Using Artificial Intelligence to Help Military Aircraft Fly Performance-Based Navigation

The recent geopolitical situation in Europe has highlighted the need for an integrated aviation infrastructure that supports operations conducted by both civil and military aircraft, ensuring the adequate levels of safety and system interoperability. A Global Positioning System (GPS) Precise Positioning Service (PPS) receiver on board of military aircraft is an encrypted solution which supports navigation and time, as well as other demanding military applications. Therefore, a GPS PPS receiver is expected to be a dual-use solution to conduct military operations, as well as flying in accordance with the International Civil Aviation Organization (ICAO) rules in the European airspace, provided that it meets the civil performance requirements. Dual-use of GPS PPS opens the door for reducing retrofit and integration costs by avoiding the need to install GPS Standard Positioning Service (SPS) receivers, which equip civil aviation, in military aircraft, while maintaining the appropriate levels of safety and capacity of the network. Therefore, the research question is what performancemeasurement solution or tool could be created so that military can demonstrate the compliance of GPS PPS towards PerformanceBased Navigation (PBN), while overcoming security constraints. The developed solution adheres to the existing performance requirements of regulated airspace on the basis of the verification of performance of military aircraft when relying on the GPS PPS receiver. The solution uses position data. A simulated automatic dependent surveillance-broadcast (ADS-B) trajectory for military aircraft is constructed by using a machine learning model that makes use of both correlated position reports for a flight (CPF) and ADS-B data from civil aircraft. The usage of CPF and ADS-B data from a number of civil flights arriving at London Stansted and Oslo Gardermoen airports allows the position enhancement of the CPF data. This is necessary since the latter relies on Mode S radars and a lower level of accuracy is expected compared to ADS-B. Such approach is key since the vast majority of military aircraft are not equipped with ADS-B and only CPF data about the flight is available. By applying machine learning to those datasets, the reconstruction of the accurate trajectory of military aircraft is generated for the approach and landing phase of the flight. This makes it possible to compare with the coded flight trajectory and assess the performance of the aircraft towards PBN accuracy requirements ultimately. The most accurate trajectory reconstruction results from a hybrid approach where Bi-Long short-term memory networks (Bi-LSTM) is used in the segments where the aircraft makes a turn, and a simple interpolation is used for the straight component of the flight path. The usability rate of the solution of around 90% for both Oslo and Stansted airports brings significant confidence for the operational application of the tool for the verification of the performance of GPS PPS-equipped military aircraft, regardless of the procedure design in place. As a minimum baseline, the benefit of the reduction of flight trials for the demonstration of compliance of GPS PPS can be achieved through the application of the tool to military flights, and continuous verification of the performance of the aircraft subsequently.