Adaptive Optimal Control for the Altitude of the Quadcopter UAV

In this paper, we study the adaptive optimal altitude control problem of the quadrotor UAV. The dynamics of the quadcopter’s altitude is modeled as a linear integrator model with nonlinear model uncertainties. Different from traditional design method which requires the complete knowledge of system model, we design a novel data-driven altitude controller based on the approximate dynamic programming. The controller is designed as the combination of a baseline controller and a virtual controller, where the virtual controller is to be learned. Then we use an off-policy iteration algorithm to learn the optimal controller. Simulation results show that the learned control policy outperforms the initial control policy, which validates our method.