Content area
Abstract
As safety aspects on systems reliability becomes more apparent, our ability to predict human operator's performance for control tasks in human-machine interactions is increasingly important. Unlike a machine's performance, a human operator's performance is quite dependent upon not only the design of the interaction systems, but also characteristics of human behavior. However, the performance of human operators is quite unpredictable and nonprobabilistic, especially in abnormal situations. Due to the complexity of the effects of the human operators' characteristics on their performance in abnormal situations, the probabilistic analysis of human performance has resulted in a great amount of uncertainty in the prediction of human reliability with interactive systems.
In this study, the effects of the human operators' characteristics on control tasks were analyzed in two control task domains by using a viewpoint of fuzzy theory and a hierarchy structure of task complexity. The two control task groups observed were: operators in the nuclear power plant control room and motor vehicle drivers, and their ability to avoid accidents. For the crews' control tasks in the control room, the application of a possibility distribution to the fuzzy characteristics of human operators showed narrower uncertainty intervals in the prediction of operator task reliability than did the application of probability distribution. In this study, another fuzzy regression method was proposed to reduce the error of estimation in the fuzzy regression analysis.
For driving tasks, a hierarchical structure of human behavior was hypothesized on the basis of task complexity. A path diagram was constructed for the hierarchical structure of human information processing abilities and accident rates. The path analysis showed less uncertainty in the effects of basic information processing abilities on accident rate, compared to the results from conventional correlation analysis.
For the operator's control tasks in interactive systems, fuzzification of human behavior or consideration of hierarchical structure of task complexity will be more effective in the prediction of operator's performance in abnormal situations.
