Mental Workload Estimation Based on Physiological Features for Pilot-UAV Teaming Applications.

Area of Science:

• Neuroergonomics
• Human-Computer Interaction
• Artificial Intelligence

Background:

• Manned-Unmanned Teaming (MUM-T) involves human pilots cooperating with aerial robots.
• No prior studies have neuroergonomically assessed mental workload (MW) in MUM-T or used physiological signal classification.
• The impact of signal non-stationarity and validation methods on MW classification accuracy is understudied.

Purpose of the Study:

• To characterize and estimate MW in MUM-T using physiological signals.
• To evaluate the influence of different validation procedures on MW classification accuracy.

Main Methods:

• A search and rescue (S&R) scenario was simulated with 14 participants acting as pilots cooperating with three Unmanned Aerial Vehicles (UAVs).
• Missions induced high and low MW levels, measured via self-report, behavior, and physiological data (cerebral, cardiac, oculomotor).
• Supervised classification pipelines using physiological features were benchmarked, comparing traditional vs. ecological validation.

Main Results:

• MW significantly impacted all physiological measures.
• Intra-subject classification accuracy reached 75% using ECG, EEG, and ET features with traditional validation.
• Ecological validation significantly reduced intra-subject accuracy but improved inter-subject classification (59.8%) beyond other methods.

Conclusions:

• Physiological measures effectively reflect MW in MUM-T settings.
• Traditional validation overestimates classification accuracy; ecological validation is crucial for operational relevance.
• Further research is needed to enhance MW monitoring in real-world MUM-T contexts.