XRHQ
Netherlands
Defense

Royal Netherlands Air Force trials brainwave-adaptive VR flight simulator training

Fifteen student pilots flew a VR PC-7 cockpit simulator while an EEG model set difficulty; ten preferred it, but flight performance was unchanged.

Jeremy Dalton

Added on
Royal Netherlands Air Force pilot training
Photo: RNAF

Sign in to view 4 more media items

Data Points (6)

This information is available to registered members of XRHQ. Sign up for free below.

The Royal Netherlands Air Force has trialled a prototype virtual reality flight-training system that reads a trainee pilot's brain activity and changes the difficulty of the next exercise accordingly. Pilots flew a VR simulator of the PC-7 Turbo Trainer built by the Dutch simulation firm multiSIM, with a physical cockpit mock-up and a control-loaded stick and rudder pedals, and a Varjo Aero headset for the out-the-window view. A wireless 32-channel EEG system (g.Nautilus PRO, g.tec medical engineering) recorded brain signals at 250 Hz. An external computer ran a pre-trained, subject-independent passive brain-computer interface in Python that classified each 120-second trial as high or low cognitive workload and, over an ethernet link, instructed the simulator to step the next trial one level down or up through five levels of degraded visibility: clear weather, mist, heavy fog, mist plus a failed artificial horizon, and heavy fog with a false horizon.

Fifteen Royal Netherlands Air Force student pilots (2 women, 13 men; mean age 24.5; mean 62.6 logged flight hours) each flew one task under the neuro-adaptive system and another under a conventional fixed sequence of increasing difficulty, without being told which was which. The researchers found no significant difference between the two conditions on flight performance, self-reported workload, engagement or simulator sickness; in both conditions performance fell as subjective workload rose. In post-session interviews, ten of the fifteen pilots said that once the system was explained to them they preferred the adaptive version, describing it as more realistic and less predictable. Six pilots showed almost no variation in detected workload across trials, indicating that the pre-trained classifier did not generalise to every individual.

The work was led by Dr Evy van Weelden of the Royal Netherlands Aerospace Centre (NLR) with colleagues at Tilburg University and Vrije Universiteit Amsterdam, and was funded through the MasterMinds project of the RegionDeal Mid- and West-Brabant, co-funded by the Dutch Ministry of Economic Affairs and the Municipality of Tilburg. It is a research prototype evaluated in a controlled experiment approved by Tilburg University's research ethics committee, not an operational training system. The simulator it ran on belongs to the air force's AIR (Ambitie, Innovatie en Resultaat) innovation unit, which had a stored PC-7 cockpit mock-up rebuilt and fitted with a multiSIM VR flight simulation for research into virtual reality in the elementary military pilot training course.

Quotes

If you're not in the field, it sounds very sci-fi. But for me, it's normal. I just see data.

Project Summary

This information is available to registered members of XRHQ. Sign up for free below.

Experience

Available to XRHQ members

Hardware

Available to XRHQ members

Software

Available to XRHQ members

Partners & Service Providers

Available to XRHQ members

Sources

Available to XRHQ members

Know more about this project? Help us fill in any missing details.

Edit this project →

SUBMITTED BY

Jeremy Dalton

Founder, XRHQ

This write-up is free to reuse with credit — CC BY 4.0.