Abstract :
A systematic method is proposed for the development, optimization, and comparison of human controller models. The method is suitable for any model, including multiparameter systems. The evaluation criteria for assessing model quality are based on three separate components: 1) the cost or criterion function, 2) the comparison between the input/output functions of the human operator and those of the model, and 3) characteristic values and functions of statistical signal theory (mean values, auto- and crosscorrelation functions, power spectral density functions, and histograms of time function data). A nonlinear multiparameter human operator model is presented which considers the complex input information rate in a single display. The nonlinear features of the model are brought about by a modified threshold element and a decision algorithm. A random search technique is used for parameter optimization. Different display content arrangements as well as various transfer functions of the controlled element are explained by different optimized parameter combinations. The comparison with the well-known quasi-linear describing function for the human operator shows a marked superiority of the nonlinear model.
