AN ADAPTIVE PATTERN RECOGNIZING MODEL OF THE HUMAN OPERATOR ENGAGED IN A TIME VARYING CONTROL TASK.
PURDUE UNIV LAFAYETTE IND
Pagination or Media Count:
A model is presented in this study which describes the input-output response of a human operator engaged in the task of manually controlling a certain class of plants. The problem is approached from an engineering point of view in which the human operator is viewed as an adaptive controller performing the functions of identification, decision, and modification while manually controlling the plant. The emphasis in this study is not to merely reproduce the human operator response by arbitrarily constructing an adaptive controller and then minimizing the differences in the responses, but to incorporate the known physiological structure of the human in the model and synthesize an adaptive controller that identifies, decides and modifies in a manner that resembles the human operator as closely as possible. The control task treated is a one-dimensional compensatory visual-manual tracking task which visually displays the error to the subject and a control mechanism is provided to allow error compensation. From experimental tests it was found that the human operator uses pattern recognition to classify a limited variety of plants into one of three categories. The nature of the recognition problem is investigated and the decision surfaces associated with the pattern recognition process measured. A pattern recognition process was included in the model to simulate this part of the plant identification process. Author