The Development of an Accurate Internal Model for High Order Systems

Abstract

In order to effectively control or monitor a complex system, the operator must construct an internal model of the system dynamics. An operator's internal model is defined as the internal or conceptual representation of how the operator thinks the system should or will work. The purpose of the research reported in this presentation is to: 1) determine methods to identify the form of the operator's internal model; 2) develop guidelines for system display design and training so that costly errors can be avoided; and 3) determine system display design and the type of training needed so that an accurate internal model is developed. To experimentally study the development of an operator's internal model, a dynamic high order control system was modeled and displayed to subjects over a CRT screen. Subjects were required to control the system through inputs on a joystick. The dynamics of this high order system were described through a set of differential equations. Without some kind of assistance, the solution to a set of differential equations is difficult to conceptualize, and, consequently, it is difficult for an operator to form an accurate internal model of the system dynamics. To help the operator internalize the system dynamics, upon control input the solutions to the differential equations were presented spatially on the CRT through computer augmentation. In the experiments reported, subjects were trained under two conditions: 1) with computer augmentation (the spatial representation of the solution); and 2) without computer augmentation. After training (about five hours per subject) performance was measured in terms of performance accuracy and mental workload needed to perform the task. The differences between the two groups in their ability to internalize the differential equation solutions was examined by providing the operator with a desired state of the system at a specified time and then testing the operator to see if he could choose the correct solution. Results indicated that with the right kind of training and the correct computer-controlled display augmentation, operators could accurately internalize the complex relationships needed to control this high order system.