Ergonomic Workload Assessment of Order Picking Operations Based on Machine Learning with sEMG Signals

The Industry 5.0 concept emphasizes human-centric manufacturing and logistics systems, encouraging decision-makers to consider human well-being in management operations alongside efficiency and cost. This study addresses the challenge of physiological load and health implications for order picking workers in warehousing systems, aiming at comprehensively assessing the ergonomic workload level of workers. We conduct a laboratory experiment and collect both subjective fatigue perception and objective physiological signals, especially surface electromyography (sEMG) signals of order pickers using Borg’s CR-10 scale and wearable physiological devices. Experimental data was processed into relevant features of basic physiology, task, and sEMG signals for order picking operations. We then developed six machine learning methods for model training and evaluation to achieve an accurate portrayal of ergonomic workload, and CatBoost achieved the best training performance. Further ablation analysis revealed that incorporating sEMG features allows for a more accurate description of ergonomic workload. This research provides a scientific approach for assessing ergonomic workload, which can benefit academics, managers, and decision-makers in developing human-centric order picking systems that balance efficiency with worker well-being, in line with the principles of Industry 5.0.