Unmanned Aerial Vehicles Control Study Using Deep Deterministic Policy Gradient

In order to reduce the dependence of controller design on accurate dynamic models of controlled objects and improve the self-learning ability of control systems, a model-free control algorithm of Unmanned Aerial Vehicles (UAVs) is presented. The controller proposed in this paper is based on deep deterministic policy gradient (DDPG) algorithm, and it is able to control UAV to a specified position and attitude without knowing the UAV's dynamic model. Furthermore, the controller is trained by flight data to obtain the learning ability. Simulation results demonstrate that the convergence speed of learning is fast, and the control effect can meet requirements of UAV's position and attitude precision when the UAV's dynamic model is unknown. Besides, the controller can quickly adapt to different environments.