Reinforcement Learning Control for 6 DOF Flight of Fixed-Wing Aircraft

Flight control is a key technique for the autonomous unmanned aircraft. The traditional model-based controller design approaches often aim at concrete plant and are short in universality. Reinforcement learning provides a general controller design paradigm that is adaptive, optimized, model-free and widely applicable, and it is a promising way for the intelligent control. In contrast to the 3 Degree-of-freedom (DOF) flight, the 6 DOF motion better describes the aircraft real flight, while the implementation of the intelligent control is much harder. Based on the multiple continuous states input and multiple continuous action output deep reinforcement learning, the integrated intelligent control for the cruise flight, directly from the vehicle flight states to the aero-surfaces and thrust control, is developed for the full-sized fixed-wing aircraft. It avoids the artificial trajectory and attitude loop separation in the traditional controller design, and is advantageous to the exploitation of the aerodynamics and the nonlinear inertia coupling.