Neural adaptive back stepping flight controller for a ducted fan UAV

Abstract

In this paper, we present a neural adaptive back-stepping flight controller for a ducted fan UAV whose dynamics is characterized by uncertainties and highly coupled nonlinearities. The proposed neural adaptive back-stepping controller can handle unknown nonlinearities, unmodeled dynamics and external wind disturbances. A single layer radial basis function network is used to approximate the virtual control law derived using back stepping approach, which provides necessary stability and tracking performances. The neural controller parameters are adapted online using Lyapunov based update laws. The proposed controller is evaluated using nonlinear desktop simulation model of a typical ducted fan UAV performing bop-up maneuver. Three neural adaptive controllers are implemented to handle attitude command altitude hold system, one in each body axis. A separate neural controller is implemented to track the height command for autonomous takeoff and landing. The results indicate that the proposed controller can stabilize the ducted fan UAV and provide necessary tracking performance.