Computational modeling to predict pilot's expectation of the aircraft state given vestibular and visual cues

Abstract

Loss of Control (LOC) accidents are a major threat for aviation, and contribute the highest risk for fatalities in all aviation accidents. The major contributor to LOC accidents is pilot spatial disorientation (SD), which accounts for roughly 32% of all LOC accidents. A pilot experiences SD during flight when he/she fails to sense correctly the motion, and/or attitude of the aircraft. In essence, the pilot's expectation of the aircraft's state deviates from reality. This deviation results from a number of underlying mechanisms of SD, such as distraction, failure to monitor flight instruments, and vestibular illusions. Previous researchers have developed computational models to understand those mechanisms. However, the models are limited in scope, as they do not model pilot expertise and have a small span of flight regimes to test with. This research proposes a new pilot model to predict the best-possible-pilot-expectation of the aircraft state given vestibular and visual cues. The proposed pilot model is in the form of a model-based observer (MBO), which provides the infrastructure needed to establish an expert pilot model. Experts are known to form an internal model of the operated system due to training/experience, which allows the expert to generate internal expectations of the system states. Pilot's internal expectations are enhanced by the presence of information fed through the pilot's sensory systems. The proposed pilot model integrates a continuous vestibular sensory model and a discrete visual-sampling sensory model to take account for the influence of the pilot's sensory system on his/her expectation of the aircraft state. The computational model serves to investigate the underlying mechanisms of SD during flight and provide a quantitative analysis tool to support flight deck and countermeasure designs.