Cognitive Load-Aware Virtual Reality Training Platform for Safe Drone Operations in Construction

Yuming Zhang; Houtan Jebelli

Abstract:

Drones have become indispensable in construction projects, performing tasks such as site surveying, 3D mapping, progress monitoring, and inspecting hazardous or hard-to-reach areas. Their integration has significantly enhanced efficiency, precision, and safety. However, effective human-drone interaction remains challenging due to the complexity of construction environments and the necessity for precise drone control, which requires workers to undergo extensive training. Traditional training programs primarily focus on task-specific skills, often neglecting the influence of workers' cognitive and physiological states on learning effectiveness and performance. This oversight can lead to increased cognitive fatigue, reduced learning outcomes, and heightened safety risks. To address these limitations, this study introduces a Virtual Reality (VR)-based adaptive training platform designed to optimize worker-drone interactions by monitoring cognitive load through physiological indicators such as heart rate and skin conductance. Utilizing a Long Short-Term Memory (LSTM) model, the platform accurately predicts cognitive load in real time and dynamically adjusts the training pace to align with individual cognitive states. In experimental evaluations, the adaptive training platform achieved a cognitive load prediction accuracy of 92% and reduced task completion time by 69.9%, outperforming non-adaptive training methods by 27.6% in worker-drone interaction scenarios. By personalizing the training experience based on real-time cognitive assessments, the platform enhances worker preparedness, promoting safer and more efficient drone operations on construction sites. This research underscores the critical role of integrating physiological data into adaptive training systems to meet the multifaceted demands of modern construction environments.