| dc.description.abstract | An alerting system is automation designed to reduce the likelihood of undesirable
outcomes that are due to rare failures in a human-controlled system. It accomplishes this
by monitoring the system, and issuing warning messages to the human operators when
thought necessary to head off a problem. On examination of existing and recently
proposed logics for alerting it appears that few commonly accepted principles guide the
design process. Different logics intended to address the same hazards may take disparate
forms and emphasize different aspects of performance, because each reflects the intuitive
priorities of a different designer. Because performance must be satisfactory to all users
of an alerting system (implying a universal meaning of acceptable performance) and not
just one designer, a proposed logic often undergoes significant piecemeal modification
before gaining general acceptance. This report is an initial attempt to clarify the common
performance goals by which an alerting system is ultimately judged. A better
understanding of these goals will hopefully allow designers to reach the final logic in a
quicker, more direct and repeatable manner. As a case study, this report compares three
alerting logics for collision prevention during independent approaches to parallel
runways, and outlines a fourth alternative incorporating elements of the first three, but
satisfying stated requirements. | en |
