Objectives
The European ATM system is evolving. Changes to its architecture create potential impacts that extend beyond their local application. Modifications introduced both through SESAR initiatives, and wider technological and policy changes, may result in a priori unrelated sub-systems impacting each other due to high levels of coupling. Moreover, the coupling of these sub-systems might be different across various stakeholders, as propagation effects (such as delay or cost) in the network are not necessarily the same from each stakeholder's perspective (e.g., how delay is propagated by flights or by passengers). Domino defines new centrality and causality metrics specifically designed for the ATM domain. It furnishes an agent-based model, which captures the behaviour of key ATM stakeholders and systems such as airspace users, flights, the Network Manager, arrival managers, and passengers.
Domino’s metrics have been tested by introducing three new mechanisms: 4D trajectory adjustment (based on dynamic cost indexing and wait-for-passenger rules at hubs); flight prioritisation (based on UDPP principles to prioritise flights at arrival ATFM regulations), and; flight arrival coordination (which deploys the principles of E-AMAN, whilst introducing the consideration of different functionalities and arrival manager horizons). These mechanisms have been evaluated individually and conjointly, and considering different uptakes and implementation levels: baseline (based on current practices); advanced (foreseen evolution), and; ‘exploratory advanced’ capabilities.
The project showed that the new network metrics based on centrality and causality metrics capture important aspects of the air transportation system. Centrality metrics have shown to be important to capture the loss of connectivity experienced by passengers, while causality metrics have proved to be complex but valuable to truly represent network-based effects of mechanisms with respect to the propagation of disruption.
The full agent-based model of the ATM system developed, ‘ABM-Mercury’, comprises eight agent types (with approximately 32 000 instances of them in a given execution), with their behaviour being driven by their estimation of costs. The model considers a full day of ECAC operations, including flights (with reactionary delay) and passengers itineraries (with connections). The model is executed in an event-driven simulator environment from which the state of the system emerges. The code is sufficiently generic to be re-deployed in diverse context situations and can be applied to other (e.g. historical) datasets and expanded with new agents, mechanisms and strategies.
