CASSIOPEIA

About

CASSIOPEIA - Complex Adaptative Systems for Optimisation of Performance in ATM - puts at work the techniques of Complex Systems to build up the scientific foundations of an innovative approach to the understanding of ATM performance, especially well-suited to represent the diversity of involved and to apprehend and manage the irreductible uncertainties which impact on ATM processes.

The specific information about CASSIOPEIA project extension, named "DCI-4HD2D", can be found here

The air transport complexity problem

The understanding of Air Transport in general and Air Traffic Management in particular has been a field deeply analyzed from many different perspectives. Despite large efforts, there is still some phenomena whose behavior is far from being totally understood, especially phenomena associated to the as a heavily networked system of systems, in which many different independent elements interact with each other in a non-linear fashion. This interactions create emergent behavior that go beyond the concepts that were originally defined in the so-called socio-technical system specifications.

Despite many and large previous efforts of research, the relation between network behavior and performance is still not thoroughly understood. This is obviously a major drawback for a sensible assignment of solutions and resources to the challenges the ATM system is facing today and in the future, and ultimately for defining the right governance mechanisms that should address delay propagation, capacity limits, network congestion and other phenomena, whose origin may be local but which can produce large-scale effects, as recently exemplified by the Eyjafjalla crisis.

The air transport modelling problem

The role of the modelling, simulation and analysis community is critical. When a founded data-driven modelling exercise is correctly able to derive the understanding of the key forces of the system, it enables the decision makers to the foresighting of potential future scenarios for the network and to reach the performance they target in different configurations. The vast amount of data available nowadays has allowed to uncover the fact that most systems are not deterministic and, in fact, they behave as complex systems, like they have been called in the recent years by the scientific community. Complexity Science is a new area of study that research on the understanding of the behavior and performance optimisation of these systems.

Modelling the air transport system has researchers all over the world doing repetitive tasks. Even the same researchers in a single company usually have to do the same tasks every time they design a model with a specific purpose. These tasks include defining the attributes of the different agents of the model such as aircraft performance, airport characteristics, air navigation system structure, or operational procedures, and implement all this into a specific software language. This is due to the fact that all agents and attributes need to be specifically designed for the purpose of the model in terms of scale and detail, in order to have a homogeneous level of certainty (and uncertainty).

Besides the definition of the generalities mentioned above, air transport modelling has other commonalities. When designing a model, researchers try to represent reality up to a certain point. The purpose of each model usually entails to include a change in the way agents interact and observe the reactions of the system.

This problem represents a high amount of time, effort and economic losses both for private companies and public entities.

The solution proposed

The CASSIOPEIA Project, Complex Adaptive Systems for Optimisation of Performance in ATM, aims at building the scientific foundations of a paradigm for ATM performance modelling based on Complex Systems Science. For this purpose, CASSIOPEIA has used an agent-based modeling approach to represent the most relevant ATM processes at different temporal and spatial scales, which have made possible to apprehend explicitly the different business perspectives, embedding a comprehensive modelling of the uncertainty associated to ATM processes. In order to fulfill this concept, CASSIOPEIA adapted and combined techniques to tackle the emergent phenomena, entanglement of multiple spatial and temporal scales and the non-determinism and uncertainty in the ATM modelling world.

The Cassiopeia team, understanding deeply the challenges of Air Transport and Air Traffic Management, looked into the development of breakthroughs in ATM modelling paradigms that would help the ATM research community and the SESAR programme to accomplish their goals by integrating a further understanding of ATM phenomena in their models, simulators, systems, tools and decision making processes.

Case studies proposed

In order to show the potential of the software platform and the modeling techniques, three Case Studies were selected and explored once the software was completed. These Case Studies were selected with extensive consultation from ATM stakeholders based on their relevance and alignment with the project’s objectives.

The purpose of the Case Studies was to guide the modeling, providing a preliminary validation of a model and giving a first assessment of a model´s capabilities. These three case studies will be conducted at the final stage of the project and will encompass different types of modeling techniques.

The Case Studies are as follows, each exploring the potential outcomes of a specific proposed regulation.

Case study 1

This case study analyses the impact and implications of airport environmental regulations on the airlines affected and on the rest of network, taking into account, particularly, the rescheduling of flights and a regionalization of night traffic and performing a cost-benefit analysis for quantifying the economic cost of each measure on the airline operators versus the environmental benefits (noise alleviation, reduction of emissions).

Health problems related to aircraft noise are being studied worldwide. Studies regarding the impact that aircraft noise has on human health have proved the relationship between the noise and loss of sleep or cardiovascular arousal. There are also indirect consequences which may include high blood pressure, hypertension, diabetes, obesity or heart disease among others. Aircraft noise impact around the major European airports has become one of the most important potential restriction to air transport in the European Union. Noise reduction is approached from four different measures: reduction of noise at the source, operational procedures, land use planning and operative restrictions. Each airport would apply the combination of them that better fits its particular features.

Operational procedures have also been developed allowing the design of arrival and departure procedures to avoid flying over most populated areas and continuous approaches, maintaining highest altitude over populated areas for the longest time possible. The use of full engine power, specific runways or thresholds have been considered at different airports.

While the International Civil Aviation Organisation (ICAO) considers operative restrictions as the last resource, applicable when the other three have been insufficient to solve noise problems, setting limits to the operation of the noisiest aircraft types, in particular at night, is now very frequent in the major European airports. In some cases, a total prohibition of commercial night flights (night curfew) has been imposed. The impact of such an extraordinary measure does not include only the obvious direct effects on the airlines or the airport, but also an indirect socioeconomic impact on the communities. For the airlines, it is important to note that night flights are usually made by cargo or hub airlines. These types of airlines usually follow hub-and-spoke strategies, which means that they fly to or from a base. Hence, those flights affected will be cancelled if they cannot by reallocated to a close schedule; and if too many flights are cancelled, it may be interesting for the airline to move the hub to another airport.

The case study analyses the impact that a curfew regulation would have on the air traffic network if it were applied to the 10 busiest European Airports. Airlines and airports are classified and modelled in such a way that each subcategory searches for its best solution, as they would in real life. In the regulated airports there would be a cumulative capacity demand just before and after curfew, and each airport would sell the slots to those airlines which have a hub at the airport, many flights, heavy aircraft, and other criteria to maximise their profit. The airlines also have some decision-making to do. In this case study, based on the slots offered by the airport, they will decide whether to stay at the airport, move to a regional airport or cancel the flight, the decision will be influenced by the type of company, its relation with the airport, the distance to the alternative airport, and the time interval between the original and the new schedule. The previous and next flights by the same aircraft are checked to ensure that a change in schedule will not overload the capacity at the connecting airports.

The model analyses different indicators to measure different performance aspects of the agents involved. These metrics allows the regulator to understand the economic, operational and social impact (noise levels reduction at populated areas) of the regulation imposed.

While the model is based on this regulation, the software platform allows the users to easily change the inputs of the regulations, such as changing the curfew, to optimise the regulation.

Stakeholders involved in the process may be classified in four different categories:

• Authorities imposing the restrictions, either national or local ones
• Airports applying the restrictions, considered as independent entities (no coordinated policies of facilities owned by the same company, like Charles de Gaulle and Orly, belonging to Aéroports de Paris, are considered)
• Airlines operating commercial flights at the airports
• Local communities of residents living around the airports

The central scenario for Case Study 1 is the generalization of the ban on night flights in airports. Also, a sensitivity analysis is performed on the influence of the size of the time interval in which the night ban is applied. The results of the simulations can be looked at from different perspectives, expressed by the different Performance Indicators, with the purpose of highlighting the repercussions on the different stakeholders:

Economic impact on airlines: Results show that the restrictions imply a relatively important economic impact on airlines, both for network carriers with a hub at the regulated airport and for the rest of the companies, regardless the decisions these companies may take in response to the restriction. The same conclusion applies to low cost carriers. Indicators show that even if only a small percentage of flights are affected, the revenues of the airline may suffer a non-negligible impact on profits. These indicators are sensitive to the variation of the size of the night ban interval with greater impact at night than in the morning.

Economic impact on airports: This impact is different depending on the airport, and varies from one airport to another. This fact depends on the number of night flights that each airport had before the night ban implementation, and the possibilities of that airport to accommodate potential re-scheduling as a consequence of the restriction. Results also show how alternate airports, where airlines move the flights that cannot re-schedule in the restricted airport, increase their economic results, in a larger proportion in airports relatively small, compared to the restricted airport.

Socio-economic impact on local communities: Impact on local communities, both in terms of yearly economic losses and jobs affected by having lower traffic is very important. It has to be taken into account that these socio-economic figures measure the full range of effects: direct, indirect, induced and catalytic.

Environmental impact: Regarding noise, obviously the night ban eliminates the noise problem during the period at which the ban is applied, but it shifts the noise load to the hours adjacent to the ban and also to the alternate airport, where an important portion of the flights are moved. The impact of the restrictions on emissions affecting local air quality in the airport area is almost negligible.

Final Study Report Case Study 1

Case study 2

Designing an ATFM slot exchange procedure, when game theory and agent-based modelling collide

The European airspace is structured in sectors. Each sector corresponds to a physical airspace volume. Air traffic flying across each sector is managed individually, in coordination with adjacent sectors. The maximum number of aircraft allowed in a given sector during a given timeframe is known as the sector’s current capacity. The maximum capacity of a sector is limited by a number of factors like complexity, staff, equipment, weather, etc., and, therefore, it fluctuates throughout the day. Nevertheless, individual flight plans are evaluated in advance so that capacity is never surpassed.

However, when any sector capacity suddenly drops, flight demand may no longer be accommodated. If the difference between demand and capacity is notable, a sector is tagged as regulated. This regulation reduces the actual traffic flows in the affected sectors by imposing delays to on-ground aircraft. Each flight affected is given an ATFM slot. These slots are usually distributed using a first-filed, first-served basis, so that the delay is, roughly, uniformly distributed, ensuring no flight is given preferential treatment for any reason other than schedule. However, airlines might be interested in prioritizing valuable flights (e.g. flights early in the morning with a large number of flight legs ahead).

For that reason, one of the cornerstones of the SESAR program is to promote Collaborative Decision Making (CDM) processes. These processes should enable the affected Airline Operators (AOs) to negotiate and ultimately agree on an ATFM (regulated) slot distribution according to their own preferences and not only to single-flight efficiency, as in some cases agreed distributions may be suboptimal in terms of single-flight delays but reduce cost overall.

Nevertheless, the design of such CDM processes is not an easy task. CASSIOPEIA’s Case Study 2 focuses on the design and test of one possible CDM solution, that regarding the distribution of ATFM slots. First, the exchange mechanism is designed using game theory, and secondly, this mechanism is tested using CASSIOPEIA’s agent based model. A number of assumptions need to be made in order to properly evaluate the exchange mechanism and therefore it is not the aim of the case study to provide any insight into real-world outcomes if this mechanism were ever to become functional. In contrast, this case study focuses on how well different tactics and strategies may perform, how they may interact and couple, and how well the proposed exchange mechanism behaves under degenerate or untrustworthy behaviors. This is of particular importance as explained below.

In game theory, one-on-one CDM mechanisms in which a good is traded offering an economic compensation in return are usually referred to as ‘bilateral exchange mechanisms’ or ‘bilateral trading’. Despite the particularities of the traded goods, desirable properties of bilateral trading are the following:

• Individual rationality, a participant will only bid or sell when a non-negative payoff occurs (in other words, there is no forced participation. It is up to the participants whether they want to exchange their goods or just remain as they are).
• Budged balance, amounts are transferred among participants, so that prices paid and received add up to zero (or to put it more simply, there is no external funding or profit).
• Incentive compatibility, there is no way for a participant to increase its payoff by misrepresenting its cost, therefore lying shall have no reward.

Unfortunately, according to the Impossibility Theorem (Myerson and Satterthwaite, 1983), a classical result in game theory, there is no bilateral exchange mechanism for which 1, 2 and 3 can possibly hold simultaneously. Since it is not clear how to force AOs to participate in a new slot exchange scheme, it seems reasonable to apply individual rationality. It should also be very unlikely to see a slot distribution mechanism funded externally, or making any profit, so it seems a good idea to incorporate budged balance. This forces us to relax the incentive compatibility condition.

In addition, the ATFM slot exchange mechanism has been designed so that the following additional properties are also met:

• Minimum disclosed information, the participants do not share their true cost, nor the maximum amount to bid, but just the amount they are willing to pay in exchange for a given slot.
• Anonymous bidding, the amounts are offered without revealing the identity of the bidder, nor the rest of the offers/exchanges taking place in the same iteration.
• Feasibility of the solution, the final solution achieved needs to be feasible from the ATM perspective.

Of course, in order to fulfill these additional properties, an arbitrator figure should be introduced, namely: the bid coordinator. The bid coordinator would collect all bidding information (who is bidding, how much, and what for), offer the amounts anonymously, listen to the replies, and communicate the results back to the bidders. The last property would most likely be carried out by the current airspace operator in charge. Very likely the network manager, or a new section inside, will act as bid coordinator. For simplicity, throughout the case study we implicitly assume that the network coordinator takes over the bid coordination and ensures that these three additional properties are fulfilled.

Study Report - Case Study 2

Case study 3

Dynamic cost indexing

Case study 3 will assess the impact of airlines using Dynamic Cost Indexing to minimize delay costs using variable aircraft speeds. The introduction of extended dynamic cost index in the Air Navigation System will create a series of changes in the different Performance Areas: predictability, cost efficiency, flexibility, environmental impact and uncertainty.

This model allows the Airlines' Operations Centers (AOCs) to obtain a clear picture of the impact that delays have on subsequent flights. The Agent Based design of the model, allows the AOCs to communicate with the different aircraft, and apply different strategies based on their independent situation.

Each time an aircraft is ready to load passengers, it requests its AOC how long it should wait for delayed connecting passengers, and the cost index that should be applied in flight. The AOC calculates the delay cost generated by waiting for those passengers, and also the delay cost generated at the end of the flight, taking into account different factors as explained below. Once in cruise, the aircraft requests again for a review of the cost index, taking into account the uncertainties up to that point.

A function of delay cost is generated for every flight and the airlines then identify the best cost index for each flight, this cost index calculation will modify the speed of the aircraft, increasing or decreasing fuel consumption and emissions, saving on the overall delay costs.

The data used in case study 3 refers to the tactical costs of delay to the airline. Some of these costs are based on calculations presented in " European airline delay cost reference values" (Cook and Tanner, 2011), which is a standard cost reference (also used by EUROCONTROL). The costs used in case study 3 are:

• Crew and maintenance costs: These costs include the costs of additional crew hours and maintenance associated to the delay incurred by the aircraft. The costs associated to cruise and arrival management are not contemplated since they would be similar with or without delay.

• Passenger costs: The calculations for the cost of delay to the airlines are driven significantly by the costs of delayed passengers, so it is very important to build a good model for the allocation of passengers to the flights. The costs of passenger delay is classified as either "hard" or "soft" as described in Cook and Tanner (2012). "Hard" costs are due to passenger rebooking, compensations and care. "Soft" costs may appear in several different ways; for example, a passenger with a flexible ticket may arrive at an airport and decide to take a competitors on-time flight, or, due to a delay on one occasion, a passenger may be deterred from an unpunctual airline. The "soft" costs are harder to quantify. For the calculation of CI by the AOC due to connecting passenger delay, the connections of the different passengers must be known as well as the number of passengers on each flight to know how many can be reallocated to the next flight. Using the passenger connection data from Zurich Airport, we are able to update the cost index so that it is possible to know which passengers will miss the connections and how much it will cost the airline. The explicit cost of allocation uses the fact that passengers are allocated to each aircraft load factors. The costs of a delay is also affected by the type of airline operation (full-service, regional, LCC or charter).

• Fuel burn: The optimal cost solution, expressed in cost index (CI), will be somewhere between maximum fuel conservation, CI(0), and minimum flight time, CI(max), almost corresponding to the maximum and minimum operating speeds of the aircrafts. This range of CI will be reduced due to comfort (for passengers and crew due to noise in cabin and cockpit) and reduced efficiency on fuel burn. Airlines usually define a limiting CI value for their operations which varies depending on aircraft type. The maximum speed variations are around 8% and the minimum around 4%, Airbus suggests working speed envelopes of 4-6% but in case study 3 we will use the ranges of +/- 3%.

E.02.14 - D4.3 - CASSIOPEIA - Study Report: Case Study 3

Final Technical Report

CASSIOPEIA technical presentation video

YouTube Video

GoogleDrive

Research Team

Coordinator: Innaxis

Partners: University of Westminster, Universidad Politecnica de Madrid