To model human operator (HO) dynamics in manual tracking tasks, an ensemble of models, each for a certain class of inputs, seems to be needed. By placing in a linear framework the modeling studies so far conducted, it is evident that different hypotheses have been proposed to explain the observed input dependence of the estimated HO (linear) models. Here, the authors examine these hypotheses and propose that the systemic notion of task dependence must be used to model this system. They have explored ways of deriving quantitative measures of the system task-dependent characteristics, using autoregressive moving-average (ARMA) models of input-output data obtained from a series of pursuit manual tracking experiments. These experiments utilized sum-of-sinusoids and random ternary inputs of various bandwidths. The resulting model parameters indicate significant task dependence of the HO dynamic characteristics. The effect of amplitude nonlinearities was examined and found to be statistically insignificant.
展开▼
