these paradigms and techniques to ATM leads to insights in their own right, in addition to supporting the more applied discussions on the topics already introduced. One modelling paradigm central to complex socio-technical systems analysis is ABMS, which is introduced and illustrated in Chapter 6. Another modelling paradigm, CNT, is explored in Chapter 2. This is a technique which not only produces compact representations but also characterises the structure of a complex system across domains as varied as transport, utilities and biology. Andrew Cook and Massimiliano Zanin introduce basic metrics that measure properties of networks, their substructures and their dynamics (further metrics focusing on output, which equates to per...