A survey of human pilot models for study of Pilot-Induced Oscillation (PIO) in longitudinal aircraft motion

Abstract

AbstractPilot-Induced Oscillation (PIO), although an old issue, still poses a significant threat to aviation safety. The introduction of new systems in modern aircraft modifies the human–machine interaction and makes it necessary for research to revisit the subject from time to time. Given the need of aircraft manufacturers to constantly perform PIO tests, this study analysed the feasibility of using three different computational pilot models (Tustin, Crossover and Precision) to simulate PIO conditions. Three aircraft models with different levels of propensity to PIO (original, low propensity and high propensity) were tested, as well as two pilot gain conditions (normal and high). Data were collected for a purely longitudinal synthetic task through simulations conducted in MATLAB®. PIO conditions were detect using a tuned PIO detection algorithm (ROVER). Data were analysed in terms of both whether the pilot models triggered a PIO condition and for how long the condition was sustained. The results indicated that the three pilot models only provoked PIO conditions when high gain inputs were applied. Additionally, Crossover was the only pilot model to trigger a PIO for the three aircraft models. There were also significant differences between the pilot models in the total PIO time, as the Tustin model typically sustained the oscillatory condition for longer.