Backstepping actor-critic optimal control for full-pose tracking of tilt-rotor quadcopters

Tilt-rotor quadcopters enable flight under non-zero attitudes, offering enhanced maneuverability for complex aerial tasks. However, despite their actuation redundancy, the translational and rotational dynamics remain partially coupled, posing challenges for accurate full-pose tracking. This paper proposes a backstepping actor-critic optimal control framework for an 8-input tilt-rotor quadcopter. A backstepping controller is developed to guarantee closed-loop stability, while a control strategy is designed to mitigate coupling effects and achieve effective decoupling between position and attitude. An actor-critic reinforcement learning mechanism is further incorporated to approximate the optimal control policy and improve tracking performance. Simulation and experimental results demonstrate accurate tracking, effective disturbance rejection, and robustness to parameter variations under non-zero attitude conditions.