In June 2020, we kicked off Dispatcher3, an Innovative Action within the frame of CleanSky 2, led by the University of Westminster (UoW) in collaboration with Innaxis, the Universitat Politècnica de Catalunya (UPC), Vueling Airlines, PACE Aerospace Engineering & Information Technology and the ANSP skeyes. This 2.5 year long project aims to improve dispatching processes by providing an infrastructure that is capable of leveraging historical data and machine learning techniques to systematically estimate variability between planned and executed flight plans.

These differences in flight plans are driven by a number of uncertainty factors, e.g. the weather or the delay due to holding upon arrival. In the current concept of operations, strategically, an airline considers its business objectives in order to identify the key performance areas (KPAs) that are of its interest and defines how they will be monitored with individual key performance indicators (KPIs). Such information is then used by schedule planners, dispatchers and pilots during the operations or to define specific flight policies.

This is where machine learning jumps in: by building intelligent estimators that utilise historical data to generate more accurate forecasts of various indicators, thus reducing uncertainty, machine learning can provide predictive capabilities that dispatchers, crew, network planners and other interested users can rely on to make more informed decisions when selecting flight plans or building schedule networks. To achieve that goal, Dispatcher3 relies on a cloud-based infrastructure powered by DataBeacon to store and process data. It aims to develop two modules:

• module of predictive capabilities, which will provide a number of machine learning based indicator estimators
• module of advice capabilities, whose main objective is to take the raw prediction from the previous module and turn it into advice for a certain type of user, thus reducing information overflow and improving user experience

Various user types can benefit from the software prototype developed in Dispatcher3:

• dispatchers, who can use it to identify the precursors of the different variations between planning and execution and highlight the factors influencing these variabilities
• crew who could benefit from having an indication of the variances that they can expect during their flight and follow-up rotations
• duty managers who can benefit from enhanced predictive capabilities to identify which flights might suffer from disruptions in the network with a few hours of look-ahead
• tactical planners, schedule planners, etc.

Dispatcher3 was presented during the 2020 edition of SESAR Innovation Days (SIDs). The project teaser video is available at the following link and if you want to hear more details about the project methodology and other possible use cases, grab a drink, sit back and enjoy this five minute long video presentation originally shown at SIDs 2020.