12 May 2016 Using machine learning and real-time workload assessment in a high-fidelity UAV simulation environment
Author Affiliations +
Abstract
Future unmanned vehicle operations will see more responsibilities distributed among fewer pilots. Current systems typically involve a small team of operators maintaining control over a single aerial platform, but this arrangement results in a suboptimal configuration of operator resources to system demands. Rather than devoting the full-time attention of several operators to a single UAV, the goal should be to distribute the attention of several operators across several UAVs as needed. Under a distributed-responsibility system, operator task load would be continuously monitored, with new tasks assigned based on system needs and operator capabilities. The current paper sought to identify a set of metrics that could be used to assess workload unobtrusively and in near real-time to inform a dynamic tasking algorithm. To this end, we put 20 participants through a variable-difficulty multiple UAV management simulation. We identified a subset of candidate metrics from a larger pool of pupillary and behavioral measures. We then used these metrics as features in a machine learning algorithm to predict workload condition every 60 seconds. This procedure produced an overall classification accuracy of 78%. An automated tasker sensitive to fluctuations in operator workload could be used to efficiently delegate tasks for teams of UAV operators.
© (2016) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Samuel S. Monfort, Samuel S. Monfort, Ciara M. Sibley, Ciara M. Sibley, Joseph T. Coyne, Joseph T. Coyne, "Using machine learning and real-time workload assessment in a high-fidelity UAV simulation environment", Proc. SPIE 9851, Next-Generation Analyst IV, 98510B (12 May 2016); doi: 10.1117/12.2219703; https://doi.org/10.1117/12.2219703
PROCEEDINGS
10 PAGES


SHARE
Back to Top