Learning-based Path Tracking Control of a Flapping-wing Micro Air Vehicle

Abstract

Flapping-wing micro air vehicles (FWMAVs) become promising research platforms due to their advantages

such as various maneuverability, and concealment. However, periodic and unsteady airflows generated

by flapping-wing motion make their dynamics time-varying and highly non-linear. Therefore, it is

difficult to apply model-based control techiques and even if applied, performance is not so good. Consequently, in practical control problems, simple model-free controllers or rule-based controllers are widely

used, but they are impractical to track diverse flight trajectories and require many trials and errors for gaintunings. In this paper, we suggest a model-based control strategy for FWMAV using learning architecture.

For this task, we construct a ground station for logging flight data and control inputs, and train dynamics

with a neural network. Then, we apply model predictive control (MPC) to the trained model. We validate

our method by hardware experiments and compare with other control methods.