Flight research has shown that adaptive flight control systems can be susceptible to adverse pilot-controller interactions, including pilot-in-the-loop oscillations (PIO). Conventional PIO analysis is performed using a static pilot model and a linear, time-invariant model of the aircraft and its control system. For aircraft with time-varying dynamics such as damage or failure, pilot technique may change rapidly to maintain control. An adaptive pilot model can be used to evaluate potential interactions between the pilot and an adaptive flight controller.
An ad hoc algorithm was designed for the real-time adaptation of a pilot model describing a pitch angle gross acquisition task. The pilot model consists of an adaptive gain and a fixed time delay. The adaptive gain is adjusted based on the magnitude of the tracking error and the aggressiveness of the response.
The algorithm was shown to match human pilot adaptation (simulation) to changes in dynamics including nominal, increased, and decreased loop gain; increased and decreased damping; and increased and decreased natural frequency. The algorithm also exhibits the classic pilot crossover theory response of –20dB/dec@0dB.
Unique features of the pilot model are:
- Continuous adaptations in which gain adjustments are updated continuously and naturally converge to a solution when the aircraft dynamics are time-invariant.
- Bidirectional adaptation in which the pilot gain increases or decreases as required.
- Explicit modeling of the pilot’s ability to phase-correlate command/response.
- Adaptation is normalized to be proportional to the piloting task.
This work was done by Curt Hanson of Armstrong Flight Research Center. For more information, contact the Armstrong Technology Transfer Office at 661-276-3967. DRC-012-007