Recognizing Pilot State: Enabling Tailored In-Flight Assistance Through Machine Learning

Motivation: Moving towards the highly controversial single pilot cockpit, more and more automation capabilities are added to today’s airliners 1. However, to operate safely without a pilot monitoring, avionics systems in future cockpits will have to be able to intelligently assist the remaining pilot. One critical enabler for proper assistance is a reliable classification of the pilot’s state, both in normal conditions and more critically in abnormal situations like an equipment failure. Only with a good assessment of the pilot’s state, the cockpit can adapt to the pilot’s current needs, i.e. alert, adapt displays, take over tasks, monitor procedures, etc. [2]