Periodic Reporting for period 2 - Pilot3 (A software engine for multi-criteria decision support in flight management)
Reporting period: 2020-11-01 to 2022-01-31
Pilot3, a Clean Sky 2 Innovation Action, developed a software engine for supporting crew for civil flights. It provides a set of trajectory alternatives enabling the crew to select the most suitable one considering multi-criteria business objectives of the airline (achieving the objective of measuring airlines' performance) and operational uncertainties (meeting the objective of considering operational environment uncertainties on the solutions). Currently, when disruption occurs pilots do not count on a system which enables them to have a full understanding of the trade-offs involved in trajectory changes leading to sub-optimal decisions in cost and environmental impact.
Pilot3 integrates airlines' objectives, defined as on-time performance (OTP) and cost (fuel, passenger disruptions and other costs), when generating alternatives (as defined in the objective of understanding airlines flight management policies). Sub-objectives are considered as part of the ranking of solutions. The optimisation framework, which considers the business objectives and operational limitations, has been identified achieving another objective of the project. This framework enriches airlines policies, which is an objective of the project.
Pilot3 tackles two of the main problems faced during the flight: visibility on the end objectives of the airlines, by computing the expected costs as a function of delay at the gate (including intrinsic uncertainties, e.g. passengers missing connections and expected reactionary costs), and by incorporating the modelling of uncertainties due to operational aspects (e.g. holdings at arrival or taxi-in time). This achieves the objective of considering the overall impact of each trajectory options, considering network effects and passengers’ itineraries.
Pilot3 provides indicators needed to estimate the expected cost of delay and operational uncertainties (holding, sequencing and merging distance, taxi time). The individual estimators used can be configured, selecting heuristic or machine learning based models which can be either updated with data while airborne or rely on pre-departure information. This flexibility is a requirement to meet one of the objectives of Pilot3.
Pilot3 integrates airlines' objectives, defined as on-time performance (OTP) and cost (fuel, passenger disruptions and other costs), when generating alternatives (as defined in the objective of understanding airlines flight management policies). Sub-objectives are considered as part of the ranking of solutions. The optimisation framework, which considers the business objectives and operational limitations, has been identified achieving another objective of the project. This framework enriches airlines policies, which is an objective of the project.
Pilot3 tackles two of the main problems faced during the flight: visibility on the end objectives of the airlines, by computing the expected costs as a function of delay at the gate (including intrinsic uncertainties, e.g. passengers missing connections and expected reactionary costs), and by incorporating the modelling of uncertainties due to operational aspects (e.g. holdings at arrival or taxi-in time). This achieves the objective of considering the overall impact of each trajectory options, considering network effects and passengers’ itineraries.
Pilot3 provides indicators needed to estimate the expected cost of delay and operational uncertainties (holding, sequencing and merging distance, taxi time). The individual estimators used can be configured, selecting heuristic or machine learning based models which can be either updated with data while airborne or rely on pre-departure information. This flexibility is a requirement to meet one of the objectives of Pilot3.
The consortium considered of paramount importance to incorporate the views of relevant stakeholders in these early stages of the project. A workshop with the Advisory Board was carried out (FEB20) and a follow-up survey and bilateral discussions with some members of the Board and the Topic Manager conducted. These activities led to the definition of the objectives to be considered by the optimiser, the addition of the Operational ATM Estimator module, to estimate operational uncertainties, and provided input regarding the most suitable methods for optimisation (reported in D1.1 - Technical resources and problem definition (MAR20)).
The multi-criteria decision making technique selection process, with input from the Advisory Board (workshop and survey), not only selected a method but defined the prototype optimisation framework. These considerations, including a thorough literature review were reported in D2.1 - Trade-off report on multi-criteria decision making techniques (MAY20) reaching MS2.
The data collection and management focused on the interaction with stakeholders. These activities were reported in D3.1 - Airlines data collection report (JUL20) leading to MS3 - Airlines data collected and analysed.
A verification and validation plan was defined as part of WP5 and reported in D5.1 - Verification and validation plan (AUG20).
Two prototype releases have been delivered. The first one in JUL21, as reported in D4.1 and D4.2. This first proof of concept prototype was presented to the Advisory Board in an on-line workshop where feedback was gathered for the prioritisation of the final modelling activities, achieving MS4 - First release results review.
The final part of the project focused on the final implementation and integration of functionalities along with the validation activities (WP5) for the final release. This prototype was presented to the Advisory Board in a workshop (JAN22) where feedback on future potential developments was gathered, and some ideas on the path toward industrialisation identified (WP6). A follow-up survey was used to validate the design of the HMI. This final release concluded the development of Pilot3 with the production of D4.3 and D4.4. The known issues and next steps were reported in D6.1 - System evolution and uptake.
The final prototype is modular and configurable. Heuristic and machine learning estimators can be used and the optimisation of the vertical profile of the flight is performed considering the total expected cost with two approaches: optimising the cost index as a proxy control variable, or with a grid-search on the altitude and speed domain. The prototype and its components have been validated, as reported in D5.2 - Verification and validation report, reaching MS5 - Final acceptance.
Communication and dissemination activities include: the communication, dissemination and exploitation plan (D7.1 - MAY20), the project website, publication of six blog entries, participation at two editions of AGIFORS conference (one of them awarded as best conference presentation), participation at the PACEDays event 2020, two poster (with video) presentations at SESAR Innovation Days 2020 and 2021, and more than 25 further dissemination actions via social media (LinkedIn). Two journal papers with contributions from Pilot3 were submitted but not yet published. All these actions are reported in D7.2 - Project communication, dissemination and exploitation report.
Finally, the project management has been done as part of WP8 leading to D8.1 - Project management plan (NOV19), D8.2 - Proof of signature of Consortium Agreement (JAN20), six periodic monitoring reports (PRM), intermediate review meeting (IRM, MS3.1) and technical discussions with Topic Manager and Project Officer to plan the final development cycle (AUG21).
The multi-criteria decision making technique selection process, with input from the Advisory Board (workshop and survey), not only selected a method but defined the prototype optimisation framework. These considerations, including a thorough literature review were reported in D2.1 - Trade-off report on multi-criteria decision making techniques (MAY20) reaching MS2.
The data collection and management focused on the interaction with stakeholders. These activities were reported in D3.1 - Airlines data collection report (JUL20) leading to MS3 - Airlines data collected and analysed.
A verification and validation plan was defined as part of WP5 and reported in D5.1 - Verification and validation plan (AUG20).
Two prototype releases have been delivered. The first one in JUL21, as reported in D4.1 and D4.2. This first proof of concept prototype was presented to the Advisory Board in an on-line workshop where feedback was gathered for the prioritisation of the final modelling activities, achieving MS4 - First release results review.
The final part of the project focused on the final implementation and integration of functionalities along with the validation activities (WP5) for the final release. This prototype was presented to the Advisory Board in a workshop (JAN22) where feedback on future potential developments was gathered, and some ideas on the path toward industrialisation identified (WP6). A follow-up survey was used to validate the design of the HMI. This final release concluded the development of Pilot3 with the production of D4.3 and D4.4. The known issues and next steps were reported in D6.1 - System evolution and uptake.
The final prototype is modular and configurable. Heuristic and machine learning estimators can be used and the optimisation of the vertical profile of the flight is performed considering the total expected cost with two approaches: optimising the cost index as a proxy control variable, or with a grid-search on the altitude and speed domain. The prototype and its components have been validated, as reported in D5.2 - Verification and validation report, reaching MS5 - Final acceptance.
Communication and dissemination activities include: the communication, dissemination and exploitation plan (D7.1 - MAY20), the project website, publication of six blog entries, participation at two editions of AGIFORS conference (one of them awarded as best conference presentation), participation at the PACEDays event 2020, two poster (with video) presentations at SESAR Innovation Days 2020 and 2021, and more than 25 further dissemination actions via social media (LinkedIn). Two journal papers with contributions from Pilot3 were submitted but not yet published. All these actions are reported in D7.2 - Project communication, dissemination and exploitation report.
Finally, the project management has been done as part of WP8 leading to D8.1 - Project management plan (NOV19), D8.2 - Proof of signature of Consortium Agreement (JAN20), six periodic monitoring reports (PRM), intermediate review meeting (IRM, MS3.1) and technical discussions with Topic Manager and Project Officer to plan the final development cycle (AUG21).
The multi-criteria decision making techniques selection was performed following a domain-driven approach. This led to a solution which is tailored to the trajectories for tactical crew support problems. This represents a progress beyond the state of the art, as far as the consortium is aware. The consideration of explicit cost functions and modelling of the impact of operational uncertainties (such as holding or taxi) are also new contributions to the field of tactical trajectory optimisation. The optimisation of the trajectory considering this total expected cost is an improvement with respect to the classical optimisation based on pre-computed cost indexes.
The modular solution of Pilot3 enables us to produce benefits on different fields (cost estimation, operational uncertainty estimation and trajectory optimisation). These elements could potentially be used beyond Pilot3 on different solutions, e.g. the cost estimations could be used as part of decision making processes performed on-ground by the airlines operating centre, or the trajectory optimiser could incorporate new objectives such as the environmental impact of the flight.
The Pilot3 solution could enable crew to make more informed decisions, particularly regarding the expected uncertainties at arrival leading to more efficient and environmentally optimal solutions.
The modular solution of Pilot3 enables us to produce benefits on different fields (cost estimation, operational uncertainty estimation and trajectory optimisation). These elements could potentially be used beyond Pilot3 on different solutions, e.g. the cost estimations could be used as part of decision making processes performed on-ground by the airlines operating centre, or the trajectory optimiser could incorporate new objectives such as the environmental impact of the flight.
The Pilot3 solution could enable crew to make more informed decisions, particularly regarding the expected uncertainties at arrival leading to more efficient and environmentally optimal solutions.