Modeling Pilot Cognitive Behavior for Predicting Performance and Workload Effects of Cockpit Automation

The objective of this research was to develop a model of pilot cognitive behavior to predict performance and workload while using varying degrees of cockpit automation to serve as a basis for future systems design. A cognitive task analysis (CTA) was conducted on expert pilot performance a flight control panel (FCP), control-display unit (CDU) and flight management system, and an enhanced CDU (CDU+) providing pre-programmed arrivals from air traffic control in a simulated landing and approach task. Cognitive models were developed from the CTA using an enhanced form of the GOMS language, including a set of additional task operators, to represent pilot actions on cockpit interfaces. Pilot performance and workload data from a parallel empirical study of the same flight tasks were used as a basis for validating the cognitive model output. Indices of automation complexity were formulated based on counts of task methods and steps, required chunks of information, and information transactions coded in the enhanced GOMS models. These indices revealed high complexity for the FCP mode and low complexity for the prototype CDU+ mode. The automation index values were positively and significantly correlated with pilot heart rate (as an objective measure of workload) and vertical path deviation error from the experimental data set. The computational cognitive models of pilot behavior in using forms of cockpit automation were demonstrated to be a viable tool for predicting pilot workload and flight performance under high workload flight conditions