Manned System Design Using Operator Measures and Criteria 1

Abstract

In existing man-machine systems, the human operator is typically presented a predesigned display and he must adapt his control techniques in an attempt to optimize overall system performance. Ultimate performance achieved may not rise to the designer's expectation or to the operator's capability because the design process does not account for the operator's adaptability. In this research program, ship control performance of the Officer of the Deck (OOD) was observed and analyzed in a series of simulator experiments involving ship transit and obstacle avoidance. Three types of displays were included in the analysis. OOD control rules, and measures and criteria which describe the control technique used, as well as the criteria employed in controlling the ship, were derived from the individual performance data. The effect of display type on the measures and criteria which were apparently optimized by the subjects was determined.

CONCLUSIONS: Analysis results show that performance differences between OLD and PACS, and RVV and PACS displays are significant at the α = 0.10 level. Summary measures which reveal this performance difference are time-to-CPA and distance travelled. Results confirm that time-to-CPA is an important measure since with it significant differences in performance can be shown. Analysis results also reveal the nature of the performance differences. From these results it is concluded not only that there are significant differences in performance with different displays, but also that performance measures which detect responses leading to a critical condition are more sensitive than summary measures of the critical conditions themselves.

Analysis results reveal that models of human operators controlling ships must include a purview factor. The purview factor used here was a radius within which a contact is considered. The model ignores contacts outside that radius. The purview function is required to represent all performances, i.e., without regard to display type.

The effect of OMAC parameter values on the comparison of optimal control to the subject control data obtained in the experiment. Results show that performance is at least a function of transit time and CPA/TCPA weighting as well as purview. Results suggests that for the contact density of the experiment problem, PACS and RVV can be rated as at least 12.5 mile displays and that OLD is a 10 mile display.

Consider now the primary purpose of the research as identified by the four questions presented in the introduction which concern methods of modeling the human operator. A method of computing the apparent OMAC from a set of candidate OMAC, has been demonstrated. The apparent OMAC has the parameter values of those analyzed that provide the best comparison to the experimental data. Also demonstrated is a method for using the OMAC to predict operator control actions in specific problem situations. The method uses an optimal control synthesis employing the representative OMAC as the criteria.

It is important to know whether or not a display configuration can affect the apparent OMAC. The results described here show that performance with different displays is significantly different and also, the results support the hypothesis that purview changes with display type.