Reducing Errors in Automated Cockpits Through Cognitive Engineering Principles

Abstract

Most navigational errors in automated aircraft are the result of incorrect input to the aircraft's autoflight system. The complaint is heard that the pilot is not “in the loop.” These concerns can be traced directly to the information available to the pilot. If we evaluate the task of navigating a highly-automated aircraft from a cognitive engineering perspective, we can design flight deck procedures that minimize the navigational errors and increase the detection of those errors before they result in an undesirable outcome. This paper evaluates what control action the pilot takes, what information is needed to verify the correctness of the control input, and what can be done to get the needed information when it is needed. Actual incidents of navigational errors resulting from improper input are reviewed. It is argued that both pilots should verify the input to the flight management system at the time of input. This is the only reliabl means of detecting input errors. Procedures are recommended for pilot confirmation at the time of input that take advantage of human cognition to reduce the probability that input errors will go undetected.