NeuroTeaming: Using Power Spectral Density for Adjusting Teaming Dynamics in Pilot-AI Task Allocation

Paul, Tanya; Lafond, Daniel; Marois, Alexandre · 2024 · Crossref

DOI: 10.54941/ahfe1004735

archive: archived pipeline: cataloged

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of optimizing task allocation in human-autonomy teaming (HAT), specifically between human operators and autonomous unmanned aerial vehicles (UAVs). The research investigates whether electroencephalogram (EEG) power spectral density (PSD) can serve as a neurophysiological indicator of task-specific capacities, thereby enabling adaptive adjustments to teaming dynamics within a coactive design framework. The motivation stems from the need to prevent maladaptive automation use, such as over-reliance or skills degradation, by transparently identifying which agent is best suited for specific tasks based on complementary capabilities. The experiment involved ten participants (4 women, 6 men) who performed a simulated semi-autonomous UAV search and rescue mission using a PlayStation 3 controller in an Unreal Engine environment. Participants were randomly assigned to one of two groups: Group 0 (G0), trained on target identification and optimal path finding, or Group 1 (G1), trained on obstacle avoidance. Each group underwent four training sessions followed by a test mission. EEG data was collected via Conscious Lab SUPRA headphones, capturing signals from 16 electrodes across frontal, central, parietal, occipital, and temporal regions. PSD was extracted using a 2000 ms window, focusing on five frequency bands: delta (1–4 Hz), theta (4–7 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–45 Hz). Performance metrics included collision counts, targets reached, and time spent outside the area of interest. Results indicated that G1 (obstacle avoidance) demonstrated higher task capacities and lower dependency on automated support compared to G0. Statistically significant differences in performance were observed across all subtasks (Mann-Whitney test, p < 0.005). EEG analysis revealed distinct activation patterns between groups; G1 exhibited higher PSD levels across all five frequency bands compared to G0. The most prominent neural differences were localized in the frontal and central brain regions, with significant variations in alpha, beta, and gamma bands. Specifically, G1 showed higher activation in frontal electrodes (FP2, F4) and central electrodes (C2), while G0 displayed lower variability and activation peaks in different channels. No significant differences were found in the occipital lobe. These findings suggest that EEG PSD bands can effectively differentiate between skill-based capacities in UAV operations, providing a basis for developing adaptive, neurophysiological-driven task-allocation systems. The distinct neural signatures associated with specific training domains imply that brain-computer interfaces can monitor operator states to dynamically adjust automated support, enhancing transparency and explainability in HAT. Future research should explore implementing these features in closed-loop systems for real-time adaptive autonomy in manned-unmanned teaming.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.8-27b-gittensor on 2026-08-23; verification: pending re-verification.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• neuro workload indices