Abstract
With the introduction of a new regulatory framework enabling beyond-visual-line-of-sight (BVLOS) flights over populated areas, fixed-wing UAVs capable of long-distance flight are increasingly required for extended, wide-area observation missions, such as in disaster zones. Consequently, our research group is developing a flight control system based on the open-source software ArduPilot, targeting a cruising speed of 60 m/s. Proper tuning of the attitude controller is crucial for stable autonomous flight. However, the standard tuning assistance features in ArduPilot require a high level of operator skill and are difficult to execute during high-speed flight. To address this issue, we developed and verified a new tuning process using Virtual Reference Feedback Tuning (VRFT), a data-driven control approach. This paper reports on the results of that process.
