Assessment of the Accuracy of Simulation of the Oculo-Motor System Based on Volterra Models

Integral nonlinear models are used to create mathematical models of the human oculo-motor system (OMS). These models take into account both inertial and nonlinear properties of the objects under study. To obtain empirical data for model construction, experimental studies are conducted with OMS using «input-output» data. Visual stimuli are used as test signals, displayed on a computer monitor at various distances from the starting position, which formally corresponds to the action of step signals with varying amplitudes on the object of study. In this process, the responses of the OMS are recorded using innovative eye-tracking technology. Mathematical models in the form of Volterra series and polynomials are employed for computer modeling of the OMS. The aim of this research is to analyze the accuracy of OMS identification as multidimensional transient functions based on eye-tracking data, examining the dependency of computation errors for models of different orders on the amplitudes and quantities of the test signals used. The subject of the study includes various methods for identifying OMS, algorithms, and Python-based software tools for computing the dynamic characteristics of OMS using eye-tracking technology. The research explores identification methods: compensation, approximation, and least squares methods. The accuracy of the linear, quadratic, and cubic OMS models is evaluated. The most accurate models, constructed from real experimental data, are found to be quadratic or cubic OMS models obtained using the least squares method with three test signals.