Abstract: Modeling and trajectory tracking control for a multi-rotor unmanned aerial vehicle (UAV) is studied. Nonlinear kinematics and dynamics models of the aircraft are established. A robust adaptive trajectory tracking control scheme based on fully tuned radial basis function neural network (FTRBFNN) and command
filtered backstepping is proposed for the multi-rotor aircraft. In the scheme, a position error PID controller of the aircraft is developed to eliminate deviation of the flight trajectory from the desired trajectory, and construct attitude angle commands for the attitude control loop. FTRBFNN is then used to estimate composite disturbance of the rotational dynamics. To avoid calculating pseudo control signal derivative analytically, and decrease dependence on the analytic model in the standard backstepping design, a command filtered backstepping technique is used to design the attitude controller. The technique uses a filter to calculate derivatives of the virtual control signal, instead of using analytical differentiation. It thus significantly simplifies
backstepping implementation and saves control energy. Correctness and effectiveness of the proposed robust adaptive trajectory tracking scheme are verified through simulation experiment.

Key words:  multi-rotor aircraft, robust adaptive, neural network (NN), command filter, backstepping