Modeling Human Multichannel Perception and Control Using Linear Time-Invariant Models

Abstract

This paper introduces a two-step identification method of human multichannel perception and control. In the first step, frequency response functions are identified using linear time-invariant models. The analytical predictions of bias and variance in the estimated frequency response functions are validated using Monte Carlo simulations of a closed-loop control task and contrasted to a conventional method using Fourier coefficients. For both methods, the analytical predictions are reliable, but the linear time-invariant method has lower bias and variance than Fourier coefficients. It is further shown that the linear time-invariant method is more robust to higher levels of pilot remnant. Finally, both methods were successfully applied to experimental data from closed-loop control tasks with pilots.