HARVIS: dynamic rerouting assistant using deep learning techniques for Single Pilot Operations (SPO)

C Bejarano , A L Rodríguez Vázquez , A Colomer (1) , J Cantero (1) , A Ferreira (2) , L Moens (2) , A Duchevet (3) , J-P Imbert (3) , T de La Hogue (3)

1 UV - Universitat de València
2 DeepBlue
3 ENAC - Ecole Nationale de l'Aviation Civile

When a diversion due to an emergency is required in flight, and especially in SOP operations during the approach and landing phases, the pilot's workload increases considerably. To help gathering the relevant information and calculating the new alternative flight paths, we developed a digital assistant for dynamic rerouting, which uses deep learning algorithms to calculate the new trajectories for landing based on data collected from previous flights and METAR information. This AI assistant would be capable of improving pilots' decision-making when a rerouting is required. The rerouting assistant HMI prototype was developed and validated in feedback sessions with pilots, Advisory Board workshops with several stakeholders and a human-in-the-loop with 8 professional pilots at ENAC A320 research simulator. The concept of the assistant was appraised positively, although an effort is still to be done to improve the overall interaction and interface with the assistant to obtain some benefits in terms of workload. Aspects such as workload, usability, trust and explainability and training are analysed and discussed in the results.

Mots clés: Artificial intelligence; deep learning; dynamic rerouting; pilot decision support tool; human in the loop simulation Artificial intelligence; human in the loop simulation

Domaines: Informatique [cs]

C Bejarano, A L Rodríguez Vázquez, A Colomer, J Cantero, A Ferreira, et al.. HARVIS: dynamic rerouting assistant using deep learning techniques for Single Pilot Operations (SPO). 34th Conference of the European Association for Aviation Psychology, Sep 2022, Gibraltar, Gibraltar. pp.262 - 269, ⟨10.1016/j.trpro.2022.12.026⟩. ⟨hal-04017389⟩