Privacy-Preserving and Fairness-Aware Optimization in the Time-Critical Setting of Air Traffic Flow Management

Description

Optimization problems involving multiple stakeholders are often solved using heuristics. In air traffic flow management, for example, flights are assigned to slots on a first-come, first-served basis if there is an imbalance between the demand for arriving flights and available airport capacity. While this approach ensures safe and smooth airport operations, optimizing the assignment of flights to slots based on airline and airport preferences may improve overall efficiency. However, the preferences of an airline, for example, are sensitive information that must be protected from stakeholders and others. Privacy-preserving implementations of optimization algorithms protect sensitive information but may not finish within the time available for optimization, which can be in the order of minutes or even seconds. Furthermore, an assignment may improve overall efficiency by favoring the flights of some airlines and disadvantaging the flights of others. However, the airlines have competing business interests, and therefore no airline should have an advantage or disadvantage over the others over time. In this presentation, we propose separating the search for solutions from the evaluation of solutions for privacy-preserving optimization in time-critical settings. The search for solutions uses evolutionary algorithms such as genetic algorithms, while the evaluation of solutions uses cryptographic methods such as secure multi-party computation to perform computations on encrypted data. We further introduce ideas for ensuring fairness among competing stakeholders in such a framework. Our research is part of the HARMONIC industrial research project, funded by the SESAR Joint Undertaking within the Horizon Europe program, with partners Frequentis AG, Swiss International Air Lines, Zurich Airport, EUROCONTROL, skyguide, and AIT Austrian Institute of Technology, among others.
Keywords: Air traffic management, evolutionary optimization, secure multiparty computation, privacy-preserving computation
Eingeladen von Prof. Dr. Eri Itoh.

Period	31 Oct 2025
Held at	Research Center for Advanced Science and Technology (RCAST), University of Tokyo, Japan
Degree of Recognition	International