Modelling human variability in DES using a modified multivariate accommodation technique

Current discrete-event simulations (DES) models of human performance often assume human tasks are independent and identically distributed. However, literature suggests that individual task times between tasks requiring similar human resources are correlated. Thus, this incorrect assumption may produce models that under predict the variability in human and total system performance. This research discusses proposed modifications of the boundary zone method (BZM), a multivariate accommodation technique developed for anthropometrics that can be combined with multi-stage simulation techniques to overcome this limitation of DES human performance modelling (HPM). A modified BZM algorithm tailored for HPM is applied to multiple datasets and shown to characterise the variability of the performance envelope for individuals with differing performance. The modified human performance BZM (HPBZM) algorithm was effective in analysing data from three large human performance datasets. In a representative dataset, the algorithm was able to generate 23 representative boundary cases that characterise 86.5% of a highly correlated dataset containing data from 901 individuals while retaining information critical to modelling performance variability in DES.