Pilot Skill Aware Fault Tolerant Flight Control

Current efforts in making flight controllers more intelligent fall in two broad categories: (i) modeling of the human pilot and (ii) machine learning. The former builds mathematical models while the latter uses human-generated data to teach a computer, but both are mostly intended to mimic the human pilot. These approaches are limited in scope because they only imitate, instead of learning from the pilot and adjusting to the pilot’s needs.

This research effort is directed towards the design and development of a flight control architecture that enables the onboard computers to adjust the level of automation of the platform based on the strengths and weaknesses of the pilot along with mission demands. Furthermore, this new intelligent flight controller will be able to adjust in real time to system failures, making the platform a fault-tolerant UAS that works with the pilot.

Preliminary analysis indicates that certain trends or characteristics of pilots could be identified from an analysis of “pilot workload”, defined as a product of the amount of activity on the controls and the deviation from desired (= zero) attitude. Current literature on the characterization of pilot skills and abilities has shown that it is possible to set pilots apart based on flight and mission performance. As can be seen from Figures below, the preliminary results follow this trend, and measurable differences can be identified from flight data, even when the pilots were instructed to follow the same mission profile.

Publications

• 1. Belt, S., Gururajan, S., and Wu, X., “Evaluation of Workload and Performance during Primary Flight Training with Motion Cueing Seat in an Advanced Aviation Training Device,” SAE Int. J. Aerosp. 13(1):2020, doi:10.4271/01-13-01-0006.

  2. Garcia Lorca, F., Gururajan, S., Belt, S., "Parametric analysis of clustering algorithms for the generation of pilot performance profiles", Oral Presentation at 2017 SAE AeroTech Conference & Exhibition, Ft. Worth, TX, September 2017.

  3. Federico Garcia Lorca, Srikanth Gururajan, and Stephen Belt. "Characterization of Pilot Profiles Through Non-Parametric Classification of Flight Data", AIAA Information Systems-AIAA Infotech @ Aerospace, AIAA SciTech Forum, (AIAA 2017-0914). http://dx.doi.org/10.2514/6.2017-0914.