Neuro-adaptive Augmented Dynamic Inversion Controller for Quadrotors

Abstract

Abstract In this paper, an adaptive nonlinear control design for quadrotors is presented using nonlinear dynamic inversion and model-following neuro-adaptive techniques. The baseline controller is designed using dynamic inversion approach that exploits the time scale separation principle. The design works well if the model is perfectly known. However, the quadrotor system can have uncertainty in parameters. To tackle this, a neuro-adaptive controller is augmented with the baseline controller. The approach uses a single layer neural network to learn unknown dynamics and an adaptive law is employed to ensure that the quadrotor behave in the desired manner. A Lyapunov approach is used to show that the approximated dynamics remains bounded.