Differentiating Workload using Pilot's Stick Input in a Virtual Reality Flight Task (2309.09619v2)

Abstract: High-risk operational tasks such as those in aviation require training environments that are realistic and capable of inducing high levels of workload. Virtual Reality (VR) offers a simulated 3D environment for immersive, safe and valid training of pilots. An added advantage of such training environments is that they can be personalized to enhance learning, e.g., by adapting the simulation to the user's workload in real-time. The question remains how to reliably and robustly measure a pilot's workload during the training. In this study, six novice military pilots (average of 34.33 flight hours) conducted a speed change maneuver in a VR flight simulator. In half of the runs an auditory 2-back task was added as a secondary task. This led to trials of low and high workload which we compared using the pilot's control input in longitudinal (i.e., pitch) and lateral (i.e., roll) directions. We extracted Pilot Inceptor Workload (PIW) from the stick data and conducted a binary logistic regression to determine whether PIW is predictive of task-induced workload. The results show that inputs on the stick along its longitudinal direction were predictive of workload (low vs. high) when performing a speed change maneuver. Given that PIW may be a task-specific measure, future work may consider (neuro)physiological predictors. Nonetheless, the current paper provides evidence that measuring PIW in a VR flight simulator yields real-time and non-invasive means to determine workload.