Bayesian network for order picker burden factor estimation considering individual differences

Recently, order picking, which is the most labour-intensive task in warehouse, has promoted the introduction of automated guided vehicles (AGVs). Most studies on human factors in order picking using AGVs have focused on the physical workload (PWL) of the worker, called the picker, but it is also important to consider mental workload (MWL) and individual differences. In this paper, we propose a Bayesian network (BN) considering individual differences for estimating the picker’s burden factor by conducting a replicated experiment of order picking with AGVs installed. BN is constructed from task data, heart rate variability indices, questionnaire data for PWL and MWL (the NASA task load index, Borg Scale CR10). To evaluate the performance of BN, the classification accuracy of Physical demand and Rate of Perceived Exertion are calculated and compared with three machine learning (random forest. neural network, support vector machine). The results indicate that the classification accuracy of BN is comparable to or higher than that of the comparison methods. Additionally, the BN demonstrates a highly interpretable and flexible model for uncertain data, such as HRV indicators and questionnaire data with individual differences.