Identification of Linear Models of a Tandem-Wing Quadplane Drone: Preliminary Results

Abstract

A tandem-wing quadplane drone has been built to study control strategies and develop high-performance onboard controllers. In hover flight, the quadplane behaves like a classic quadcopter. Highly non-linear dynamics of the orientation stabilization need a state-of-the-art Model Predictive Controller (MPC). To develop such a controller, an accurate model of the drone needs to be identified – ideally, a linear model. This paper present preliminary results of identifying two linear models: a State-Space Model derived from Newton dynamic principles and a novel Recurrent Neural Network based linear model.