An adaptive system model of human precision grip

Abstract

Eight human test subjects attempted to make the position signal from an instrumented manipulandum match a pyramidical desired position signal displayed on a computer monitor. The eight test subjects used the manipulandum with three different levels of stiffness. A transfer function was developed to represent the human-manipulandum system from the data on the first step, when the test subjects initially displaced the manipulandum to learn the stiffness of the system. Another transfer function was formed from the data on the subsequent steps in the pyramidical trajectory, after significant displacement of the manipulandum occurred. Neither of these transfer functions accurately model the system dynamics for the duration of the experiments. For this reason, an adaptive system model is developed to account for the learning process of the human test subjects as they displaced the manipulandum to gain knowledge of the system stiffness. The adaptive system model was subsequently validated by comparison to the human test subject data. An examination of the average absolute error between the position predicted by the adaptive model and the actual experimental data yields an overall average error of 0.26mm for all three levels of stiffness. A statistical analysis shows that the adaptive model predicts a significantly lower amount of error than the other two transfer functions.