Assessment of Operator Productivity in Intelligent Systems when Solving Test Problems under Conditions of Uncertainty

Abstract

A cognitive approach to assessing the performance of operators under conditions of uncertainty to increase the efficiency of human-machine interfaces of intelligent systems is presented. A models for solving a sequence of tasks by an operator in the form of Markov chains, a modification of stochastic processes of Ornstein-Uhlenbeck and Vasicek are developed. Algorithms for identifying model parameters from experimental data are developed. The experimental data were obtained as a result of testing models of cognitive-style potential (CSP) of operators. Computer variants of methods for diagnosing the cognitive sphere are implemented. Using machine learning methods based on cognitive models, a system for the results predictions of operators’ work is implemented. To train operators in solving problems of object recognition, overcoming obstacles and pursuing a goal, a simulator with a subsystem for recording the time of execution of actions, errors and evaluation of the results of the mission has been developed. Examples of solving test tasks in a sonar monitoring system saturated with information of various modality are given.