Constrained Imitation Learning for a Flapping Wing Unmanned Aerial Vehicle

This letter presents a data-driven optimal control policy for micro flapping wing unmanned aerial vehicle. First, set of trajectories are computed off-line based on geometric formulation dynamics that captures the nonlinear coupling between large angle motion and quasi-steady aerodynamics. Then, it is transformed into feedback system according to framework imitation learning. In particular, an additional constraint incorporated through learning process enhance stability properties resulting controlled dynamics. Compared with conventional methods, proposed constrained eliminates need generate on-line, without sacrificing stability. As such, computational efficiency substantially improved. Furthermore, this establishes first stabilizes coupled longitudinal lateral vehicle relying averaging or linearization. These illustrated by numerical examples simulated model inspired Monarch butterflies.