Design of a Sensorized Safety Helmet for Stress and Cognitive Workload for Automated Warehouses Operators

In modern smart warehouses, human–machine collaboration remains essential to maintain flexibility in increasingly automated logis- tics operations. Despite the rise of automation, several tasks, such as order picking, still rely on human operators whose performance may be affected by mental fatigue, stress, or cognitive overload. Therefore, the monitoring of these conditions is essential, to ensure both operational ef- ficiency and worker safety. To address this challenge, a sensorized safety helmet was developed to detect early signs of reduced alertness and in- creased cognitive workload during manual picking activities within auto- mated warehouse environments. The system integrates multiple sensing modalities, including electrodermal activity (EDA) and an eye-tracking camera, all managed by an embedded processor. The prototype was vali- dated in a logistics laboratory replicating warehouse operations, where 23 participants performed picking tasks of increasing difficulty. A supervised learning framework was employed to investigate the predictive relation- ship between EDA and oculometric features with cognitive workload. Four classifiers were trained using NASA-TLX scores and self-reports as ground truth labels to categorize workload-related perceptual dimen- sions. Results show F1 scores above 72% for NASA-TLX subdomains and above 81% for emotional dimension. These values demonstrate that the integration of EDA and eye-related metrics effectively distinguishes between cognitive states, highlighting the potential of multi-feature sens- ing for adaptive safety systems that enhance both productivity and well- being in industrial contexts.