Behavior Based Task and High Workload Determination of Pilots Guiding Multiple UAVs

Abstract

Guiding multiple UAVs equipped with state-of-the-art automation by just one pilot usually means a large number and variety of supervision and monitoring tasks, interrupted by time-critical re-planning and reconfiguration tasks as reaction to unexpected events. To support pilots, especially in critical workload situations, by a cognitive associate system requires the machine awareness of the actual task the operator is working on, and the ability to detect critical workload situations to initiate assistant system interventions. This article describes an approach of building human operator behavior models to determine both, the current task of the operator, and derivations in task accomplishment, the latter observable by self-adapting strategies exposed during high workload conditions. Therefore, laboratory experiments were conducted utilizing a virtual flight simulator to stimulate pilot's workload during the guidance of multiple UAVs and to record their manual and visual interactions. These interactions represent the input data to train task specific operator behavior models by applying the Hidden Markov theory. Using Hidden Markov based behavior models allows the inference of tasks and their derivations from observable operator interactions. In this article, we describe the experimental findings, the methods applied, and the modelling approach.