At the end of this 2nd reporting period, the overall achievement can be set at 100%. Out of the 5 objectives described above, all of them have been successfully achieved.
We have been able to answer positively the Research questions posed at the beginning of the project
Research Questions (RQ)
RQ#1: Can trajectory level uncertainty be modelled, assimilated on a cycle based, and propagated?
RQ#2: Can ATM network uncertainty be modelled (including thunderstorms), cyclically assimilated, and propagated?
RQ#3: Can a robust operational plan for ATM system resilience be found?
RQ#4: Can this advanced functionality (build upon successful achievement of RQ#1 to RQ#3) be implemented in operational dispatching tools such as FLIGHTKEYS’s one?
Conclusion #1: The proposed study has the applicability and suitability for uncertainty propagation in the aircraft trajectory prediction process. It has been shows how, when applying the framework to a relevant scenario within the European air traffic, the results obtained for estimating the probability distribution of the flight times resembles the actual values observed, even when considering weather uncertainties and thunderstorms as disruptive events. It obtains results comparable to those retrieved by using a complex aircraft trajectory prediction tool while reducing the involved computational time and complexity, as it avoids the application of computationally demanding methods (e.g. a Monte Carlo simulation).
Conclusion #2: All in all, the results prove that ATM network uncertainty can be modeled and propagated. We have observed that regarding the problems of modeling network propagations, a combination of mathematical model and data-driven approach provides more flexibility and robustness compared to a fully switch over from mathematical modeling to purely data-driven approach. This way it becomes possible to fully harness both the stability of mathematical models and the insights of data-driven models.
Conclusion #3: The results show that a robust operational plan can be obtained for a set of flights or for the complete set of flights.
The answer to RQ#4 is affirmative:
Conclusion #4: All in all, we have shown that all the methodologies, processes, and algorithms developed within START project could be integrated in a flight dispatching tool such as FK5D. It would, of course, require further development and integration, but this can be considered as future research and development.
