Tilbake til søkeresultatene

BIA-Brukerstyrt innovasjonsarena

From manual to AI-assisted airport safety reporting through human-AI teaming (FLAIT)

Alternativ tittel: Fra manuell til AI-assistert safety-rapportering på lufthavner gjennom menneske-AI-samarbeid (FLAIT)

Tildelt: kr 15,4 mill.

Prosjekteier er Airside Innovation AS som også er ansvarlig for å kommersialisere hovedinnovasjonen fra prosjektet. Prosjektpartnerne er Opscom Systems (Safety Management System), Avinor (Lufthavneier) og Widerøe Ground Handling (ground handling på lufthavn), sammen med to forskningsorganisasjoner, NORCE og SINTEF. SINTEF er prosjektleder. Prosjektet har også en rådgivende komite med Luftfartstilsynet, London Luton Lufthavn og Václav Havel Airport Praha. Prosjektet startet i august 2022 og vi har hatt en oppstartsmøte med alle prosjektpartnerne for å sette i gang prosjektaktivitetene.

The apron (where aircraft are parked at the gate) is one of the busiest areas on airports and up to 243 000 people are injured each year in apron-related accidents or incidents. Moreover, airlines suffer a cost of USD 10 billion due to aircraft ground damage. Unreported safety-related occurrences (i.e., events, incidents, accidents) poses the highest risk to flight safety. Although AI-based (Artificial Intelligence) "assistants" are widely used for applications such as accounting, etc., airport apron safety reporting involves mostly time-consuming and error prone manual procedures. We will bridge this gap and develop, demonstrate and verify a Decision Support System prototype - "the FLAIT Assistant" - for automatic classification and reporting of occurrences in the apron area through human-AI collaboration. The FLAIT Assistant and accompanying safety assurance procedures will be designed such that those involved in safety reporting can work with the Assistant as a human-AI team. Such teaming augments human capabilities and raises performance beyond that of either entity. To develop and deploy the FLAIT Assistant we need to address several research challenges including how to 1) enable all-weather object (humans, vehicles, equipment) detection and classification within a unified "bird's eye view" of the apron area, 2) realize an explainable spatio-temporal system for detection and classification of both routine operations and occurrences at the apron, 3) enable and verify evidence-based and trustworthy occurrence reporting that facilitates joint human-AI learning and handles potential mismatches between human and AI reasoning. Our "unfair advantage" for succeeding in developing and commercializing is our team's combination of in-depth knowledge of the airport sector, safety assurance, explainable AI, and experience with deploying AI in the airport sector - and that our team is composed of the value chain from supplier to end-user and a world-class advisory board.

Budsjettformål:

BIA-Brukerstyrt innovasjonsarena